wip: shader compilation benchmark

2024-01-16 17:03:41 -08:00
260 changed files with 2882 additions and 5656 deletions
@@ -10,7 +10,7 @@ on:
 jobs:
  build-android:
    name: build-android
-    runs-on: macos-14
+    runs-on: macos-latest

    steps:
      - uses: actions/checkout@v3.3.0
@@ -13,7 +13,7 @@ on:
 jobs:
  cocoapods-deploy:
    name: cocoapods-deploy
-    runs-on: macos-14
+    runs-on: macos-latest
    steps:
      - name: Check out iOS/CocoaPods directory
        uses: Bhacaz/checkout-files@49fc3050859046bf4f4873678d46099985640e89
@@ -10,7 +10,7 @@ on:
 jobs:
  build-ios:
    name: build-ios
-    runs-on: macos-14
+    runs-on: macos-latest

    steps:
      - uses: actions/checkout@v3.3.0
@@ -10,7 +10,7 @@ on:
 jobs:
  build-mac:
    name: build-mac
-    runs-on: macos-14
+    runs-on: macos-latest

    steps:
      - uses: actions/checkout@v3.3.0
@@ -11,7 +11,7 @@ on:
 jobs:
  npm-deploy:
    name: npm-deploy
-    runs-on: macos-14
+    runs-on: macos-latest
    steps:
      - uses: actions/checkout@v3.3.0
        with:
@@ -15,7 +15,7 @@ jobs:

    strategy:
      matrix:
-        os: [macos-14, ubuntu-22.04]
+        os: [macos-latest, ubuntu-22.04]

    steps:
      - uses: actions/checkout@v3.3.0
@@ -40,7 +40,7 @@ jobs:

  build-android:
    name: build-android
-    runs-on: macos-14
+    runs-on: macos-latest

    steps:
      - uses: actions/checkout@v3.3.0
@@ -54,7 +54,7 @@ jobs:

  build-ios:
    name: build-iOS
-    runs-on: macos-14
+    runs-on: macos-latest

    steps:
      - uses: actions/checkout@v3.3.0
@@ -67,7 +67,7 @@ jobs:

  build-web:
    name: build-web
-    runs-on: macos-14
+    runs-on: macos-latest

    steps:
      - uses: actions/checkout@v3.3.0
@@ -31,7 +31,7 @@ jobs:

    strategy:
      matrix:
-        os: [macos-14, ubuntu-22.04]
+        os: [macos-latest, ubuntu-22.04]

    steps:
      - name: Decide Git ref
@@ -65,7 +65,7 @@ jobs:

  build-web:
    name: build-web
-    runs-on: macos-14
+    runs-on: macos-latest
    if: github.event_name == 'release' || github.event.inputs.platform == 'web'

    steps:
@@ -98,7 +98,7 @@ jobs:

  build-android:
    name: build-android
-    runs-on: macos-14
+    runs-on: macos-latest
    if: github.event_name == 'release' || github.event.inputs.platform == 'android'

    steps:
@@ -129,7 +129,7 @@ jobs:
      - name: Sign sample-gltf-viewer
        run: |
          echo "${APK_KEYSTORE_BASE64}" > filament.jks.base64
-          base64 --decode -i filament.jks.base64 > filament.jks
+          base64 --decode filament.jks.base64 > filament.jks
          BUILD_TOOLS_VERSION=$(ls ${ANDROID_HOME}/build-tools | sort -V | tail -n 1)
          APKSIGNER=${ANDROID_HOME}/build-tools/${BUILD_TOOLS_VERSION}/apksigner
          IN_FILE="out/sample-gltf-viewer-release.apk"
@@ -152,7 +152,7 @@ jobs:

  build-ios:
    name: build-ios
-    runs-on: macos-14
+    runs-on: macos-latest
    if: github.event_name == 'release' || github.event.inputs.platform == 'ios'

    steps:
@@ -10,7 +10,7 @@ on:
 jobs:
  build-web:
    name: build-web
-    runs-on: macos-14
+    runs-on: macos-latest

    steps:
      - uses: actions/checkout@v3.3.0
@@ -634,9 +634,9 @@ function(combine_static_libs TARGET OUTPUT DEPS)
    # Loop through the dependent libraries and query their location on disk.
    set(DEPS_FILES )
    foreach(DEPENDENCY ${DEPS})
-        if (TARGET ${DEPENDENCY})
+        if(TARGET ${DEPENDENCY})
            get_property(dep_type TARGET ${DEPENDENCY} PROPERTY TYPE)
-            if (dep_type STREQUAL "STATIC_LIBRARY")
+            if(dep_type STREQUAL "STATIC_LIBRARY")
                list(APPEND DEPS_FILES "$<TARGET_FILE:${DEPENDENCY}>")
            endif()
        endif()
@@ -706,7 +706,7 @@ function(get_resgen_vars ARCHIVE_DIR ARCHIVE_NAME)
        set(RESGEN_OUTPUTS "${OUTPUTS}" PARENT_SCOPE)
        set(RESGEN_FLAGS -qx ${ARCHIVE_DIR} -p ${ARCHIVE_NAME} PARENT_SCOPE)
        set(RESGEN_SOURCE "${ARCHIVE_DIR}/${ARCHIVE_NAME}${ASM_SUFFIX}.S" PARENT_SCOPE)
-        set(RESGEN_SOURCE_FLAGS "-I'${ARCHIVE_DIR}' ${ASM_ARCH_FLAG}" PARENT_SCOPE)
+        set(RESGEN_SOURCE_FLAGS "-I${ARCHIVE_DIR} ${ASM_ARCH_FLAG}" PARENT_SCOPE)
    endif()
 endfunction()

@@ -721,6 +721,7 @@ add_subdirectory(${LIBRARIES}/filabridge)
 add_subdirectory(${LIBRARIES}/filaflat)
 add_subdirectory(${LIBRARIES}/filagui)
 add_subdirectory(${LIBRARIES}/filameshio)
+add_subdirectory(${LIBRARIES}/geometry)
 add_subdirectory(${LIBRARIES}/gltfio)
 add_subdirectory(${LIBRARIES}/ibl)
 add_subdirectory(${LIBRARIES}/iblprefilter)
@@ -748,9 +749,6 @@ add_subdirectory(${EXTERNAL}/jsmn/tnt)
 add_subdirectory(${EXTERNAL}/stb/tnt)
 add_subdirectory(${EXTERNAL}/getopt)

-# Note that this has to be placed after mikktspace in order for combine_static_libs to work.
-add_subdirectory(${LIBRARIES}/geometry)
-
 if (FILAMENT_BUILD_FILAMAT OR IS_HOST_PLATFORM)
    # spirv-tools must come before filamat, as filamat relies on the presence of the
    # spirv-tools_SOURCE_DIR variable.
@@ -7,3 +7,5 @@ for next branch cut* header.
 appropriate header in [RELEASE_NOTES.md](./RELEASE_NOTES.md).

 ## Release notes for next branch cut
+
+- Metal: fix some shader artifacts by disabling fast math optimizations.
@@ -31,7 +31,7 @@ repositories {
 }

 dependencies {
-    implementation 'com.google.android.filament:filament-android:1.50.5'
+    implementation 'com.google.android.filament:filament-android:1.49.3'
 }
 ```

@@ -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', '~> 1.50.5'
+pod 'Filament', '~> 1.49.3'
 ```

 ### Snapshots
@@ -7,28 +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).

-## v1.50.5
-
- android: NDK 26.1.10909125 is used by default
- android: Minimum API level on Android is now API 21 instead of API 19. This allows the use of OpenGL ES 3.1
- rendering: New PBR Neutral tone mapper, designed to preserve materials color appearance
- android: Change default frameRateOptions.interval to 1.0
-
-## v1.50.4
-
-
-## v1.50.3
-
-
-## v1.50.2
-
-
-## v1.50.1
-
- Metal: fix some shader artifacts by disabling fast math optimizations.
- backend: remove `atan2` overload which had a typo and wasn't useful. Fixes b/320856413.
- utils: remove usages of `SpinLock`. Fixes b/321101014.
-
 ## v1.50.0
 - engine: TAA now supports 4x upscaling [BETA] [⚠️ **New Material Version**]

@@ -80,7 +80,7 @@ buildscript {

    ext.versions = [
        'jdk': 17,
-        'minSdk': 21,
+        'minSdk': 19,
        'targetSdk': 34,
        'compileSdk': 34,
        'kotlin': '1.9.21',
@@ -484,10 +484,7 @@ extern "C" JNIEXPORT void JNICALL Java_com_google_android_filament_Engine_nSetBu
 extern "C" JNIEXPORT void JNICALL Java_com_google_android_filament_Engine_nSetBuilderConfig(JNIEnv*,
        jclass, jlong nativeBuilder, jlong commandBufferSizeMB, jlong perRenderPassArenaSizeMB,
        jlong driverHandleArenaSizeMB, jlong minCommandBufferSizeMB, jlong perFrameCommandsSizeMB,
-        jlong jobSystemThreadCount,
-        jlong textureUseAfterFreePoolSize, jboolean disableParallelShaderCompile,
-        jint stereoscopicType, jlong stereoscopicEyeCount,
-        jlong resourceAllocatorCacheSizeMB, jlong resourceAllocatorCacheMaxAge) {
+        jlong jobSystemThreadCount, jlong stereoscopicEyeCount) {
    Engine::Builder* builder = (Engine::Builder*) nativeBuilder;
    Engine::Config config = {
            .commandBufferSizeMB = (uint32_t) commandBufferSizeMB,
@@ -496,12 +493,7 @@ extern "C" JNIEXPORT void JNICALL Java_com_google_android_filament_Engine_nSetBu
            .minCommandBufferSizeMB = (uint32_t) minCommandBufferSizeMB,
            .perFrameCommandsSizeMB = (uint32_t) perFrameCommandsSizeMB,
            .jobSystemThreadCount = (uint32_t) jobSystemThreadCount,
-            .textureUseAfterFreePoolSize = (uint32_t) textureUseAfterFreePoolSize,
-            .disableParallelShaderCompile = (bool) disableParallelShaderCompile,
-            .stereoscopicType = (Engine::StereoscopicType) stereoscopicType,
            .stereoscopicEyeCount = (uint8_t) stereoscopicEyeCount,
-            .resourceAllocatorCacheSizeMB = (uint32_t) resourceAllocatorCacheSizeMB,
-            .resourceAllocatorCacheMaxAge = (uint8_t) resourceAllocatorCacheMaxAge,
    };
    builder->config(&config);
 }
@@ -47,11 +47,6 @@ Java_com_google_android_filament_ToneMapper_nCreateFilmicToneMapper(JNIEnv*, jcl
    return (jlong) new FilmicToneMapper();
 }

-extern "C" JNIEXPORT jlong JNICALL
-Java_com_google_android_filament_ToneMapper_nCreatePBRNeutralToneMapper(JNIEnv*, jclass) {
-    return (jlong) new PBRNeutralToneMapper();
-}
-
 extern "C" JNIEXPORT jlong JNICALL
 Java_com_google_android_filament_ToneMapper_nCreateAgxToneMapper(JNIEnv*, jclass, jint look) {
    return (jlong) new AgxToneMapper(AgxToneMapper::AgxLook(look));
@@ -158,16 +158,6 @@ public class Engine {
        FEATURE_LEVEL_3,
    };

-    /**
-     * The type of technique for stereoscopic rendering
-     */
-    public enum StereoscopicType {
-        /** Stereoscopic rendering is performed using instanced rendering technique. */
-        INSTANCED,
-        /** Stereoscopic rendering is performed using the multiview feature from the graphics backend. */
-        MULTIVIEW,
-    };
-
    /**
     * Constructs <code>Engine</code> objects using a builder pattern.
     */
@@ -221,10 +211,7 @@ public class Engine {
            nSetBuilderConfig(mNativeBuilder, config.commandBufferSizeMB,
                    config.perRenderPassArenaSizeMB, config.driverHandleArenaSizeMB,
                    config.minCommandBufferSizeMB, config.perFrameCommandsSizeMB,
-                    config.jobSystemThreadCount,
-                    config.textureUseAfterFreePoolSize, config.disableParallelShaderCompile,
-                    config.stereoscopicType.ordinal(), config.stereoscopicEyeCount,
-                    config.resourceAllocatorCacheSizeMB, config.resourceAllocatorCacheMaxAge);
+                    config.jobSystemThreadCount, config.stereoscopicEyeCount);
            return this;
        }

@@ -361,35 +348,6 @@ public class Engine {
         */
        public long jobSystemThreadCount = 0;

-        /**
-         * Number of most-recently destroyed textures to track for use-after-free.
-         *
-         * This will cause the backend to throw an exception when a texture is freed but still bound
-         * to a SamplerGroup and used in a draw call. 0 disables completely.
-         *
-         * Currently only respected by the Metal backend.
-         */
-        public long textureUseAfterFreePoolSize = 0;
-
-        /**
-         * Set to `true` to forcibly disable parallel shader compilation in the backend.
-         * Currently only honored by the GL backend.
-         */
-        public boolean disableParallelShaderCompile = false;
-
-        /**
-         * The type of technique for stereoscopic rendering.
-         *
-         * This setting determines the algorithm used when stereoscopic rendering is enabled. This
-         * decision applies to the entire Engine for the lifetime of the Engine. E.g., multiple
-         * Views created from the Engine must use the same stereoscopic type.
-         *
-         * Each view can enable stereoscopic rendering via the StereoscopicOptions::enable flag.
-         *
-         * @see View#setStereoscopicOptions
-         */
-        public StereoscopicType stereoscopicType = StereoscopicType.INSTANCED;
-
        /**
         * The number of eyes to render when stereoscopic rendering is enabled. Supported values are
         * between 1 and Engine#getMaxStereoscopicEyes() (inclusive).
@@ -398,18 +356,6 @@ public class Engine {
         * @see Engine#getMaxStereoscopicEyes
         */
        public long stereoscopicEyeCount = 2;
-
-        /*
-         * Size in MiB of the frame graph texture cache. This should be adjusted based on the
-         * size of used render targets (typically the screen).
-         */
-        public long resourceAllocatorCacheSizeMB = 64;
-
-        /*
-         * This value determines for how many frames are texture entries kept in the cache.
-         * The default value of 30 corresponds to about half a second at 60 fps.
-         */
-        public long resourceAllocatorCacheMaxAge = 30;
    }

    private Engine(long nativeEngine, Config config) {
@@ -1281,9 +1227,7 @@ public class Engine {
    private static native void nSetBuilderConfig(long nativeBuilder, long commandBufferSizeMB,
            long perRenderPassArenaSizeMB, long driverHandleArenaSizeMB,
            long minCommandBufferSizeMB, long perFrameCommandsSizeMB, long jobSystemThreadCount,
-            long textureUseAfterFreePoolSize, boolean disableParallelShaderCompile,
-            int stereoscopicType, long stereoscopicEyeCount,
-            long resourceAllocatorCacheSizeMB, long resourceAllocatorCacheMaxAge);
+            long stereoscopicEyeCount);
    private static native void nSetBuilderFeatureLevel(long nativeBuilder, int ordinal);
    private static native void nSetBuilderSharedContext(long nativeBuilder, long sharedContext);
    private static native long nBuilderBuild(long nativeBuilder);
@@ -101,7 +101,7 @@ public class Renderer {
        /**
         * Desired frame interval in unit of 1 / DisplayInfo.refreshRate.
         */
-        public float interval = 1.0f;
+        public float interval = 1.0f / 60.0f;

        /**
         * Additional headroom for the GPU as a ratio of the targetFrameTime.
@@ -16,14 +16,12 @@ package com.google.android.filament;
 * <li>Configurable tone mapping operators</li>
 * <ul>
 *   <li>GenericToneMapper</li>
- *   <li>AgXToneMapper</li>
 * </ul>
 * <li>Fixed-aesthetic tone mapping operators</li>
 * <ul>
 *   <li>ACESToneMapper</li>
 *   <li>ACESLegacyToneMapper</li>
 *   <li>FilmicToneMapper</li>
- *   <li>PBRNeutralToneMapper</li>
 * </ul>
 * <li>Debug/validation tone mapping operators</li>
 * <ul>
@@ -102,21 +100,11 @@ public class ToneMapper {
        }
    }

-    /**
-     * Khronos PBR Neutral tone mapping operator. This tone mapper was designed
-     * to preserve the appearance of materials across lighting conditions while
-     * avoiding artifacts in the highlights in high dynamic range conditions.
-     */
-    public static class PBRNeutralToneMapper extends ToneMapper {
-        public PBRNeutralToneMapper() {
-            super(nCreatePBRNeutralToneMapper());
-        }
-    }
-
    /**
     * AgX tone mapping operator.
     */
    public static class Agx extends ToneMapper {
+
        public enum AgxLook {
            /**
             * Base contrast with no look applied
@@ -245,7 +233,6 @@ public class ToneMapper {
    private static native long nCreateACESToneMapper();
    private static native long nCreateACESLegacyToneMapper();
    private static native long nCreateFilmicToneMapper();
-    private static native long nCreatePBRNeutralToneMapper();
    private static native long nCreateAgxToneMapper(int look);
    private static native long nCreateGenericToneMapper(
            float contrast, float midGrayIn, float midGrayOut, float hdrMax);
@@ -1882,7 +1882,7 @@ public class View {
     * Options for Temporal Anti-aliasing (TAA)
     * Most TAA parameters are extremely costly to change, as they will trigger the TAA post-process
     * shaders to be recompiled. These options should be changed or set during initialization.
-     * `filterWidth`, `feedback` and `jitterPattern`, however, can be changed at any time.
+     * `filterWidth`, `feedback` and `jitterPattern`, however, could be changed at any time.
     *
     * `feedback` of 0.1 effectively accumulates a maximum of 19 samples in steady state.
     * see "A Survey of Temporal Antialiasing Techniques" by Lei Yang and all for more information.
@@ -1925,7 +1925,6 @@ public class View {
            UNIFORM_HELIX_X4,
            HALTON_23_X8,
            HALTON_23_X16,
-            HALTON_23_X32,
        }

        /**
@@ -1940,10 +1939,6 @@ public class View {
         * texturing lod bias (typically -1 or -2)
         */
        public float lodBias = -1.0f;
-        /**
-         * post-TAA sharpen, especially useful when upscaling is true.
-         */
-        public float sharpness = 0.0f;
        /**
         * enables or disables temporal anti-aliasing
         */
@@ -1,5 +1,5 @@
 GROUP=com.google.android.filament
-VERSION_NAME=1.50.5
+VERSION_NAME=1.49.3

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

@@ -30,7 +30,7 @@ android {
    compileSdkVersion versions.compileSdk
    defaultConfig {
        applicationId "com.google.android.filament.gltf"
-        minSdkVersion versions.minSdk
+        minSdkVersion 19
        targetSdkVersion versions.targetSdk
    }

@@ -1 +1 @@
-26.1.10909125
+25.1.8937393
@@ -125,13 +125,7 @@ if [[ "${has_universal}" == "true" ]]; then

        arch_output="${OUTPUT_PATH%.a}_${arch}.a"
        arch_outputs+=("$arch_output")
-
-        archives=()
-        while IFS=  read -r -d $'\0'; do
-            archives+=("$REPLY")
-        done < <(find "$(pwd)/${archs_temp_dir}/${arch}" -iname '*.a' -print0)
-
-        combine_static_libs "$arch_output" "$archives"
+        combine_static_libs "$arch_output" $(find "$(pwd)/${archs_temp_dir}/${arch}" -iname '*.a')
    done

    # Finally, combine the single-architecture archives into a universal binary.
@@ -21,7 +21,7 @@ set(CMAKE_SYSTEM_NAME Linux)
 set(CMAKE_SYSTEM_VERSION 1)

 # android
-set(API_LEVEL 21)
+set(API_LEVEL 19)

 # architecture
 set(ARCH armv7a-linux-androideabi)
@@ -21,7 +21,7 @@ set(CMAKE_SYSTEM_NAME Linux)
 set(CMAKE_SYSTEM_VERSION 1)

 # android
-set(API_LEVEL 21)
+set(API_LEVEL 19)

 # architecture
 set(ARCH i686-linux-android)
@@ -66,7 +66,6 @@ set(SRCS
        src/Froxelizer.cpp
        src/Frustum.cpp
        src/HwRenderPrimitiveFactory.cpp
-        src/HwVertexBufferInfoFactory.cpp
        src/IndexBuffer.cpp
        src/IndirectLight.cpp
        src/InstanceBuffer.cpp
@@ -138,7 +137,6 @@ set(SRCS

 set(PRIVATE_HDRS
        src/Allocators.h
-        src/Bimap.h
        src/BufferPoolAllocator.h
        src/ColorSpaceUtils.h
        src/Culler.h
@@ -149,7 +147,6 @@ set(PRIVATE_HDRS
        src/FrameSkipper.h
        src/Froxelizer.h
        src/HwRenderPrimitiveFactory.h
-        src/HwVertexBufferInfoFactory.h
        src/Intersections.h
        src/MaterialParser.h
        src/PerViewUniforms.h
@@ -215,7 +212,6 @@ set(MATERIAL_SRCS
        src/materials/antiAliasing/fxaa.mat
        src/materials/antiAliasing/taa.mat
        src/materials/blitLow.mat
-        src/materials/blitArray.mat
        src/materials/bloom/bloomDownsample.mat
        src/materials/bloom/bloomDownsample2x.mat
        src/materials/bloom/bloomDownsample9.mat
@@ -242,7 +238,6 @@ set(MATERIAL_SRCS
        src/materials/resolveDepth.mat
        src/materials/separableGaussianBlur.mat
        src/materials/skybox.mat
-        src/materials/shadowmap.mat
        src/materials/ssao/bilateralBlur.mat
        src/materials/ssao/bilateralBlurBentNormals.mat
        src/materials/ssao/mipmapDepth.mat
@@ -23,6 +23,7 @@
 #include <utils/ostream.h>

 #include <stddef.h>
+#include <stdint.h>

 namespace filament::backend {

@@ -17,6 +17,8 @@
 #ifndef TNT_FILAMENT_BACKEND_CALLBACKHANDLER_H
 #define TNT_FILAMENT_BACKEND_CALLBACKHANDLER_H

+#include <utils/compiler.h>
+
 namespace filament::backend {

 /**
@@ -28,8 +28,7 @@

 #include <math/vec4.h>

-#include <array>        // FIXME: STL headers are not allowed in public headers
-#include <type_traits>  // FIXME: STL headers are not allowed in public headers
+#include <array>    // FIXME: STL headers are not allowed in public headers

 #include <stddef.h>
 #include <stdint.h>
@@ -1212,14 +1211,6 @@ enum class Workaround : uint16_t {
    DISABLE_THREAD_AFFINITY
 };

-//! The type of technique for stereoscopic rendering
-enum class StereoscopicType : uint8_t {
-    // Stereoscopic rendering is performed using instanced rendering technique.
-    INSTANCED,
-    // Stereoscopic rendering is performed using the multiview feature from the graphics backend.
-    MULTIVIEW,
-};
-
 } // namespace filament::backend

 template<> struct utils::EnableBitMaskOperators<filament::backend::ShaderStageFlags>
@@ -17,15 +17,17 @@
 #ifndef TNT_FILAMENT_BACKEND_HANDLE_H
 #define TNT_FILAMENT_BACKEND_HANDLE_H

+#include <utils/compiler.h>
 #if !defined(NDEBUG)
-#include <utils/ostream.h>
+#include <utils/Log.h>
 #endif
 #include <utils/debug.h>

-#include <type_traits> // FIXME: STL headers are not allowed in public headers
-
 #include <stdint.h>

+#include <limits>
+#include <type_traits>
+
 namespace filament::backend {

 struct HwBufferObject;
@@ -39,7 +41,6 @@ struct HwStream;
 struct HwSwapChain;
 struct HwTexture;
 struct HwTimerQuery;
-struct HwVertexBufferInfo;
 struct HwVertexBuffer;

 /*
@@ -53,7 +54,7 @@ struct HwVertexBuffer;
 class HandleBase {
 public:
    using HandleId = uint32_t;
-    static constexpr const HandleId nullid = HandleId{ UINT32_MAX };
+    static constexpr const HandleId nullid = HandleId{ std::numeric_limits<HandleId>::max() };

    constexpr HandleBase() noexcept: object(nullid) {}

@@ -63,6 +64,14 @@ public:
    // clear the handle, this doesn't free associated resources
    void clear() noexcept { object = nullid; }

+    // compare handles
+    bool operator==(const HandleBase& rhs) const noexcept { return object == rhs.object; }
+    bool operator!=(const HandleBase& rhs) const noexcept { return object != rhs.object; }
+    bool operator<(const HandleBase& rhs) const noexcept { return object < rhs.object; }
+    bool operator<=(const HandleBase& rhs) const noexcept { return object <= rhs.object; }
+    bool operator>(const HandleBase& rhs) const noexcept { return object > rhs.object; }
+    bool operator>=(const HandleBase& rhs) const noexcept { return object >= rhs.object; }
+
    // get this handle's handleId
    HandleId getId() const noexcept { return object; }

@@ -94,14 +103,6 @@ struct Handle : public HandleBase {

    explicit Handle(HandleId id) noexcept : HandleBase(id) { }

-    // compare handles of the same type
-    bool operator==(const Handle& rhs) const noexcept { return getId() == rhs.getId(); }
-    bool operator!=(const Handle& rhs) const noexcept { return getId() != rhs.getId(); }
-    bool operator<(const Handle& rhs) const noexcept { return getId() < rhs.getId(); }
-    bool operator<=(const Handle& rhs) const noexcept { return getId() <= rhs.getId(); }
-    bool operator>(const Handle& rhs) const noexcept { return getId() > rhs.getId(); }
-    bool operator>=(const Handle& rhs) const noexcept { return getId() >= rhs.getId(); }
-
    // type-safe Handle cast
    template<typename B, typename = std::enable_if_t<std::is_base_of<T, B>::value> >
    Handle(Handle<B> const& base) noexcept : HandleBase(base) { } // NOLINT(hicpp-explicit-conversions,google-explicit-constructor)
@@ -115,19 +116,18 @@ private:

 // Types used by the command stream
 // (we use this renaming because the macro-system doesn't deal well with "<" and ">")
-using BufferObjectHandle     = Handle<HwBufferObject>;
-using FenceHandle            = Handle<HwFence>;
-using IndexBufferHandle      = Handle<HwIndexBuffer>;
-using ProgramHandle          = Handle<HwProgram>;
-using RenderPrimitiveHandle  = Handle<HwRenderPrimitive>;
-using RenderTargetHandle     = Handle<HwRenderTarget>;
-using SamplerGroupHandle     = Handle<HwSamplerGroup>;
-using StreamHandle           = Handle<HwStream>;
-using SwapChainHandle        = Handle<HwSwapChain>;
-using TextureHandle          = Handle<HwTexture>;
-using TimerQueryHandle       = Handle<HwTimerQuery>;
-using VertexBufferHandle     = Handle<HwVertexBuffer>;
-using VertexBufferInfoHandle = Handle<HwVertexBufferInfo>;
+using BufferObjectHandle    = Handle<HwBufferObject>;
+using FenceHandle           = Handle<HwFence>;
+using IndexBufferHandle     = Handle<HwIndexBuffer>;
+using ProgramHandle         = Handle<HwProgram>;
+using RenderPrimitiveHandle = Handle<HwRenderPrimitive>;
+using RenderTargetHandle    = Handle<HwRenderTarget>;
+using SamplerGroupHandle    = Handle<HwSamplerGroup>;
+using StreamHandle          = Handle<HwStream>;
+using SwapChainHandle       = Handle<HwSwapChain>;
+using TextureHandle         = Handle<HwTexture>;
+using TimerQueryHandle      = Handle<HwTimerQuery>;
+using VertexBufferHandle    = Handle<HwVertexBuffer>;

 } // namespace filament::backend

@@ -20,7 +20,7 @@
 #include <backend/DriverEnums.h>
 #include <backend/Handle.h>

-#include <utils/ostream.h>
+#include <limits>

 #include <stdint.h>

@@ -33,6 +33,10 @@ struct PipelineState {
    RasterState rasterState;
    StencilState stencilState;
    PolygonOffset polygonOffset;
+    Viewport scissor{ 0, 0,
+                      (uint32_t)std::numeric_limits<int32_t>::max(),
+                      (uint32_t)std::numeric_limits<int32_t>::max()
+    };
 };

 } // namespace filament::backend
@@ -24,7 +24,6 @@

 #include <utils/compiler.h>
 #include <utils/debug.h>
-#include <utils/ostream.h>

 #include <stddef.h>
 #include <stdint.h>
@@ -19,11 +19,11 @@
 #ifndef TNT_FILAMENT_BACKEND_PLATFORM_H
 #define TNT_FILAMENT_BACKEND_PLATFORM_H

+#include <backend/DriverEnums.h>
+
 #include <utils/compiler.h>
 #include <utils/Invocable.h>

-#include <stddef.h>
-
 namespace filament::backend {

 class Driver;
@@ -41,24 +41,18 @@ public:
    struct Stream {};

    struct DriverConfig {
-        /**
-         * Size of handle arena in bytes. Setting to 0 indicates default value is to be used.
+        /*
+         * size of handle arena in bytes. Setting to 0 indicates default value is to be used.
         * Driver clamps to valid values.
         */
        size_t handleArenaSize = 0;

-        /**
-         * This number of most-recently destroyed textures will be tracked for use-after-free.
+        /*
+         * this number of most-recently destroyed textures will be tracked for use-after-free.
         * Throws an exception when a texture is freed but still bound to a SamplerGroup and used in
         * a draw call. 0 disables completely. Currently only respected by the Metal backend.
         */
        size_t textureUseAfterFreePoolSize = 0;
-
-        /**
-         * Set to `true` to forcibly disable parallel shader compilation in the backend.
-         * Currently only honored by the GL backend.
-         */
-        bool disableParallelShaderCompile = false;
    };

    Platform() noexcept;
@@ -84,7 +78,7 @@ public:
     *
     * @return nullptr on failure, or a pointer to the newly created driver.
     */
-    virtual backend::Driver* UTILS_NULLABLE createDriver(void* UTILS_NULLABLE sharedContext,
+    virtual backend::Driver* createDriver(void* sharedContext,
            const DriverConfig& driverConfig) noexcept = 0;

    /**
@@ -102,8 +96,7 @@ public:
     * cache.
     */
    using InsertBlobFunc = utils::Invocable<
-            void(const void* UTILS_NONNULL key, size_t keySize,
-                    const void* UTILS_NONNULL value, size_t valueSize)>;
+            void(const void* key, size_t keySize, const void* value, size_t valueSize)>;

    /*
     * RetrieveBlobFunc is an Invocable to an application-provided function that a
@@ -111,8 +104,7 @@ public:
     * cache.
     */
    using RetrieveBlobFunc = utils::Invocable<
-            size_t(const void* UTILS_NONNULL key, size_t keySize,
-                    void* UTILS_NONNULL value, size_t valueSize)>;
+            size_t(const void* key, size_t keySize, void* value, size_t valueSize)>;

    /**
     * Sets the callback functions that the backend can use to interact with caching functionality
@@ -165,8 +157,7 @@ public:
     * @param value         pointer to the beginning of the value data that is to be inserted
     * @param valueSize     specifies the size in byte of the data pointed to by <value>
     */
-    void insertBlob(const void* UTILS_NONNULL key, size_t keySize,
-            const void* UTILS_NONNULL value, size_t valueSize);
+    void insertBlob(const void* key, size_t keySize, const void* value, size_t valueSize);

    /**
     * To retrieve the binary value associated with a given key from the cache, a
@@ -185,8 +176,7 @@ public:
     * @return             If the cache contains a value associated with the given key then the
     *                     size of that binary value in bytes is returned. Otherwise 0 is returned.
     */
-    size_t retrieveBlob(const void* UTILS_NONNULL key, size_t keySize,
-            void* UTILS_NONNULL value, size_t valueSize);
+    size_t retrieveBlob(const void* key, size_t keySize, void* value, size_t valueSize);

 private:
    InsertBlobFunc mInsertBlob;
@@ -21,7 +21,8 @@

 #include <utils/compiler.h>

-namespace filament::backend {
+namespace filament {
+namespace backend {

 /**
 * A PresentCallable is a callable object that, when called, schedules a frame for presentation on
@@ -97,6 +98,7 @@ private:
 */
 using FrameFinishedCallback UTILS_DEPRECATED = void(*)(PresentCallable callable, void* user);

-} // namespace filament::backend
+} // namespace backend
+} // namespace filament

 #endif // TNT_FILAMENT_BACKEND_PRESENTCALLABLE
@@ -17,19 +17,17 @@
 #ifndef TNT_FILAMENT_BACKEND_PRIVATE_PROGRAM_H
 #define TNT_FILAMENT_BACKEND_PRIVATE_PROGRAM_H

+#include <utils/compiler.h>
 #include <utils/CString.h>
 #include <utils/FixedCapacityVector.h>
 #include <utils/Invocable.h>
+#include <utils/Log.h>
 #include <utils/ostream.h>

 #include <backend/DriverEnums.h>

-#include <array>    // FIXME: STL headers are not allowed in public headers
-#include <utility>  // FIXME: STL headers are not allowed in public headers
-#include <variant>  // FIXME: STL headers are not allowed in public headers
-
-#include <stddef.h>
-#include <stdint.h>
+#include <array>
+#include <variant>

 namespace filament::backend {

@@ -24,6 +24,9 @@

 #include <utils/compiler.h>

+#include <stddef.h>
+#include <stdint.h>
+
 namespace filament::backend {

 struct UTILS_PUBLIC SamplerDescriptor {
@@ -17,11 +17,9 @@
 #ifndef TNT_FILAMENT_BACKEND_TARGETBUFFERINFO_H
 #define TNT_FILAMENT_BACKEND_TARGETBUFFERINFO_H

+#include <backend/DriverEnums.h>
 #include <backend/Handle.h>

-#include <utils/ostream.h>
-
-#include <stddef.h>
 #include <stdint.h>

 namespace filament::backend {
@@ -18,13 +18,8 @@
 #define TNT_FILAMENT_BACKEND_PRIVATE_OPENGLPLATFORM_H

 #include <backend/AcquiredImage.h>
-#include <backend/DriverEnums.h>
 #include <backend/Platform.h>

-#include <utils/compiler.h>
-
-#include <stdint.h>
-
 namespace filament::backend {

 class Driver;
@@ -43,8 +38,8 @@ protected:
     * Derived classes can use this to instantiate the default OpenGLDriver backend.
     * This is typically called from your implementation of createDriver()
     */
-    static Driver* UTILS_NULLABLE createDefaultDriver(OpenGLPlatform* UTILS_NONNULL platform,
-            void* UTILS_NULLABLE sharedContext, const DriverConfig& driverConfig);
+    static Driver* createDefaultDriver(OpenGLPlatform* platform,
+            void* sharedContext, const DriverConfig& driverConfig);

    ~OpenGLPlatform() noexcept override;

@@ -71,8 +66,7 @@ public:
     * @return              The driver's SwapChain object.
     *
     */
-    virtual SwapChain* UTILS_NONNULL createSwapChain(
-            void* UTILS_NULLABLE nativeWindow, uint64_t flags) noexcept = 0;
+    virtual SwapChain* createSwapChain(void* nativeWindow, uint64_t flags) noexcept = 0;

    /**
     * Return whether createSwapChain supports the SWAP_CHAIN_CONFIG_SRGB_COLORSPACE flag.
@@ -93,14 +87,13 @@ public:
     * TODO: we need a more generic way of passing construction parameters
     *       A void* might be enough.
     */
-    virtual SwapChain* UTILS_NULLABLE createSwapChain(
-            uint32_t width, uint32_t height, uint64_t flags) noexcept = 0;
+    virtual SwapChain* createSwapChain(uint32_t width, uint32_t height, uint64_t flags) noexcept = 0;

    /**
     * Called by the driver to destroys the SwapChain
     * @param swapChain SwapChain to be destroyed.
     */
-    virtual void destroySwapChain(SwapChain* UTILS_NONNULL swapChain) noexcept = 0;
+    virtual void destroySwapChain(SwapChain* swapChain) noexcept = 0;

    /**
     * Returns the set of buffers that must be preserved up to the call to commit().
@@ -113,7 +106,7 @@ public:
     * @return buffer that must be preserved
     * @see commit()
     */
-    virtual TargetBufferFlags getPreservedFlags(SwapChain* UTILS_NONNULL swapChain) noexcept;
+    virtual TargetBufferFlags getPreservedFlags(SwapChain* swapChain) noexcept;

    /**
     * Called by the driver to establish the default FBO. The default implementation returns 0.
@@ -127,16 +120,14 @@ public:
     * @param drawSwapChain SwapChain to draw to. It must be bound to the default FBO.
     * @param readSwapChain SwapChain to read from (for operation like `glBlitFramebuffer`)
     */
-    virtual void makeCurrent(
-            SwapChain* UTILS_NONNULL drawSwapChain,
-            SwapChain* UTILS_NONNULL readSwapChain) noexcept = 0;
+    virtual void makeCurrent(SwapChain* drawSwapChain, SwapChain* readSwapChain) noexcept = 0;

    /**
     * Called by the driver once the current frame finishes drawing. Typically, this should present
     * the drawSwapChain. This is for example where `eglMakeCurrent()` would be called.
     * @param swapChain the SwapChain to present.
     */
-    virtual void commit(SwapChain* UTILS_NONNULL swapChain) noexcept = 0;
+    virtual void commit(SwapChain* swapChain) noexcept = 0;

    /**
     * Set the time the next committed buffer should be presented to the user at.
@@ -161,14 +152,14 @@ public:
     *
     * @return A Fence object. The default implementation returns nullptr.
     */
-    virtual Fence* UTILS_NULLABLE createFence() noexcept;
+    virtual Fence* createFence() noexcept;

    /**
     * Destroys a Fence object. The default implementation does nothing.
     *
     * @param fence Fence to destroy.
     */
-    virtual void destroyFence(Fence* UTILS_NONNULL fence) noexcept;
+    virtual void destroyFence(Fence* fence) noexcept;

    /**
     * Waits on a Fence.
@@ -178,7 +169,7 @@ public:
     * @return Whether the fence signaled or timed out. See backend::FenceStatus.
     *         The default implementation always return backend::FenceStatus::ERROR.
     */
-    virtual backend::FenceStatus waitFence(Fence* UTILS_NONNULL fence, uint64_t timeout) noexcept;
+    virtual backend::FenceStatus waitFence(Fence* fence, uint64_t timeout) noexcept;


    // --------------------------------------------------------------------------------------------
@@ -192,13 +183,13 @@ public:
     * @param nativeStream The native stream, this parameter depends on the concrete implementation.
     * @return A new Stream object.
     */
-    virtual Stream* UTILS_NULLABLE createStream(void* UTILS_NULLABLE nativeStream) noexcept;
+    virtual Stream* createStream(void* nativeStream) noexcept;

    /**
     * Destroys a Stream.
     * @param stream Stream to destroy.
     */
-    virtual void destroyStream(Stream* UTILS_NONNULL stream) noexcept;
+    virtual void destroyStream(Stream* stream) noexcept;

    /**
     * The specified stream takes ownership of the texture (tname) object
@@ -208,21 +199,20 @@ public:
     * @param stream Stream to take ownership of the texture
     * @param tname  GL texture id to "bind" to the Stream.
     */
-    virtual void attach(Stream* UTILS_NONNULL stream, intptr_t tname) noexcept;
+    virtual void attach(Stream* stream, intptr_t tname) noexcept;

    /**
     * Destroys the texture associated to the stream
     * @param stream Stream to detach from its texture
     */
-    virtual void detach(Stream* UTILS_NONNULL stream) noexcept;
+    virtual void detach(Stream* stream) noexcept;

    /**
     * Updates the content of the texture attached to the stream.
     * @param stream Stream to update
     * @param timestamp Output parameter: Timestamp of the image bound to the texture.
     */
-    virtual void updateTexImage(Stream* UTILS_NONNULL stream,
-            int64_t* UTILS_NONNULL timestamp) noexcept;
+    virtual void updateTexImage(Stream* stream, int64_t* timestamp) noexcept;


    // --------------------------------------------------------------------------------------------
@@ -235,13 +225,13 @@ public:
     *         implementation could just return { 0, GL_TEXTURE_2D } at this point. The actual
     *         values can be delayed until setExternalImage.
     */
-    virtual ExternalTexture* UTILS_NULLABLE createExternalImageTexture() noexcept;
+    virtual ExternalTexture *createExternalImageTexture() noexcept;

    /**
     * Destroys an external texture handle and associated data.
     * @param texture a pointer to the handle to destroy.
     */
-    virtual void destroyExternalImage(ExternalTexture* UTILS_NONNULL texture) noexcept;
+    virtual void destroyExternalImage(ExternalTexture* texture) noexcept;

    // called on the application thread to allow Filament to take ownership of the image

@@ -254,7 +244,7 @@ public:
     * @param externalImage A token representing the platform's external image.
     * @see destroyExternalImage
     */
-    virtual void retainExternalImage(void* UTILS_NONNULL externalImage) noexcept;
+    virtual void retainExternalImage(void* externalImage) noexcept;

    /**
     * Called to bind the platform-specific externalImage to an ExternalTexture.
@@ -268,8 +258,7 @@ public:
     * @param texture an in/out pointer to ExternalTexture, id and target can be updated if necessary.
     * @return true on success, false on error.
     */
-    virtual bool setExternalImage(void* UTILS_NONNULL externalImage,
-            ExternalTexture* UTILS_NONNULL texture) noexcept;
+    virtual bool setExternalImage(void* externalImage, ExternalTexture* texture) noexcept;

    /**
     * The method allows platforms to convert a user-supplied external image object into a new type
@@ -17,10 +17,11 @@
 #ifndef TNT_FILAMENT_BACKEND_OPENGL_OPENGL_PLATFORM_COCOA_GL_H
 #define TNT_FILAMENT_BACKEND_OPENGL_OPENGL_PLATFORM_COCOA_GL_H

-#include <backend/DriverEnums.h>
+#include <stdint.h>
+
 #include <backend/platforms/OpenGLPlatform.h>

-#include <stdint.h>
+#include <backend/DriverEnums.h>

 namespace filament::backend {

@@ -30,7 +30,7 @@ struct PlatformCocoaTouchGLImpl;
 class PlatformCocoaTouchGL : public OpenGLPlatform {
 public:
    PlatformCocoaTouchGL();
-    ~PlatformCocoaTouchGL() noexcept override;
+    ~PlatformCocoaTouchGL() noexcept;

    // --------------------------------------------------------------------------------------------
    // Platform Interface
@@ -17,21 +17,18 @@
 #ifndef TNT_FILAMENT_BACKEND_OPENGL_OPENGL_PLATFORM_EGL_H
 #define TNT_FILAMENT_BACKEND_OPENGL_OPENGL_PLATFORM_EGL_H

-#include <backend/DriverEnums.h>
-#include <backend/Platform.h>
-#include <backend/platforms/OpenGLPlatform.h>
+#include <stdint.h>

 #include <EGL/egl.h>
 #include <EGL/eglext.h>
-#include <EGL/eglplatform.h>

-#include <initializer_list>
+#include <backend/platforms/OpenGLPlatform.h>
+
+#include <backend/DriverEnums.h>
+
 #include <utility>
 #include <vector>

-#include <stddef.h>
-#include <stdint.h>
-
 namespace filament::backend {

 /**
@@ -151,13 +148,6 @@ protected:
        } egl;
    } ext;

-    struct SwapChainEGL : public Platform::SwapChain {
-        EGLSurface sur = EGL_NO_SURFACE;
-        Config attribs{};
-        EGLNativeWindowType nativeWindow{};
-        EGLConfig config{};
-    };
-
    void initializeGlExtensions() noexcept;

 protected:
@@ -17,14 +17,8 @@
 #ifndef TNT_FILAMENT_BACKEND_OPENGL_OPENGL_PLATFORM_EGL_ANDROID_H
 #define TNT_FILAMENT_BACKEND_OPENGL_OPENGL_PLATFORM_EGL_ANDROID_H

-#include <backend/AcquiredImage.h>
-#include <backend/Platform.h>
-#include <backend/platforms/OpenGLPlatform.h>
 #include <backend/platforms/PlatformEGL.h>

-#include <stddef.h>
-#include <stdint.h>
-
 namespace filament::backend {

 class ExternalStreamManagerAndroid;
@@ -20,18 +20,13 @@
 #include <backend/Platform.h>

 #include <bluevk/BlueVK.h>
-
 #include <utils/CString.h>
 #include <utils/FixedCapacityVector.h>
 #include <utils/PrivateImplementation.h>

-#include <string_view>
 #include <tuple>
 #include <unordered_set>

-#include <stddef.h>
-#include <stdint.h>
-
 namespace filament::backend {

 using SwapChain = Platform::SwapChain;
@@ -17,7 +17,7 @@
 #ifndef TNT_FILAMENT_BACKEND_PRIVATE_CIRCULARBUFFER_H
 #define TNT_FILAMENT_BACKEND_PRIVATE_CIRCULARBUFFER_H

-#include <utils/debug.h>
+#include <utils/compiler.h>

 #include <stddef.h>
 #include <stdint.h>
@@ -40,36 +40,28 @@ public:

    ~CircularBuffer() noexcept;

-    static size_t getBlockSize() noexcept { return sPageSize; }
-
-    // Total size of circular buffer. This is a constant.
-    size_t size() const noexcept { return mSize; }
-
-    // Allocates `s` bytes in the circular buffer and returns a pointer to the memory. All
-    // allocations must not exceed size() bytes.
-    inline void* allocate(size_t s) noexcept {
-        // We can never allocate more that size().
-        assert_invariant(getUsed() + s <= size());
+    // allocates 'size' bytes in the circular buffer and returns a pointer to the memory
+    // return the current head and moves it forward by size bytes
+    inline void* allocate(size_t size) noexcept {
        char* const cur = static_cast<char*>(mHead);
-        mHead = cur + s;
+        mHead = cur + size;
        return cur;
    }

-    // Returns true if the buffer is empty, i.e.: no allocations were made since
-    // calling getBuffer();
+    // Total size of circular buffer
+    size_t size() const noexcept { return mSize; }
+
+    // returns true if the buffer is empty (e.g. after calling flush)
    bool empty() const noexcept { return mTail == mHead; }

-    // Returns the size used since the last call to getBuffer()
-    size_t getUsed() const noexcept { return intptr_t(mHead) - intptr_t(mTail); }
+    void* getHead() const noexcept { return mHead; }

-    // Retrieves the current allocated range and frees it. It is the responsibility of the caller
-    // to make sure the returned range is no longer in use by the time allocate() allocates
-    // (size() - getUsed()) bytes.
-    struct Range {
-        void* tail;
-        void* head;
-    };
-    Range getBuffer() noexcept;
+    void* getTail() const noexcept { return mTail; }
+
+    // call at least once every getRequiredSize() bytes allocated from the buffer
+    void circularize() noexcept;
+
+    static size_t getBlockSize() noexcept { return sPageSize; }

 private:
    void* alloc(size_t size) noexcept;
@@ -77,10 +69,10 @@ private:

    // pointer to the beginning of the circular buffer (constant)
    void* mData = nullptr;
-    int mAshmemFd = -1;
+    int mUsesAshmem = -1;

    // size of the circular buffer (constant)
-    size_t const mSize;
+    size_t mSize = 0;

    // pointer to the beginning of recorded data
    void* mTail = nullptr;
@@ -19,21 +19,19 @@

 #include "private/backend/CircularBuffer.h"

+#include <utils/compiler.h>
 #include <utils/Condition.h>
 #include <utils/Mutex.h>

 #include <vector>

-#include <stddef.h>
-#include <stdint.h>
-
 namespace filament::backend {

 /*
 * A producer-consumer command queue that uses a CircularBuffer as main storage
 */
 class CommandBufferQueue {
-    struct Range {
+    struct Slice {
        void* begin;
        void* end;
    };
@@ -46,7 +44,7 @@ class CommandBufferQueue {

    mutable utils::Mutex mLock;
    mutable utils::Condition mCondition;
-    mutable std::vector<Range> mCommandBuffersToExecute;
+    mutable std::vector<Slice> mCommandBuffersToExecute;
    size_t mFreeSpace = 0;
    size_t mHighWatermark = 0;
    uint32_t mExitRequested = 0;
@@ -58,20 +56,17 @@ public:
    CommandBufferQueue(size_t requiredSize, size_t bufferSize);
    ~CommandBufferQueue();

-    CircularBuffer& getCircularBuffer() noexcept { return mCircularBuffer; }
-    CircularBuffer const& getCircularBuffer() const noexcept { return mCircularBuffer; }
-
-    size_t getCapacity() const noexcept { return mRequiredSize; }
+    CircularBuffer& getCircularBuffer() { return mCircularBuffer; }

    size_t getHighWatermark() const noexcept { return mHighWatermark; }

    // wait for commands to be available and returns an array containing these commands
-    std::vector<Range> waitForCommands() const;
+    std::vector<Slice> waitForCommands() const;

    // return the memory used by this command buffer to the circular buffer
    // WARNING: releaseBuffer() must be called in sequence of the Slices returned by
    // waitForCommands()
-    void releaseBuffer(Range const& buffer);
+    void releaseBuffer(Slice const& buffer);

    // all commands buffers (Slices) written to this point are returned by waitForCommand(). This
    // call blocks until the CircularBuffer has at least mRequiredSize bytes available.
@@ -32,13 +32,11 @@
 #include <backend/TargetBufferInfo.h>

 #include <utils/compiler.h>
-#include <utils/debug.h>
 #include <utils/ThreadUtils.h>

 #include <cstddef>
 #include <functional>
 #include <tuple>
-#include <type_traits>
 #include <utility>

 #ifndef NDEBUG
@@ -154,14 +152,14 @@ struct CommandType<void (Driver::*)(ARGS...)> {
        }

        // placement new declared as "throw" to avoid the compiler's null-check
-        inline void* operator new(std::size_t, void* ptr) {
+        inline void* operator new(std::size_t size, void* ptr) {
            assert_invariant(ptr);
            return ptr;
        }
    };
 };

-// convert a method of "class Driver" into a Command<> type
+// convert an method of "class Driver" into a Command<> type
 #define COMMAND_TYPE(method) CommandType<decltype(&Driver::method)>::Command<&Driver::method>

 // ------------------------------------------------------------------------------------------------
@@ -213,8 +211,6 @@ public:
    CommandStream(CommandStream const& rhs) noexcept = delete;
    CommandStream& operator=(CommandStream const& rhs) noexcept = delete;

-    CircularBuffer const& getCircularBuffer() const noexcept { return mCurrentBuffer; }
-
 public:
 #define DECL_DRIVER_API(methodName, paramsDecl, params)                                         \
    inline void methodName(paramsDecl) {                                                        \
@@ -29,7 +29,6 @@

 #include <functional>

-#include <stddef.h>
 #include <stdint.h>

 // Command debugging off. debugging virtuals are not called.
@@ -167,14 +167,11 @@ DECL_DRIVER_API_0(resetState)
 * -----------------------
 */

-DECL_DRIVER_API_R_N(backend::VertexBufferInfoHandle, createVertexBufferInfo,
+DECL_DRIVER_API_R_N(backend::VertexBufferHandle, createVertexBuffer,
        uint8_t, bufferCount,
        uint8_t, attributeCount,
-        backend::AttributeArray, attributes)
-
-DECL_DRIVER_API_R_N(backend::VertexBufferHandle, createVertexBuffer,
        uint32_t, vertexCount,
-        backend::VertexBufferInfoHandle, vbih)
+        backend::AttributeArray, attributes)

 DECL_DRIVER_API_R_N(backend::IndexBufferHandle, createIndexBuffer,
        backend::ElementType, elementType,
@@ -227,7 +224,11 @@ DECL_DRIVER_API_R_N(backend::SamplerGroupHandle, createSamplerGroup,
 DECL_DRIVER_API_R_N(backend::RenderPrimitiveHandle, createRenderPrimitive,
        backend::VertexBufferHandle, vbh,
        backend::IndexBufferHandle, ibh,
-        backend::PrimitiveType, pt)
+        backend::PrimitiveType, pt,
+        uint32_t, offset,
+        uint32_t, minIndex,
+        uint32_t, maxIndex,
+        uint32_t, count)

 DECL_DRIVER_API_R_N(backend::ProgramHandle, createProgram,
        backend::Program&&, program)
@@ -263,7 +264,6 @@ DECL_DRIVER_API_R_0(backend::TimerQueryHandle, createTimerQuery)
 */

 DECL_DRIVER_API_N(destroyVertexBuffer,    backend::VertexBufferHandle, vbh)
-DECL_DRIVER_API_N(destroyVertexBufferInfo,backend::VertexBufferInfoHandle, vbih)
 DECL_DRIVER_API_N(destroyIndexBuffer,     backend::IndexBufferHandle, ibh)
 DECL_DRIVER_API_N(destroyBufferObject,    backend::BufferObjectHandle, ibh)
 DECL_DRIVER_API_N(destroyRenderPrimitive, backend::RenderPrimitiveHandle, rph)
@@ -298,7 +298,7 @@ DECL_DRIVER_API_SYNCHRONOUS_0(bool, isFrameBufferFetchMultiSampleSupported)
 DECL_DRIVER_API_SYNCHRONOUS_0(bool, isFrameTimeSupported)
 DECL_DRIVER_API_SYNCHRONOUS_0(bool, isAutoDepthResolveSupported)
 DECL_DRIVER_API_SYNCHRONOUS_0(bool, isSRGBSwapChainSupported)
-DECL_DRIVER_API_SYNCHRONOUS_N(bool, isStereoSupported, backend::StereoscopicType, stereoscopicType)
+DECL_DRIVER_API_SYNCHRONOUS_0(bool, isStereoSupported)
 DECL_DRIVER_API_SYNCHRONOUS_0(bool, isParallelShaderCompileSupported)
 DECL_DRIVER_API_SYNCHRONOUS_0(bool, isDepthStencilResolveSupported)
 DECL_DRIVER_API_SYNCHRONOUS_0(uint8_t, getMaxDrawBuffers)
@@ -494,17 +494,12 @@ DECL_DRIVER_API_N(blit,
 DECL_DRIVER_API_N(draw,
        backend::PipelineState, state,
        backend::RenderPrimitiveHandle, rph,
-        uint32_t, indexOffset,
-        uint32_t, indexCount,
        uint32_t, instanceCount)

 DECL_DRIVER_API_N(dispatchCompute,
        backend::ProgramHandle, program,
        math::uint3, workGroupCount)

-DECL_DRIVER_API_N(scissor,
-        Viewport, scissor)
-

 #pragma clang diagnostic pop

@@ -22,33 +22,28 @@
 #include <utils/Allocator.h>
 #include <utils/Log.h>
 #include <utils/compiler.h>
-#include <utils/debug.h>
-#include <utils/ostream.h>
-#include <utils/Panic.h>
-
 #include <tsl/robin_map.h>
-
-#include <cstddef>
-#include <exception>
-#include <type_traits>
 #include <unordered_map>
-#include <utility>

-#include <stddef.h>
-#include <stdint.h>
+#if !defined(NDEBUG) && UTILS_HAS_RTTI
+#   define HANDLE_TYPE_SAFETY 1
+#else
+#   define HANDLE_TYPE_SAFETY 0
+#endif

-#define HandleAllocatorGL  HandleAllocator<24,  64, 136>    // ~3640 / pool / MiB
-#define HandleAllocatorVK  HandleAllocator<80, 176, 320>    // ~1820 / pool / MiB
-#define HandleAllocatorMTL HandleAllocator<48, 160, 592>    // ~1310 / pool / MiB
+#define HandleAllocatorGL  HandleAllocator<16, 64, 208>
+#define HandleAllocatorVK  HandleAllocator<16, 64, 880>
+#define HandleAllocatorMTL HandleAllocator<16, 64, 584>

 namespace filament::backend {

 /*
 * A utility class to efficiently allocate and manage Handle<>
 */
-template<size_t P0, size_t P1, size_t P2>
+template <size_t P0, size_t P1, size_t P2>
 class HandleAllocator {
 public:
+
    HandleAllocator(const char* name, size_t size) noexcept;
    HandleAllocator(HandleAllocator const& rhs) = delete;
    HandleAllocator& operator=(HandleAllocator const& rhs) = delete;
@@ -66,9 +61,14 @@ public:
     */
    template<typename D, typename ... ARGS>
    Handle<D> allocateAndConstruct(ARGS&& ... args) noexcept {
-        Handle<D> h{ allocateHandle<D>() };
+        Handle<D> h{ allocateHandle<sizeof(D)>() };
        D* addr = handle_cast<D*>(h);
        new(addr) D(std::forward<ARGS>(args)...);
+#if HANDLE_TYPE_SAFETY
+        mLock.lock();
+        mHandleTypeId[addr] = typeid(D).name();
+        mLock.unlock();
+#endif
        return h;
    }

@@ -84,7 +84,13 @@ public:
     */
    template<typename D>
    Handle<D> allocate() noexcept {
-        Handle<D> h{ allocateHandle<D>() };
+        Handle<D> h{ allocateHandle<sizeof(D)>() };
+#if HANDLE_TYPE_SAFETY
+        D* addr = handle_cast<D*>(h);
+        mLock.lock();
+        mHandleTypeId[addr] = typeid(D).name();
+        mLock.unlock();
+#endif
        return h;
    }

@@ -101,10 +107,17 @@ public:
        assert_invariant(handle);
        D* addr = handle_cast<D*>(const_cast<Handle<B>&>(handle));
        assert_invariant(addr);
+
        // currently we implement construct<> with dtor+ctor, we could use operator= also
        // but all our dtors are trivial, ~D() is actually a noop.
        addr->~D();
        new(addr) D(std::forward<ARGS>(args)...);
+
+#if HANDLE_TYPE_SAFETY
+        mLock.lock();
+        mHandleTypeId[addr] = typeid(D).name();
+        mLock.unlock();
+#endif
        return addr;
    }

@@ -121,6 +134,12 @@ public:
        D* addr = handle_cast<D*>(const_cast<Handle<B>&>(handle));
        assert_invariant(addr);
        new(addr) D(std::forward<ARGS>(args)...);
+
+#if HANDLE_TYPE_SAFETY
+        mLock.lock();
+        mHandleTypeId[addr] = typeid(D).name();
+        mLock.unlock();
+#endif
        return addr;
    }

@@ -136,8 +155,19 @@ public:
    void deallocate(Handle<B>& handle, D const* p) noexcept {
        // allow to destroy the nullptr, similarly to operator delete
        if (p) {
+#if HANDLE_TYPE_SAFETY
+            mLock.lock();
+            auto typeId = mHandleTypeId[p];
+            mHandleTypeId.erase(p);
+            mLock.unlock();
+            if (UTILS_UNLIKELY(typeId != typeid(D).name())) {
+                utils::slog.e << "Destroying handle " << handle.getId() << ", type " << typeid(D).name()
+                       << ", but handle's actual type is " << typeId << utils::io::endl;
+                std::terminate();
+            }
+#endif
            p->~D();
-            deallocateHandle<D>(handle.getId());
+            deallocateHandle<sizeof(D)>(handle.getId());
        }
    }

@@ -165,17 +195,7 @@ public:
            std::is_base_of_v<B, typename std::remove_pointer_t<Dp>>, Dp>
    handle_cast(Handle<B>& handle) noexcept {
        assert_invariant(handle);
-        auto [p, tag] = handleToPointer(handle.getId());
-
-        if (isPoolHandle(handle.getId())) {
-            // check for use after free
-            uint8_t const age = (tag & HANDLE_AGE_MASK) >> HANDLE_AGE_SHIFT;
-            auto const pNode = static_cast<typename Allocator::Node*>(p);
-            uint8_t const expectedAge = pNode[-1].age;
-            ASSERT_POSTCONDITION(expectedAge == age,
-                    "use-after-free of Handle with id=%d", handle.getId());
-        }
-
+        void* const p = handleToPointer(handle.getId());
        return static_cast<Dp>(p);
    }

@@ -190,57 +210,29 @@ public:

 private:

-    template<typename D>
-    static constexpr size_t getBucketSize() noexcept {
-        if constexpr (sizeof(D) <= P0) { return P0; }
-        if constexpr (sizeof(D) <= P1) { return P1; }
-        static_assert(sizeof(D) <= P2);
-        return P2;
-    }
-
+    // template <int P0, int P1, int P2>
    class Allocator {
        friend class HandleAllocator;
-        static constexpr size_t MIN_ALIGNMENT = alignof(std::max_align_t);
-        struct Node { uint8_t age; };
-        // Note: using the `extra` parameter of PoolAllocator<>, even with a 1-byte structure,
-        // generally increases all pool allocations by 8-bytes because of alignment restrictions.
-        template<size_t SIZE>
-        using Pool = utils::PoolAllocator<SIZE, MIN_ALIGNMENT, sizeof(Node)>;
-        Pool<P0> mPool0;
-        Pool<P1> mPool1;
-        Pool<P2> mPool2;
+        utils::PoolAllocator<P0, 16>   mPool0;
+        utils::PoolAllocator<P1, 16>   mPool1;
+        utils::PoolAllocator<P2, 16>   mPool2;
        UTILS_UNUSED_IN_RELEASE const utils::AreaPolicy::HeapArea& mArea;
    public:
+        static constexpr size_t MIN_ALIGNMENT_SHIFT = 4;
        explicit Allocator(const utils::AreaPolicy::HeapArea& area);

-        static constexpr size_t getAlignment() noexcept { return MIN_ALIGNMENT; }
-
        // this is in fact always called with a constexpr size argument
-        [[nodiscard]] inline void* alloc(size_t size, size_t, size_t, uint8_t* outAge) noexcept {
+        [[nodiscard]] inline void* alloc(size_t size, size_t alignment, size_t extra) noexcept {
            void* p = nullptr;
-            if      (size <= mPool0.getSize()) p = mPool0.alloc(size);
-            else if (size <= mPool1.getSize()) p = mPool1.alloc(size);
-            else if (size <= mPool2.getSize()) p = mPool2.alloc(size);
-            if (UTILS_LIKELY(p)) {
-                Node const* const pNode = static_cast<Node const*>(p);
-                // we are guaranteed to have at least sizeof<Node> bytes of extra storage before
-                // the allocation address.
-                *outAge = pNode[-1].age;
-            }
+                 if (size <= mPool0.getSize()) p = mPool0.alloc(size, 16, extra);
+            else if (size <= mPool1.getSize()) p = mPool1.alloc(size, 16, extra);
+            else if (size <= mPool2.getSize()) p = mPool2.alloc(size, 16, extra);
            return p;
        }

        // this is in fact always called with a constexpr size argument
-        inline void free(void* p, size_t size, uint8_t age) noexcept {
+        inline void free(void* p, size_t size) noexcept {
            assert_invariant(p >= mArea.begin() && (char*)p + size <= (char*)mArea.end());
-
-            // check for double-free
-            Node* const pNode = static_cast<Node*>(p);
-            uint8_t& expectedAge = pNode[-1].age;
-            ASSERT_POSTCONDITION(expectedAge == age,
-                    "double-free of Handle of size %d at %p", size, p);
-            expectedAge = (expectedAge + 1) & 0xF; // fixme
-
            if (size <= mPool0.getSize()) { mPool0.free(p); return; }
            if (size <= mPool1.getSize()) { mPool1.free(p); return; }
            if (size <= mPool2.getSize()) { mPool2.free(p); return; }
@@ -262,16 +254,24 @@ private:
    // allocateHandle()/deallocateHandle() selects the pool to use at compile-time based on the
    // allocation size this is always inlined, because all these do is to call
    // allocateHandleInPool()/deallocateHandleFromPool() with the right pool size.
-    template<typename D>
+    template<size_t SIZE>
    HandleBase::HandleId allocateHandle() noexcept {
-        constexpr size_t BUCKET_SIZE = getBucketSize<D>();
-        return allocateHandleInPool<BUCKET_SIZE>();
+        if constexpr (SIZE <= P0) { return allocateHandleInPool<P0>(); }
+        if constexpr (SIZE <= P1) { return allocateHandleInPool<P1>(); }
+        static_assert(SIZE <= P2);
+        return allocateHandleInPool<P2>();
    }

-    template<typename D>
+    template<size_t SIZE>
    void deallocateHandle(HandleBase::HandleId id) noexcept {
-        constexpr size_t BUCKET_SIZE = getBucketSize<D>();
-        deallocateHandleFromPool<BUCKET_SIZE>(id);
+        if constexpr (SIZE <= P0) {
+            deallocateHandleFromPool<P0>(id);
+        } else if constexpr (SIZE <= P1) {
+            deallocateHandleFromPool<P1>(id);
+        } else {
+            static_assert(SIZE <= P2);
+            deallocateHandleFromPool<P2>(id);
+        }
    }

    // allocateHandleInPool()/deallocateHandleFromPool() is NOT inlined, which will cause three
@@ -280,11 +280,9 @@ private:
    template<size_t SIZE>
    UTILS_NOINLINE
    HandleBase::HandleId allocateHandleInPool() noexcept {
-        uint8_t age;
-        void* p = mHandleArena.alloc(SIZE, alignof(std::max_align_t), 0, &age);
+        void* p = mHandleArena.alloc(SIZE);
        if (UTILS_LIKELY(p)) {
-            uint32_t const tag = (uint32_t(age) << HANDLE_AGE_SHIFT) & HANDLE_AGE_MASK;
-            return arenaPointerToHandle(p, tag);
+            return pointerToHandle(p);
        } else {
            return allocateHandleSlow(SIZE);
        }
@@ -294,51 +292,42 @@ private:
    UTILS_NOINLINE
    void deallocateHandleFromPool(HandleBase::HandleId id) noexcept {
        if (UTILS_LIKELY(isPoolHandle(id))) {
-            auto [p, tag] = handleToPointer(id);
-            uint8_t const age = (tag & HANDLE_AGE_MASK) >> HANDLE_AGE_SHIFT;
-            mHandleArena.free(p, SIZE, age);
+            void* p = handleToPointer(id);
+            mHandleArena.free(p, SIZE);
        } else {
            deallocateHandleSlow(id, SIZE);
        }
    }

-    // we handle a 4 bits age per address
-    static constexpr uint32_t HANDLE_HEAP_FLAG      = 0x80000000u;      // pool vs heap handle
-    static constexpr uint32_t HANDLE_AGE_MASK       = 0x78000000u;      // handle's age
-    static constexpr uint32_t HANDLE_INDEX_MASK     = 0x07FFFFFFu;      // handle index
-    static constexpr uint32_t HANDLE_TAG_MASK       = HANDLE_AGE_MASK;
-    static constexpr uint32_t HANDLE_AGE_SHIFT      = 27;
+    static constexpr uint32_t HEAP_HANDLE_FLAG = 0x80000000u;

    static bool isPoolHandle(HandleBase::HandleId id) noexcept {
-        return (id & HANDLE_HEAP_FLAG) == 0u;
+        return (id & HEAP_HANDLE_FLAG) == 0u;
    }

    HandleBase::HandleId allocateHandleSlow(size_t size) noexcept;
    void deallocateHandleSlow(HandleBase::HandleId id, size_t size) noexcept;

    // We inline this because it's just 4 instructions in the fast case
-    inline std::pair<void*, uint32_t> handleToPointer(HandleBase::HandleId id) const noexcept {
+    inline void* handleToPointer(HandleBase::HandleId id) const noexcept {
        // note: the null handle will end-up returning nullptr b/c it'll be handled as
        // a non-pool handle.
        if (UTILS_LIKELY(isPoolHandle(id))) {
            char* const base = (char*)mHandleArena.getArea().begin();
-            uint32_t const tag = id & HANDLE_TAG_MASK;
-            size_t const offset = (id & HANDLE_INDEX_MASK) * Allocator::getAlignment();
-            return { static_cast<void*>(base + offset), tag };
+            size_t offset = id << Allocator::MIN_ALIGNMENT_SHIFT;
+            return static_cast<void*>(base + offset);
        }
-        return { handleToPointerSlow(id), 0 };
+        return handleToPointerSlow(id);
    }

    void* handleToPointerSlow(HandleBase::HandleId id) const noexcept;

    // We inline this because it's just 3 instructions
-    inline HandleBase::HandleId arenaPointerToHandle(void* p, uint32_t tag) const noexcept {
+    inline HandleBase::HandleId pointerToHandle(void* p) const noexcept {
        char* const base = (char*)mHandleArena.getArea().begin();
-        size_t const offset = (char*)p - base;
-        assert_invariant((offset % Allocator::getAlignment()) == 0);
-        auto id = HandleBase::HandleId(offset / Allocator::getAlignment());
-        id |= tag & HANDLE_TAG_MASK;
-        assert_invariant((id & HANDLE_HEAP_FLAG) == 0);
+        size_t offset = (char*)p - base;
+        auto id = HandleBase::HandleId(offset >> Allocator::MIN_ALIGNMENT_SHIFT);
+        assert_invariant((id & HEAP_HANDLE_FLAG) == 0);
        return id;
    }

@@ -348,6 +337,9 @@ private:
    mutable utils::Mutex mLock;
    tsl::robin_map<HandleBase::HandleId, void*> mOverflowMap;
    HandleBase::HandleId mId = 0;
+#if HANDLE_TYPE_SAFETY
+    mutable std::unordered_map<const void*, const char*> mHandleTypeId;
+#endif
 };

 } // namespace filament::backend
@@ -19,9 +19,9 @@
 #ifndef TNT_FILAMENT_BACKEND_PLATFORM_FACTORY_H
 #define TNT_FILAMENT_BACKEND_PLATFORM_FACTORY_H

-#include <backend/DriverEnums.h>
+#include "backend/DriverEnums.h"

-#include <utils/compiler.h>
+#include "utils/compiler.h"

 namespace filament::backend {

@@ -17,12 +17,14 @@
 #ifndef TNT_FILAMENT_BACKEND_PRIVATE_SAMPLERGROUP_H
 #define TNT_FILAMENT_BACKEND_PRIVATE_SAMPLERGROUP_H

-#include <backend/DriverApiForward.h>
-#include <backend/DriverEnums.h>
-#include <backend/SamplerDescriptor.h>
+#include "backend/DriverApiForward.h"

+#include <utils/compiler.h>
 #include <utils/FixedCapacityVector.h>
-#include <utils/ostream.h>
+
+#include <backend/DriverEnums.h>
+#include <backend/Handle.h>
+#include <backend/SamplerDescriptor.h>

 #include <stddef.h>

@@ -16,14 +16,6 @@

 #include "private/backend/CircularBuffer.h"

-#include <utils/Log.h>
-#include <utils/Panic.h>
-#include <utils/architecture.h>
-#include <utils/ashmem.h>
-#include <utils/compiler.h>
-#include <utils/debug.h>
-#include <utils/ostream.h>
-
 #if !defined(WIN32) && !defined(__EMSCRIPTEN__) && !defined(IOS)
 #    include <sys/mman.h>
 #    include <unistd.h>
@@ -32,20 +24,23 @@
 #    define HAS_MMAP 0
 #endif

-#include <stdint.h>
-#include <stddef.h>
-#include <stdlib.h>
 #include <stdio.h>

+#include <utils/architecture.h>
+#include <utils/ashmem.h>
+#include <utils/debug.h>
+#include <utils/Log.h>
+#include <utils/Panic.h>
+
 using namespace utils;

 namespace filament::backend {

 size_t CircularBuffer::sPageSize = arch::getPageSize();

-CircularBuffer::CircularBuffer(size_t size)
-    : mSize(size) {
+CircularBuffer::CircularBuffer(size_t size) {
    mData = alloc(size);
+    mSize = size;
    mTail = mData;
    mHead = mData;
 }
@@ -90,7 +85,7 @@ void* CircularBuffer::alloc(size_t size) noexcept {
                            MAP_PRIVATE, fd, (off_t)size);
                    if (vaddr_guard != MAP_FAILED && (vaddr_guard == (char*)vaddr_shadow + size)) {
                        // woo-hoo success!
-                        mAshmemFd = fd;
+                        mUsesAshmem = fd;
                        data = vaddr;
                    }
                }
@@ -98,7 +93,7 @@ void* CircularBuffer::alloc(size_t size) noexcept {
        }
    }

-    if (UTILS_UNLIKELY(mAshmemFd < 0)) {
+    if (UTILS_UNLIKELY(mUsesAshmem < 0)) {
        // ashmem failed
        if (vaddr_guard != MAP_FAILED) {
            munmap(vaddr_guard, size);
@@ -142,9 +137,9 @@ void CircularBuffer::dealloc() noexcept {
    if (mData) {
        size_t const BLOCK_SIZE = getBlockSize();
        munmap(mData, mSize * 2 + BLOCK_SIZE);
-        if (mAshmemFd >= 0) {
-            close(mAshmemFd);
-            mAshmemFd = -1;
+        if (mUsesAshmem >= 0) {
+            close(mUsesAshmem);
+            mUsesAshmem = -1;
        }
    }
 #else
@@ -154,37 +149,23 @@ void CircularBuffer::dealloc() noexcept {
 }


-CircularBuffer::Range CircularBuffer::getBuffer() noexcept {
-    Range const range{ .tail = mTail, .head = mHead };
-
-    char* const pData = static_cast<char*>(mData);
-    char const* const pEnd = pData + mSize;
-    char const* const pHead = static_cast<char const*>(mHead);
-    if (UTILS_UNLIKELY(pHead >= pEnd)) {
-        size_t const overflow = pHead - pEnd;
-        if (UTILS_LIKELY(mAshmemFd > 0)) {
-            assert_invariant(overflow <= mSize);
-            mHead = static_cast<void*>(pData + overflow);
-            // Data         Tail  End   Head              [virtual]
-            //  v             v    v     v
-            //  +-------------:----+-----:--------------+
-            //  |             :    |     :              |
-            //  +-----:------------+--------------------+
-            //       Head          |<------ copy ------>| [physical]
-        } else {
-            // Data         Tail  End   Head
-            //  v             v    v     v
-            //  +-------------:----+-----:--------------+
-            //  |             :    |     :              |
-            //  +-----|------------+-----|--------------+
-            //        |<---------------->|
-            //           sliding window
+void CircularBuffer::circularize() noexcept {
+    if (mUsesAshmem > 0) {
+        intptr_t const overflow = intptr_t(mHead) - (intptr_t(mData) + ssize_t(mSize));
+        if (overflow >= 0) {
+            assert_invariant(size_t(overflow) <= mSize);
+            mHead = (void *) (intptr_t(mData) + overflow);
+            #ifndef NDEBUG
+            memset(mData, 0xA5, size_t(overflow));
+            #endif
+        }
+    } else {
+        // Only circularize if mHead if in the second buffer.
+        if (intptr_t(mHead) - intptr_t(mData) > ssize_t(mSize)) {
            mHead = mData;
        }
    }
    mTail = mHead;
-
-    return range;
 }

 } // namespace filament::backend
@@ -15,25 +15,14 @@
 */

 #include "private/backend/CommandBufferQueue.h"
-#include "private/backend/CircularBuffer.h"
-#include "private/backend/CommandStream.h"

-#include <utils/compiler.h>
 #include <utils/Log.h>
-#include <utils/Mutex.h>
-#include <utils/ostream.h>
-#include <utils/Panic.h>
 #include <utils/Systrace.h>
+#include <utils/Panic.h>
 #include <utils/debug.h>

-#include <algorithm>
-#include <mutex>
-#include <iterator>
-#include <utility>
-#include <vector>
-
-#include <stddef.h>
-#include <stdint.h>
+#include "private/backend/BackendUtils.h"
+#include "private/backend/CommandStream.h"

 using namespace utils;

@@ -76,53 +65,50 @@ void CommandBufferQueue::flush() noexcept {
    // always guaranteed to have enough space for the NoopCommand
    new(circularBuffer.allocate(sizeof(NoopCommand))) NoopCommand(nullptr);

-    const size_t requiredSize = mRequiredSize;
+    // end of this slice
+    void* const head = circularBuffer.getHead();

-    // get the current buffer
-    auto const [begin, end] = circularBuffer.getBuffer();
+    // beginning of this slice
+    void* const tail = circularBuffer.getTail();

-    assert_invariant(circularBuffer.empty());
+    // size of this slice
+    uint32_t const used = uint32_t(intptr_t(head) - intptr_t(tail));

-    // size of the current buffer
-    size_t const used = std::distance(
-            static_cast<char const*>(begin), static_cast<char const*>(end));
+    circularBuffer.circularize();

    std::unique_lock<utils::Mutex> lock(mLock);
-    mCommandBuffersToExecute.push_back({ begin, end });
-    mCondition.notify_one();
+    mCommandBuffersToExecute.push_back({ tail, head });

    // circular buffer is too small, we corrupted the stream
    ASSERT_POSTCONDITION(used <= mFreeSpace,
            "Backend CommandStream overflow. Commands are corrupted and unrecoverable.\n"
            "Please increase minCommandBufferSizeMB inside the Config passed to Engine::create.\n"
-            "Space used at this time: %u bytes, overflow: %u bytes",
-            (unsigned)used, unsigned(used - mFreeSpace));
+            "Space used at this time: %u bytes",
+            (unsigned)used);

    // wait until there is enough space in the buffer
    mFreeSpace -= used;
-    if (UTILS_UNLIKELY(mFreeSpace < requiredSize)) {
-
+    const size_t requiredSize = mRequiredSize;

 #ifndef NDEBUG
-        size_t const totalUsed = circularBuffer.size() - mFreeSpace;
-        slog.d << "CommandStream used too much space (will block): "
-                << "needed space " << requiredSize << " out of " << mFreeSpace
-                << ", totalUsed=" << totalUsed << ", current=" << used
-                << ", queue size=" << mCommandBuffersToExecute.size() << " buffers"
-                << io::endl;
-
-        mHighWatermark = std::max(mHighWatermark, totalUsed);
+    size_t totalUsed = circularBuffer.size() - mFreeSpace;
+    mHighWatermark = std::max(mHighWatermark, totalUsed);
+    if (UTILS_UNLIKELY(totalUsed > requiredSize)) {
+        slog.d << "CommandStream used too much space: " << totalUsed
+            << ", out of " << requiredSize << " (will block)" << io::endl;
+    }
 #endif

+    mCondition.notify_one();
+    if (UTILS_LIKELY(mFreeSpace < requiredSize)) {
        SYSTRACE_NAME("waiting: CircularBuffer::flush()");
        mCondition.wait(lock, [this, requiredSize]() -> bool {
-            // TODO: on macOS, we need to call pumpEvents from time to time
            return mFreeSpace >= requiredSize;
        });
    }
 }

-std::vector<CommandBufferQueue::Range> CommandBufferQueue::waitForCommands() const {
+std::vector<CommandBufferQueue::Slice> CommandBufferQueue::waitForCommands() const {
    if (!UTILS_HAS_THREADING) {
        return std::move(mCommandBuffersToExecute);
    }
@@ -137,7 +123,7 @@ std::vector<CommandBufferQueue::Range> CommandBufferQueue::waitForCommands() con
    return std::move(mCommandBuffersToExecute);
 }

-void CommandBufferQueue::releaseBuffer(CommandBufferQueue::Range const& buffer) {
+void CommandBufferQueue::releaseBuffer(CommandBufferQueue::Slice const& buffer) {
    std::lock_guard<utils::Mutex> const lock(mLock);
    mFreeSpace += uintptr_t(buffer.end) - uintptr_t(buffer.begin);
    mCondition.notify_one();
@@ -19,7 +19,6 @@

 #include <backend/PixelBufferDescriptor.h>

-#include <cstring>
 #include <stddef.h>
 #include <stdint.h>

@@ -21,12 +21,7 @@

 #include <backend/AcquiredImage.h>
 #include <backend/BufferDescriptor.h>
-#include <backend/DriverEnums.h>

-#include <utils/compiler.h>
-#include <utils/debug.h>
-#include <utils/Log.h>
-#include <utils/ostream.h>
 #include <utils/Systrace.h>

 #include <math/half.h>
@@ -34,13 +29,6 @@
 #include <math/vec3.h>
 #include <math/vec4.h>

-#include <functional>
-#include <mutex>
-#include <utility>
-
-#include <stddef.h>
-#include <stdint.h>
-
 using namespace utils;
 using namespace filament::math;

@@ -131,8 +119,7 @@ void DriverBase::purge() noexcept {
 // ------------------------------------------------------------------------------------------------

 void DriverBase::scheduleDestroySlow(BufferDescriptor&& buffer) noexcept {
-    auto const handler = buffer.getHandler();
-    scheduleCallback(handler, [buffer = std::move(buffer)]() {
+    scheduleCallback(buffer.getHandler(), [buffer = std::move(buffer)]() {
        // user callback is called when BufferDescriptor gets destroyed
    });
 }
@@ -49,25 +49,21 @@ struct AcquiredImage;
 struct HwBase {
 };

-
-struct HwVertexBufferInfo : public HwBase {
+struct HwVertexBuffer : public HwBase {
+    AttributeArray attributes{};          // 8 * MAX_VERTEX_ATTRIBUTE_COUNT
+    uint32_t vertexCount{};               //   4
    uint8_t bufferCount{};                //   1
    uint8_t attributeCount{};             //   1
-    bool padding[2]{};                    //   2
-    HwVertexBufferInfo() noexcept = default;
-    HwVertexBufferInfo(uint8_t bufferCount, uint8_t attributeCount) noexcept
-            : bufferCount(bufferCount),
-              attributeCount(attributeCount) {
-    }
-};
+    bool padding{};                       //   1
+    uint8_t bufferObjectsVersion{};       //   1 -> total struct is 136 bytes

-struct HwVertexBuffer : public HwBase {
-    uint32_t vertexCount{};               //   4
-    uint8_t bufferObjectsVersion{0xff};   //   1
-    bool padding[3]{};                    //   2
    HwVertexBuffer() noexcept = default;
-    explicit HwVertexBuffer(uint32_t vertextCount) noexcept
-            : vertexCount(vertextCount) {
+    HwVertexBuffer(uint8_t bufferCount, uint8_t attributeCount, uint32_t elementCount,
+            AttributeArray const& attributes) noexcept
+            : attributes(attributes),
+              vertexCount(elementCount),
+              bufferCount(bufferCount),
+              attributeCount(attributeCount) {
    }
 };

@@ -92,6 +88,11 @@ struct HwIndexBuffer : public HwBase {
 };

 struct HwRenderPrimitive : public HwBase {
+    uint32_t offset{};
+    uint32_t minIndex{};
+    uint32_t maxIndex{};
+    uint32_t count{};
+    uint32_t maxVertexCount{};
    PrimitiveType type = PrimitiveType::TRIANGLES;
 };

@@ -16,22 +16,9 @@

 #include "private/backend/HandleAllocator.h"

-#include <backend/Handle.h>
-
-#include <utils/Allocator.h>
-#include <utils/Log.h>
 #include <utils/Panic.h>
-#include <utils/compiler.h>
-#include <utils/debug.h>
-#include <utils/ostream.h>
-
-#include <algorithm>
-#include <exception>
-#include <limits>
-#include <mutex>

 #include <stdlib.h>
-#include <string.h>

 namespace filament::backend {

@@ -41,34 +28,14 @@ template <size_t P0, size_t P1, size_t P2>
 UTILS_NOINLINE
 HandleAllocator<P0, P1, P2>::Allocator::Allocator(AreaPolicy::HeapArea const& area)
        : mArea(area) {
-
-    // The largest handle this allocator can generate currently depends on the architecture's
-    // min alignment, typically 8 or 16 bytes.
-    // e.g. On Android armv8, the alignment is 16 bytes, so for a 1 MiB heap, the largest handle
-    //      index will be 65536. Note that this is not the same as the number of handles (which
-    //      will always be less).
-    // Because our maximum representable handle currently is 0x07FFFFFF, the maximum no-nonsensical
-    // heap size is 2 GiB, which amounts to 7.6 millions handles per pool (in the GL case).
-    size_t const maxHeapSize = std::min(area.size(), HANDLE_INDEX_MASK * getAlignment());
-
-    if (UTILS_UNLIKELY(maxHeapSize != area.size())) {
-        slog.w << "HandleAllocator heap size reduced to "
-               << maxHeapSize << " from " << area.size() << io::endl;
-    }
-
-    // make sure we start with a clean arena. This is needed to ensure that all blocks start
-    // with an age of 0.
-    memset(area.data(), 0, maxHeapSize);
-
-    // size the different pools so that they can all contain the same number of handles
-    size_t const count = maxHeapSize / (P0 + P1 + P2);
-    char* const p0 = static_cast<char*>(area.begin());
-    char* const p1 = p0 + count * P0;
-    char* const p2 = p1 + count * P1;
-
-    mPool0 = Pool<P0>(p0, count * P0);
-    mPool1 = Pool<P1>(p1, count * P1);
-    mPool2 = Pool<P2>(p2, count * P2);
+    // TODO: we probably need a better way to set the size of these pools
+    const size_t unit = area.size() / 32;
+    const size_t offsetPool1 =      unit;
+    const size_t offsetPool2 = 16 * unit;
+    char* const p = (char*)area.begin();
+    mPool0 = PoolAllocator< P0, 16>(p, p + offsetPool1);
+    mPool1 = PoolAllocator< P1, 16>(p + offsetPool1, p + offsetPool2);
+    mPool2 = PoolAllocator< P2, 16>(p + offsetPool2, area.end());
 }

 // ------------------------------------------------------------------------------------------------
@@ -106,17 +73,11 @@ template <size_t P0, size_t P1, size_t P2>
 HandleBase::HandleId HandleAllocator<P0, P1, P2>::allocateHandleSlow(size_t size) noexcept {
    void* p = ::malloc(size);
    std::unique_lock lock(mLock);
-
-    HandleBase::HandleId id = (++mId) | HANDLE_HEAP_FLAG;
-
-    ASSERT_POSTCONDITION(mId < HANDLE_HEAP_FLAG,
-            "No more Handle ids available! This can happen if HandleAllocator arena has been full"
-            " for a while. Please increase FILAMENT_OPENGL_HANDLE_ARENA_SIZE_IN_MB");
-
+    HandleBase::HandleId id = (++mId) | HEAP_HANDLE_FLAG;
    mOverflowMap.emplace(id, p);
    lock.unlock();

-    if (UTILS_UNLIKELY(id == (HANDLE_HEAP_FLAG | 1u))) { // meaning id was zero
+    if (UTILS_UNLIKELY(id == (HEAP_HANDLE_FLAG|1u))) { // meaning id was zero
        PANIC_LOG("HandleAllocator arena is full, using slower system heap. Please increase "
                  "the appropriate constant (e.g. FILAMENT_OPENGL_HANDLE_ARENA_SIZE_IN_MB).");
    }
@@ -125,7 +86,7 @@ HandleBase::HandleId HandleAllocator<P0, P1, P2>::allocateHandleSlow(size_t size

 template <size_t P0, size_t P1, size_t P2>
 void HandleAllocator<P0, P1, P2>::deallocateHandleSlow(HandleBase::HandleId id, size_t) noexcept {
-    assert_invariant(id & HANDLE_HEAP_FLAG);
+    assert_invariant(id & HEAP_HANDLE_FLAG);
    void* p = nullptr;
    auto& overflowMap = mOverflowMap;

@@ -20,7 +20,6 @@
 #include "MetalUtils.h"

 #include <utils/Panic.h>
-#include <utils/Log.h>

 namespace filament::backend {

@@ -19,7 +19,6 @@
 #include "MetalContext.h"

 #include <utils/Panic.h>
-#include <utils/Log.h>
 #include <utils/trap.h>

 #include <thread>
@@ -133,9 +133,12 @@ private:
        mHandleAllocator.deallocate(handle, p);
    }

-    inline void setRenderPrimitiveBuffer(Handle<HwRenderPrimitive> rph, PrimitiveType pt,
+    inline void setRenderPrimitiveBuffer(Handle<HwRenderPrimitive> rph,
            Handle<HwVertexBuffer> vbh, Handle<HwIndexBuffer> ibh);

+    inline void setRenderPrimitiveRange(Handle<HwRenderPrimitive> rph, PrimitiveType pt,
+            uint32_t offset, uint32_t minIndex, uint32_t maxIndex, uint32_t count);
+
    void finalizeSamplerGroup(MetalSamplerGroup* sg);
    void enumerateBoundBuffers(BufferObjectBinding bindingType,
            const std::function<void(const BufferState&, MetalBuffer*, uint32_t)>& f);
@@ -43,40 +43,6 @@ namespace filament {
 namespace backend {

 Driver* MetalDriverFactory::create(MetalPlatform* const platform, const Platform::DriverConfig& driverConfig) {
-#if 0
-    // this is useful for development, but too verbose even for debug builds
-    // For reference on a 64-bits machine in Release mode:
-    //    MetalTimerQuery              :  16       few
-    //    HwStream                     :  24       few
-    //    MetalIndexBuffer             :  40       moderate
-    //    MetalFence                   :  48       few
-    //    MetalBufferObject            :  48       many
-    // -- less than or equal 48 bytes
-    //    MetalSamplerGroup            : 112       few
-    //    MetalProgram                 : 144       moderate
-    //    MetalTexture                 : 152       moderate
-    //    MetalVertexBuffer            : 152       moderate
-    // -- less than or equal 160 bytes
-    //    MetalSwapChain               : 184       few
-    //    MetalRenderTarget            : 272       few
-    //    MetalRenderPrimitive         : 584       many
-    // -- less than or equal to 592 bytes
-
-    utils::slog.d
-           << "\nMetalSwapChain: " << sizeof(MetalSwapChain)
-           << "\nMetalBufferObject: " << sizeof(MetalBufferObject)
-           << "\nMetalVertexBuffer: " << sizeof(MetalVertexBuffer)
-           << "\nMetalIndexBuffer: " << sizeof(MetalIndexBuffer)
-           << "\nMetalSamplerGroup: " << sizeof(MetalSamplerGroup)
-           << "\nMetalRenderPrimitive: " << sizeof(MetalRenderPrimitive)
-           << "\nMetalTexture: " << sizeof(MetalTexture)
-           << "\nMetalTimerQuery: " << sizeof(MetalTimerQuery)
-           << "\nHwStream: " << sizeof(HwStream)
-           << "\nMetalRenderTarget: " << sizeof(MetalRenderTarget)
-           << "\nMetalFence: " << sizeof(MetalFence)
-           << "\nMetalProgram: " << sizeof(MetalProgram)
-           << utils::io::endl;
-#endif
    return MetalDriver::create(platform, driverConfig);
 }

@@ -279,16 +245,10 @@ void MetalDriver::finish(int) {
    [oneOffBuffer waitUntilCompleted];
 }

-void MetalDriver::createVertexBufferInfoR(Handle<HwVertexBufferInfo> vbih, uint8_t bufferCount,
-        uint8_t attributeCount, AttributeArray attributes) {
-    construct_handle<MetalVertexBufferInfo>(vbih, *mContext,
-            bufferCount, attributeCount, attributes);
-}
-
-void MetalDriver::createVertexBufferR(Handle<HwVertexBuffer> vbh,
-        uint32_t vertexCount, Handle<HwVertexBufferInfo> vbih) {
-    MetalVertexBufferInfo const* const vbi = handle_cast<const MetalVertexBufferInfo>(vbih);
-    construct_handle<MetalVertexBuffer>(vbh, *mContext, vertexCount, vbi->bufferCount, vbih);
+void MetalDriver::createVertexBufferR(Handle<HwVertexBuffer> vbh, uint8_t bufferCount,
+        uint8_t attributeCount, uint32_t vertexCount, AttributeArray attributes) {
+    construct_handle<MetalVertexBuffer>(vbh, *mContext, bufferCount,
+            attributeCount, vertexCount, attributes);
 }

 void MetalDriver::createIndexBufferR(Handle<HwIndexBuffer> ibh, ElementType elementType,
@@ -356,9 +316,11 @@ void MetalDriver::createSamplerGroupR(

 void MetalDriver::createRenderPrimitiveR(Handle<HwRenderPrimitive> rph,
        Handle<HwVertexBuffer> vbh, Handle<HwIndexBuffer> ibh,
-        PrimitiveType pt) {
+        PrimitiveType pt, uint32_t offset,
+        uint32_t minIndex, uint32_t maxIndex, uint32_t count) {
    construct_handle<MetalRenderPrimitive>(rph);
-    MetalDriver::setRenderPrimitiveBuffer(rph, pt, vbh, ibh);
+    MetalDriver::setRenderPrimitiveBuffer(rph, vbh, ibh);
+    MetalDriver::setRenderPrimitiveRange(rph, pt, offset, minIndex, maxIndex, count);
 }

 void MetalDriver::createProgramR(Handle<HwProgram> rph, Program&& program) {
@@ -444,10 +406,6 @@ void MetalDriver::createTimerQueryR(Handle<HwTimerQuery> tqh, int) {
    // nothing to do, timer query was constructed in createTimerQueryS
 }

-Handle<HwVertexBufferInfo> MetalDriver::createVertexBufferInfoS() noexcept {
-    return alloc_handle<MetalVertexBufferInfo>();
-}
-
 Handle<HwVertexBuffer> MetalDriver::createVertexBufferS() noexcept {
    return alloc_handle<MetalVertexBuffer>();
 }
@@ -512,12 +470,6 @@ Handle<HwTimerQuery> MetalDriver::createTimerQueryS() noexcept {
    return alloc_and_construct_handle<MetalTimerQuery, HwTimerQuery>();
 }

-void MetalDriver::destroyVertexBufferInfo(Handle<HwVertexBufferInfo> vbih) {
-    if (vbih) {
-        destruct_handle<MetalVertexBufferInfo>(vbih);
-    }
-}
-
 void MetalDriver::destroyVertexBuffer(Handle<HwVertexBuffer> vbh) {
    if (vbh) {
        destruct_handle<MetalVertexBuffer>(vbh);
@@ -770,14 +722,8 @@ bool MetalDriver::isSRGBSwapChainSupported() {
    return false;
 }

-bool MetalDriver::isStereoSupported(backend::StereoscopicType stereoscopicType) {
-    switch (stereoscopicType) {
-    case backend::StereoscopicType::INSTANCED:
-        return true;
-    case backend::StereoscopicType::MULTIVIEW:
-        // TODO: implement multiview feature in Metal.
-        return false;
-    }
+bool MetalDriver::isStereoSupported() {
+    return true;
 }

 bool MetalDriver::isParallelShaderCompileSupported() {
@@ -1083,14 +1029,23 @@ void MetalDriver::endRenderPass(int dummy) {
    mContext->currentRenderPassEncoder = nil;
 }

-void MetalDriver::setRenderPrimitiveBuffer(Handle<HwRenderPrimitive> rph, PrimitiveType pt,
+void MetalDriver::setRenderPrimitiveBuffer(Handle<HwRenderPrimitive> rph,
        Handle<HwVertexBuffer> vbh, Handle<HwIndexBuffer> ibh) {
    auto primitive = handle_cast<MetalRenderPrimitive>(rph);
    auto vertexBuffer = handle_cast<MetalVertexBuffer>(vbh);
    auto indexBuffer = handle_cast<MetalIndexBuffer>(ibh);
-    MetalVertexBufferInfo const* const vbi = handle_cast<MetalVertexBufferInfo>(vertexBuffer->vbih);
-    primitive->setBuffers(vbi, vertexBuffer, indexBuffer);
+    primitive->setBuffers(vertexBuffer, indexBuffer);
+}
+
+void MetalDriver::setRenderPrimitiveRange(Handle<HwRenderPrimitive> rph,
+        PrimitiveType pt, uint32_t offset, uint32_t minIndex, uint32_t maxIndex,
+        uint32_t count) {
+    auto primitive = handle_cast<MetalRenderPrimitive>(rph);
    primitive->type = pt;
+    primitive->offset = offset * primitive->indexBuffer->elementSize;
+    primitive->count = count;
+    primitive->minIndex = minIndex;
+    primitive->maxIndex = maxIndex > minIndex ? maxIndex : primitive->maxVertexCount - 1;
 }

 void MetalDriver::makeCurrent(Handle<HwSwapChain> schDraw, Handle<HwSwapChain> schRead) {
@@ -1592,8 +1547,7 @@ void MetalDriver::finalizeSamplerGroup(MetalSamplerGroup* samplerGroup) {
    }
 }

-void MetalDriver::draw(PipelineState ps, Handle<HwRenderPrimitive> rph,
-        uint32_t const indexOffset, uint32_t const indexCount, uint32_t const instanceCount) {
+void MetalDriver::draw(PipelineState ps, Handle<HwRenderPrimitive> rph, uint32_t instanceCount) {
    ASSERT_PRECONDITION(mContext->currentRenderPassEncoder != nullptr,
            "Attempted to draw without a valid command encoder.");
    auto primitive = handle_cast<MetalRenderPrimitive>(rph);
@@ -1610,7 +1564,7 @@ void MetalDriver::draw(PipelineState ps, Handle<HwRenderPrimitive> rph,
        return;
    }

-    functions.validate();
+    ASSERT_PRECONDITION(bool(functions), "Attempting to draw with an invalid Metal program.");

    auto [fragment, vertex] = functions.getRasterFunctions();

@@ -1731,6 +1685,36 @@ void MetalDriver::draw(PipelineState ps, Handle<HwRenderPrimitive> rph,
        mContext->currentPolygonOffset = ps.polygonOffset;
    }

+    // Set scissor-rectangle.
+    // In order to do this, we compute the intersection between:
+    //  1. the scissor rectangle
+    //  2. the render target attachment dimensions (important, as the scissor can't be set larger)
+    // fmax/min are used below to guard against NaN and because the MTLViewport/MTLRegion
+    // coordinates are doubles.
+    MTLRegion scissor = mContext->currentRenderTarget->getRegionFromClientRect(ps.scissor);
+    const float sleft = scissor.origin.x, sright = scissor.origin.x + scissor.size.width;
+    const float stop = scissor.origin.y, sbottom = scissor.origin.y + scissor.size.height;
+
+    // Attachment extent
+    const auto attachmentSize = mContext->currentRenderTarget->getAttachmentSize();
+    const float aleft = 0.0f, atop = 0.0f;
+    const float aright = static_cast<float>(attachmentSize.x);
+    const float abottom = static_cast<float>(attachmentSize.y);
+
+    const auto left   = std::fmax(sleft, aleft);
+    const auto right  = std::fmin(sright, aright);
+    const auto top    = std::fmax(stop, atop);
+    const auto bottom = std::fmin(sbottom, abottom);
+
+    MTLScissorRect scissorRect = {
+        .x      = static_cast<NSUInteger>(left),
+        .y      = static_cast<NSUInteger>(top),
+        .width  = static_cast<NSUInteger>(right  - left),
+        .height = static_cast<NSUInteger>(bottom - top)
+    };
+
+    [mContext->currentRenderPassEncoder setScissorRect:scissorRect];
+
    // Bind uniform buffers.
    MetalBuffer* uniformsToBind[Program::UNIFORM_BINDING_COUNT] = { nil };
    NSUInteger offsets[Program::UNIFORM_BINDING_COUNT] = { 0 };
@@ -1808,10 +1792,10 @@ void MetalDriver::draw(PipelineState ps, Handle<HwRenderPrimitive> rph,
    id<MTLCommandBuffer> cmdBuffer = getPendingCommandBuffer(mContext);
    id<MTLBuffer> metalIndexBuffer = indexBuffer->buffer.getGpuBufferForDraw(cmdBuffer);
    [mContext->currentRenderPassEncoder drawIndexedPrimitives:getMetalPrimitiveType(primitive->type)
-                                                   indexCount:indexCount
+                                                   indexCount:primitive->count
                                                    indexType:getIndexType(indexBuffer->elementSize)
                                                  indexBuffer:metalIndexBuffer
-                                            indexBufferOffset:indexOffset * primitive->indexBuffer->elementSize
+                                            indexBufferOffset:primitive->offset
                                                instanceCount:instanceCount];
 }

@@ -1884,38 +1868,6 @@ void MetalDriver::dispatchCompute(Handle<HwProgram> program, math::uint3 workGro
    [computeEncoder endEncoding];
 }

-void MetalDriver::scissor(Viewport scissorBox) {
-    // Set scissor-rectangle.
-    // In order to do this, we compute the intersection between:
-    //  1. the scissor rectangle
-    //  2. the render target attachment dimensions (important, as the scissor can't be set larger)
-    // fmax/min are used below to guard against NaN and because the MTLViewport/MTLRegion
-    // coordinates are doubles.
-    MTLRegion scissor = mContext->currentRenderTarget->getRegionFromClientRect(scissorBox);
-    const float sleft = scissor.origin.x, sright = scissor.origin.x + scissor.size.width;
-    const float stop = scissor.origin.y, sbottom = scissor.origin.y + scissor.size.height;
-
-    // Attachment extent
-    const auto attachmentSize = mContext->currentRenderTarget->getAttachmentSize();
-    const float aleft = 0.0f, atop = 0.0f;
-    const float aright = static_cast<float>(attachmentSize.x);
-    const float abottom = static_cast<float>(attachmentSize.y);
-
-    const auto left   = std::fmax(sleft, aleft);
-    const auto right  = std::fmin(sright, aright);
-    const auto top    = std::fmax(stop, atop);
-    const auto bottom = std::fmin(sbottom, abottom);
-
-    MTLScissorRect scissorRect = {
-            .x      = static_cast<NSUInteger>(left),
-            .y      = static_cast<NSUInteger>(top),
-            .width  = static_cast<NSUInteger>(right  - left),
-            .height = static_cast<NSUInteger>(bottom - top)
-    };
-
-    [mContext->currentRenderPassEncoder setScissorRect:scissorRect];
-}
-
 void MetalDriver::beginTimerQuery(Handle<HwTimerQuery> tqh) {
    ASSERT_PRECONDITION(!isInRenderPass(mContext),
            "beginTimerQuery must be called outside of a render pass.");
@@ -21,7 +21,6 @@
 #include "MetalUtils.h"

 #include <utils/Panic.h>
-#include <utils/Log.h>
 #include <utils/trap.h>

 #define NSERROR_CHECK(message)                                                                     \
@@ -145,18 +145,10 @@ private:
    MetalBuffer buffer;
 };

-struct MetalVertexBufferInfo : public HwVertexBufferInfo {
-    MetalVertexBufferInfo(MetalContext& context,
-            uint8_t bufferCount, uint8_t attributeCount, AttributeArray const& attributes);
-
-    AttributeArray attributes;
-};
-
 struct MetalVertexBuffer : public HwVertexBuffer {
-    MetalVertexBuffer(MetalContext& context,
-            uint32_t vertexCount, uint32_t bufferCount, Handle<HwVertexBufferInfo> vbih);
+    MetalVertexBuffer(MetalContext& context, uint8_t bufferCount, uint8_t attributeCount,
+            uint32_t vertexCount, AttributeArray const& attributes);

-    Handle<HwVertexBufferInfo> vbih;
    utils::FixedCapacityVector<MetalBuffer*> buffers;
 };

@@ -169,8 +161,7 @@ struct MetalIndexBuffer : public HwIndexBuffer {

 struct MetalRenderPrimitive : public HwRenderPrimitive {
    MetalRenderPrimitive();
-    void setBuffers(MetalVertexBufferInfo const* const vbi,
-            MetalVertexBuffer* vertexBuffer, MetalIndexBuffer* indexBuffer);
+    void setBuffers(MetalVertexBuffer* vertexBuffer, MetalIndexBuffer* indexBuffer);
    // The pointers to MetalVertexBuffer and MetalIndexBuffer are "weak".
    // The MetalVertexBuffer and MetalIndexBuffer must outlive the MetalRenderPrimitive.

@@ -245,53 +245,30 @@ void MetalSwapChain::present() {
    }
 }

-#ifndef FILAMENT_RELEASE_PRESENT_DRAWABLE_MAIN_THREAD
-#define FILAMENT_RELEASE_PRESENT_DRAWABLE_MAIN_THREAD 1
-#endif
-
-class PresentDrawableData {
-public:
-    PresentDrawableData() = delete;
-    PresentDrawableData(const PresentDrawableData&) = delete;
-    PresentDrawableData& operator=(const PresentDrawableData&) = delete;
-
-    static PresentDrawableData* create(id<CAMetalDrawable> drawable, MetalDriver* driver) {
-        assert_invariant(driver);
-        return new PresentDrawableData(drawable, driver);
-    }
-
-    static void maybePresentAndDestroyAsync(PresentDrawableData* that, bool shouldPresent) {
-        if (shouldPresent) {
-           [that->mDrawable present];
-        }
-
-#if FILAMENT_RELEASE_PRESENT_DRAWABLE_MAIN_THREAD == 1
-        // mDrawable is acquired on the driver thread. Typically, we would release this object on
-        // the same thread, but after receiving consistent crash reports from within
-        // [CAMetalDrawable dealloc], we suspect this object requires releasing on the main thread.
-        dispatch_async(dispatch_get_main_queue(), ^{ cleanupAndDestroy(that); });
-#else
-        that->mDriver->runAtNextTick([that]() { cleanupAndDestroy(that); });
-#endif
-    }
-
-private:
-    PresentDrawableData(id<CAMetalDrawable> drawable, MetalDriver* driver)
-        : mDrawable(drawable), mDriver(driver) {}
-
-    static void cleanupAndDestroy(PresentDrawableData *that) {
-        that->mDrawable = nil;
-        that->mDriver = nullptr;
-        delete that;
-    }
-
-    id<CAMetalDrawable> mDrawable;
-    MetalDriver* mDriver = nullptr;
+struct PresentDrawableData {
+    void* drawable = nullptr;
+    MetalDriver* driver = nullptr;
 };

 void presentDrawable(bool presentFrame, void* user) {
    auto* presentDrawableData = static_cast<PresentDrawableData*>(user);
-    PresentDrawableData::maybePresentAndDestroyAsync(presentDrawableData, presentFrame);
+
+    // CFBridgingRelease here is used to balance the CFBridgingRetain inside acquireDrawable.
+    id<CAMetalDrawable> drawable =
+            (id<CAMetalDrawable>)CFBridgingRelease(presentDrawableData->drawable);
+    if (presentFrame) {
+        [drawable present];
+    }
+
+    // Schedule the drawable destruction on the driver thread.
+    void* voidDrawable = (void*) CFBridgingRetain(drawable);
+    MetalDriver* driver = presentDrawableData->driver;
+    driver->runAtNextTick([voidDrawable]() {
+        // The drawable is released here.
+        CFBridgingRelease(voidDrawable);
+    });
+
+    delete presentDrawableData;
 }

 void MetalSwapChain::scheduleFrameScheduledCallback() {
@@ -300,16 +277,21 @@ void MetalSwapChain::scheduleFrameScheduledCallback() {
    }

    assert_invariant(drawable);
-
-    // Destroy this by calling maybePresentAndDestroyAsync() later.
-    auto* presentData = PresentDrawableData::create(drawable, context.driver);
-
-    FrameScheduledCallback userCallback = frameScheduledCallback;
+    FrameScheduledCallback callback = frameScheduledCallback;
+    // This block strongly captures drawable to keep it alive until the handler executes.
+    // We cannot simply reference this->drawable inside the block because the block would then only
+    // capture the _this_ pointer (MetalSwapChain*) instead of the drawable.
+    id<CAMetalDrawable> d = drawable;
    void* userData = frameScheduledUserData;
-
+    MetalDriver* driver = context.driver;
    [getPendingCommandBuffer(&context) addScheduledHandler:^(id<MTLCommandBuffer> cb) {
-        PresentCallable callable(presentDrawable, static_cast<void*>(presentData));
-        userCallback(callable, userData);
+        // CFBridgingRetain is used here to give the drawable a +1 retain count before
+        // casting it to a void*.
+        auto* presentDrawableData = new PresentDrawableData;
+        presentDrawableData->drawable = (void*) CFBridgingRetain(d);
+        presentDrawableData->driver = driver;
+        PresentCallable callable(presentDrawable, (void*) presentDrawableData);
+        callback(callable, userData);
    }];
 }

@@ -340,16 +322,9 @@ void MetalBufferObject::updateBufferUnsynchronized(void* data, size_t size, uint
    buffer.copyIntoBufferUnsynchronized(data, size, byteOffset);
 }

-MetalVertexBufferInfo::MetalVertexBufferInfo(MetalContext& context, uint8_t bufferCount,
-        uint8_t attributeCount, AttributeArray const& attributes)
-        : HwVertexBufferInfo(bufferCount, attributeCount),
-          attributes(attributes) {
-}
-
-MetalVertexBuffer::MetalVertexBuffer(MetalContext& context,
-        uint32_t vertexCount, uint32_t bufferCount, Handle<HwVertexBufferInfo> vbih)
-    : HwVertexBuffer(vertexCount), vbih(vbih), buffers(bufferCount, nullptr) {
-}
+MetalVertexBuffer::MetalVertexBuffer(MetalContext& context, uint8_t bufferCount,
+            uint8_t attributeCount, uint32_t vertexCount, AttributeArray const& attributes)
+    : HwVertexBuffer(bufferCount, attributeCount, vertexCount, attributes), buffers(bufferCount, nullptr) {}

 MetalIndexBuffer::MetalIndexBuffer(MetalContext& context, BufferUsage usage, uint8_t elementSize,
        uint32_t indexCount) : HwIndexBuffer(elementSize, indexCount),
@@ -358,12 +333,12 @@ MetalIndexBuffer::MetalIndexBuffer(MetalContext& context, BufferUsage usage, uin
 MetalRenderPrimitive::MetalRenderPrimitive()
    : bufferMapping(utils::FixedCapacityVector<Entry>::with_capacity(MAX_VERTEX_BUFFER_COUNT)) {}

-void MetalRenderPrimitive::setBuffers(MetalVertexBufferInfo const* const vbi,
-        MetalVertexBuffer* vertexBuffer, MetalIndexBuffer* indexBuffer) {
+void MetalRenderPrimitive::setBuffers(MetalVertexBuffer* vertexBuffer, MetalIndexBuffer*
+        indexBuffer) {
    this->vertexBuffer = vertexBuffer;
    this->indexBuffer = indexBuffer;

-    const size_t attributeCount = vbi->attributes.size();
+    const size_t attributeCount = vertexBuffer->attributes.size();

    auto& mapping = bufferMapping;
    mapping.clear();
@@ -404,7 +379,7 @@ void MetalRenderPrimitive::setBuffers(MetalVertexBufferInfo const* const vbi,
    };

    for (uint32_t attributeIndex = 0; attributeIndex < attributeCount; attributeIndex++) {
-        const auto& attribute = vbi->attributes[attributeIndex];
+        const auto& attribute = vertexBuffer->attributes[attributeIndex];

        // If the attribute is unused, bind it to the zero buffer. It's a Metal error for a shader
        // to read from missing vertex attributes.
@@ -30,8 +30,6 @@

 #include <array>
 #include <memory>
-#include <tuple>
-#include <variant>

 namespace filament::backend {

@@ -43,69 +41,36 @@ class MetalShaderCompiler {
 public:
    class MetalFunctionBundle {
    public:
-        using Raster = std::tuple<id<MTLFunction>, id<MTLFunction>>;
-        using Compute = id<MTLFunction>;
-        using Error = std::tuple<NSString*, NSString*>; // error message, Program name
-        struct None {};
-
-        MetalFunctionBundle() : mPrograms{None{}} {}
-
-        explicit operator bool() const { return isValid(); }
-
-        bool isValid() const noexcept {
-            return std::holds_alternative<Raster>(mPrograms) ||
-                std::holds_alternative<Compute>(mPrograms);
-        }
-
-        /**
-         * Throws an NSException if this MetalFunctionBundle either contains an error or is empty.
-         *
-         * If this MetalFunctionBundle contains an error, will throw a MetalCompilationFailure
-         * NSException with the error string and program name passed to
-         * MetalFunctionBundle::error(NSString*, NSString*).
-         *
-         * If this MetalFunctionBundle is empty, will throw a MetalEmptyFunctionBundle NSException.
-         */
-        void validate() const;
-
-        Raster getRasterFunctions() const {
-            assert_invariant(std::holds_alternative<Raster>(mPrograms));
-            return std::get<Raster>(mPrograms);
-        }
-
-        Compute getComputeFunction() const {
-            assert_invariant(std::holds_alternative<Compute>(mPrograms));
-            return std::get<Compute>(mPrograms);
-        }
-
-        static MetalFunctionBundle none() {
-            return MetalFunctionBundle(None{});
-        }
-
-        static MetalFunctionBundle raster(id<MTLFunction> fragment, id<MTLFunction> vertex) {
+        MetalFunctionBundle() = default;
+        MetalFunctionBundle(id<MTLFunction> fragment, id<MTLFunction> vertex)
+            : functions{fragment, vertex} {
            assert_invariant(fragment && vertex);
            assert_invariant(fragment.functionType == MTLFunctionTypeFragment);
            assert_invariant(vertex.functionType == MTLFunctionTypeVertex);
-            return MetalFunctionBundle(Raster{fragment, vertex});
        }
-
-        static MetalFunctionBundle compute(id<MTLFunction> compute) {
+        explicit MetalFunctionBundle(id<MTLFunction> compute) : functions{compute, nil} {
            assert_invariant(compute);
            assert_invariant(compute.functionType == MTLFunctionTypeKernel);
-            return MetalFunctionBundle(Compute{compute});
        }

-        static MetalFunctionBundle error(NSString* errorMessage, NSString* programName) {
-            return MetalFunctionBundle(Error{errorMessage, programName});
+        std::pair<id<MTLFunction>, id<MTLFunction>> getRasterFunctions() const noexcept {
+            assert_invariant(functions[0].functionType == MTLFunctionTypeFragment);
+            assert_invariant(functions[1].functionType == MTLFunctionTypeVertex);
+            return {functions[0], functions[1]};
        }

+        id<MTLFunction> getComputeFunction() const noexcept {
+            assert_invariant(functions[0].functionType == MTLFunctionTypeKernel);
+            return functions[0];
+        }
+
+        explicit operator bool() const { return functions[0] != nil; }
+
    private:
-        MetalFunctionBundle(None&& t) : mPrograms(std::move(t)) {}
-        MetalFunctionBundle(Raster&& t) : mPrograms(std::move(t)) {}
-        MetalFunctionBundle(Compute&& t) : mPrograms(std::move(t)) {}
-        MetalFunctionBundle(Error&& t) : mPrograms(std::move(t)) {}
-
-        std::variant<Raster, Compute, None, Error> mPrograms;
+        // Can hold two functions, either:
+        // - fragment and vertex (for rasterization pipelines)
+        // - compute (for compute pipelines)
+        id<MTLFunction> functions[2] = {nil, nil};
    };

    using program_token_t = std::shared_ptr<MetalProgramToken>;
@@ -118,16 +118,13 @@ void MetalShaderCompiler::terminate() noexcept {
                                                      options:options
                                                        error:&error];
        if (library == nil) {
-            NSString* errorMessage = @"unknown error";
            if (error) {
                auto description =
                        [error.localizedDescription cStringUsingEncoding:NSUTF8StringEncoding];
                utils::slog.w << description << utils::io::endl;
-                errorMessage = error.localizedDescription;
            }
            PANIC_LOG("Failed to compile Metal program.");
-            NSString* programName = [NSString stringWithFormat:@"%s", program.getName().c_str_safe()];
-            return MetalFunctionBundle::error(errorMessage, programName);
+            return {};
        }

        MTLFunctionConstantValues* constants = [MTLFunctionConstantValues new];
@@ -163,15 +160,14 @@ void MetalShaderCompiler::terminate() noexcept {
    assert_invariant(isRasterizationProgram != isComputeProgram);

    if (isRasterizationProgram) {
-        return MetalFunctionBundle::raster(fragmentFunction, vertexFunction);
+        return {fragmentFunction, vertexFunction};
    }

    if (isComputeProgram) {
-        return MetalFunctionBundle::compute(computeFunction);
+        return MetalFunctionBundle{computeFunction};
    }

-    // Should never reach here.
-    return MetalFunctionBundle::none();
+    return {};
 }

 MetalShaderCompiler::program_token_t MetalShaderCompiler::createProgram(
@@ -229,24 +225,4 @@ void MetalShaderCompiler::notifyWhenAllProgramsAreReady(
    mCallbackManager.setCallback(handler, callback, user);
 }

-UTILS_NOINLINE
-void MetalShaderCompiler::MetalFunctionBundle::validate() const {
-    if (UTILS_UNLIKELY(std::holds_alternative<Error>(mPrograms))) {
-        auto [errorMessage, programName] = std::get<Error>(mPrograms);
-        NSString* reason =
-                [NSString stringWithFormat:
-                        @"Attempting to draw with an id<MTLFunction> that failed to compile.\n"
-                        @"Program: %@\n"
-                        @"%@", programName, errorMessage];
-        [[NSException exceptionWithName:@"MetalCompilationFailure"
-                                reason:reason
-                              userInfo:nil] raise];
-    } else if (UTILS_UNLIKELY(std::holds_alternative<None>(mPrograms))) {
-        NSString* reason = @"Attempting to draw with an empty id<MTLFunction>.";
-        [[NSException exceptionWithName:@"MetalEmptyFunctionBundle"
-                                reason:reason
-                              userInfo:nil] raise];
-    }
-}
-
 } // namespace filament::backend
@@ -18,8 +18,6 @@

 #include "MetalEnums.h"

-#include <utils/Log.h>
-
 namespace filament {
 namespace backend {

@@ -77,9 +77,6 @@ void NoopDriver::finish(int) {
 void NoopDriver::destroyRenderPrimitive(Handle<HwRenderPrimitive> rph) {
 }

-void NoopDriver::destroyVertexBufferInfo(Handle<HwVertexBufferInfo> vbih) {
-}
-
 void NoopDriver::destroyVertexBuffer(Handle<HwVertexBuffer> vbh) {
 }

@@ -177,7 +174,7 @@ bool NoopDriver::isSRGBSwapChainSupported() {
    return false;
 }

-bool NoopDriver::isStereoSupported(backend::StereoscopicType) {
+bool NoopDriver::isStereoSupported() {
    return false;
 }

@@ -343,16 +340,12 @@ void NoopDriver::blit(
 }

 void NoopDriver::draw(PipelineState pipelineState, Handle<HwRenderPrimitive> rph,
-        uint32_t indexOffset, uint32_t indexCount, uint32_t instanceCount) {
+        uint32_t instanceCount) {
 }

 void NoopDriver::dispatchCompute(Handle<HwProgram> program, math::uint3 workGroupCount) {
 }

-void NoopDriver::scissor(
-        Viewport scissor) {
-}
-
 void NoopDriver::beginTimerQuery(Handle<HwTimerQuery> tqh) {
 }

@@ -427,6 +427,7 @@ void OpenGLContext::initBugs(Bugs* bugs, Extensions const& exts,
            if (strstr(renderer, "Mali-T")) {
                bugs->disable_glFlush = true;
                bugs->disable_shared_context_draws = true;
+                bugs->texture_external_needs_rebind = true;
                // We have not verified that timer queries work on Mali-T, so we disable to be safe.
                bugs->dont_use_timer_query = true;
            }
@@ -622,7 +623,6 @@ void OpenGLContext::initExtensionsGLES(Extensions* ext, GLint major, GLint minor
    ext->OES_standard_derivatives = exts.has("GL_OES_standard_derivatives"sv);
    ext->OES_texture_npot = exts.has("GL_OES_texture_npot"sv);
    ext->OES_vertex_array_object = exts.has("GL_OES_vertex_array_object"sv);
-    ext->OVR_multiview2 = exts.has("GL_OVR_multiview2"sv);
    ext->WEBGL_compressed_texture_etc = exts.has("WEBGL_compressed_texture_etc"sv);
    ext->WEBGL_compressed_texture_s3tc = exts.has("WEBGL_compressed_texture_s3tc"sv);
    ext->WEBGL_compressed_texture_s3tc_srgb = exts.has("WEBGL_compressed_texture_s3tc_srgb"sv);
@@ -687,7 +687,6 @@ void OpenGLContext::initExtensionsGL(Extensions* ext, GLint major, GLint minor)
    ext->OES_standard_derivatives = true;
    ext->OES_texture_npot = true;
    ext->OES_vertex_array_object = true;
-    ext->OVR_multiview2 = exts.has("GL_OVR_multiview2"sv);
    ext->WEBGL_compressed_texture_etc = false;
    ext->WEBGL_compressed_texture_s3tc = false;
    ext->WEBGL_compressed_texture_s3tc_srgb = false;
@@ -205,7 +205,6 @@ public:
        bool OES_standard_derivatives;
        bool OES_texture_npot;
        bool OES_vertex_array_object;
-        bool OVR_multiview2;
        bool WEBGL_compressed_texture_etc;
        bool WEBGL_compressed_texture_s3tc;
        bool WEBGL_compressed_texture_s3tc_srgb;
@@ -223,6 +222,10 @@ public:
        // Some drivers have gl state issues when drawing from shared contexts
        bool disable_shared_context_draws;

+        // Some drivers require the GL_TEXTURE_EXTERNAL_OES target to be bound when
+        // the texture image changes, even if it's already bound to that texture
+        bool texture_external_needs_rebind;
+
        // Some web browsers seem to immediately clear the default framebuffer when calling
        // glInvalidateFramebuffer with WebGL 2.0
        bool disable_invalidate_framebuffer;
@@ -430,6 +433,9 @@ private:
            {   bugs.disable_shared_context_draws,
                    "disable_shared_context_draws",
                    ""},
+            {   bugs.texture_external_needs_rebind,
+                    "texture_external_needs_rebind",
+                    ""},
            {   bugs.disable_invalidate_framebuffer,
                    "disable_invalidate_framebuffer",
                    ""},
@@ -705,7 +711,7 @@ void OpenGLContext::bindTexture(GLuint unit, GLuint target, GLuint texId, size_t
    update_state(state.textures.units[unit].targets[targetIndex].texture_id, texId, [&]() {
        activeTexture(unit);
        glBindTexture(target, texId);
-    }, target == GL_TEXTURE_EXTERNAL_OES);
+    }, (target == GL_TEXTURE_EXTERNAL_OES) && bugs.texture_external_needs_rebind);
 }

 void OpenGLContext::bindTexture(GLuint unit, GLuint target, GLuint texId) noexcept {
@@ -90,28 +90,26 @@ Driver* OpenGLDriver::create(OpenGLPlatform* const platform,
 #if 0
    // this is useful for development, but too verbose even for debug builds
    // For reference on a 64-bits machine in Release mode:
+    //    GLFence                   :   8       few
    //    GLIndexBuffer             :   8       moderate
-    //    GLSamplerGroup            :  16       few
-    //    GLSwapChain               :  16       few
-    //    GLTimerQuery              :  16       few
-    //    GLRenderPrimitive         :  20       many
-    //    GLFence                   :  24       few
-    // -- less than or equal 24 bytes
-    //    GLBufferObject            :  32       many
-    //    OpenGLProgram             :  56       moderate
-    //    GLTexture                 :  64       moderate
+    //    GLSamplerGroup            :   8       few
+    // -- less than or equal 16 bytes
+    //    GLBufferObject            :  24       many
+    //    GLSync                    :  24       few
+    //    GLTimerQuery              :  32       few
+    //    OpenGLProgram             :  32       moderate
+    //    GLRenderPrimitive         :  48       many
    // -- less than or equal 64 bytes
-    //    GLVertexBuffer            :  76       moderate
-    //    GLStream                  : 104       few
+    //    GLTexture                 :  72       moderate
    //    GLRenderTarget            : 112       few
-    //    GLVertexBufferInfo        : 132       moderate
-    // -- less than or equal to 136 bytes
+    //    GLStream                  : 184       few
+    //    GLVertexBuffer            : 200       moderate
+    // -- less than or equal to 208 bytes

    slog.d
-           << "\nGLSwapChain: " << sizeof(GLSwapChain)
+           << "HwFence: " << sizeof(HwFence)
           << "\nGLBufferObject: " << sizeof(GLBufferObject)
           << "\nGLVertexBuffer: " << sizeof(GLVertexBuffer)
-           << "\nGLVertexBufferInfo: " << sizeof(GLVertexBufferInfo)
           << "\nGLIndexBuffer: " << sizeof(GLIndexBuffer)
           << "\nGLSamplerGroup: " << sizeof(GLSamplerGroup)
           << "\nGLRenderPrimitive: " << sizeof(GLRenderPrimitive)
@@ -119,7 +117,7 @@ Driver* OpenGLDriver::create(OpenGLPlatform* const platform,
           << "\nGLTimerQuery: " << sizeof(GLTimerQuery)
           << "\nGLStream: " << sizeof(GLStream)
           << "\nGLRenderTarget: " << sizeof(GLRenderTarget)
-           << "\nGLFence: " << sizeof(GLFence)
+           << "\nGLSync: " << sizeof(GLSync)
           << "\nOpenGLProgram: " << sizeof(OpenGLProgram)
           << io::endl;
 #endif
@@ -149,9 +147,9 @@ Driver* OpenGLDriver::create(OpenGLPlatform* const platform,
 #endif

    size_t const defaultSize = FILAMENT_OPENGL_HANDLE_ARENA_SIZE_IN_MB * 1024U * 1024U;
-    Platform::DriverConfig validConfig{ driverConfig };
+    Platform::DriverConfig validConfig {driverConfig};
    validConfig.handleArenaSize = std::max(driverConfig.handleArenaSize, defaultSize);
-    OpenGLDriver* const driver = new(std::nothrow) OpenGLDriver(ec, validConfig);
+    OpenGLDriver* const driver = new OpenGLDriver(ec, validConfig);
    return driver;
 }

@@ -173,8 +171,7 @@ OpenGLDriver::OpenGLDriver(OpenGLPlatform* platform, const Platform::DriverConfi
          mContext(),
          mShaderCompilerService(*this),
          mHandleAllocator("Handles", driverConfig.handleArenaSize),
-          mSamplerMap(32),
-          mDriverConfig(driverConfig) {
+          mSamplerMap(32) {

    std::fill(mSamplerBindings.begin(), mSamplerBindings.end(), nullptr);

@@ -388,10 +385,6 @@ void OpenGLDriver::setStencilState(StencilState ss) noexcept {
 // Creating driver objects
 // ------------------------------------------------------------------------------------------------

-Handle<HwVertexBufferInfo> OpenGLDriver::createVertexBufferInfoS() noexcept {
-    return initHandle<GLVertexBufferInfo>();
-}
-
 Handle<HwVertexBuffer> OpenGLDriver::createVertexBufferS() noexcept {
    return initHandle<GLVertexBuffer>();
 }
@@ -452,21 +445,14 @@ Handle<HwTimerQuery> OpenGLDriver::createTimerQueryS() noexcept {
    return initHandle<GLTimerQuery>();
 }

-void OpenGLDriver::createVertexBufferInfoR(
-        Handle<HwVertexBufferInfo> vbih,
-        uint8_t bufferCount,
-        uint8_t attributeCount,
-        AttributeArray attributes) {
-    DEBUG_MARKER()
-    construct<GLVertexBufferInfo>(vbih, bufferCount, attributeCount, attributes);
-}
-
 void OpenGLDriver::createVertexBufferR(
        Handle<HwVertexBuffer> vbh,
-        uint32_t vertexCount,
-        Handle<HwVertexBufferInfo> vbih) {
+        uint8_t bufferCount,
+        uint8_t attributeCount,
+        uint32_t elementCount,
+        AttributeArray attributes) {
    DEBUG_MARKER()
-    construct<GLVertexBuffer>(vbh, vertexCount, vbih);
+    construct<GLVertexBuffer>(vbh, bufferCount, attributeCount, elementCount, attributes);
 }

 void OpenGLDriver::createIndexBufferR(
@@ -514,11 +500,13 @@ void OpenGLDriver::createBufferObjectR(Handle<HwBufferObject> boh,

 void OpenGLDriver::createRenderPrimitiveR(Handle<HwRenderPrimitive> rph,
        Handle<HwVertexBuffer> vbh, Handle<HwIndexBuffer> ibh,
-        PrimitiveType pt) {
+        PrimitiveType pt, uint32_t offset,
+        uint32_t minIndex, uint32_t maxIndex, uint32_t count) {
    DEBUG_MARKER()

    auto& gl = mContext;

+    GLVertexBuffer const* const eb = handle_cast<const GLVertexBuffer*>(vbh);
    GLIndexBuffer const* const ib = handle_cast<const GLIndexBuffer*>(ibh);
    assert_invariant(ib->elementSize == 2 || ib->elementSize == 4);

@@ -526,14 +514,17 @@ void OpenGLDriver::createRenderPrimitiveR(Handle<HwRenderPrimitive> rph,
    rp->gl.indicesSize = (ib->elementSize == 4u) ? 4u : 2u;
    rp->gl.vertexBufferWithObjects = vbh;
    rp->type = pt;
+    rp->offset = offset * rp->gl.indicesSize;
+    rp->count = count;
+    rp->minIndex = minIndex;
+    rp->maxIndex = maxIndex > minIndex ? maxIndex : rp->maxVertexCount - 1; // sanitize max index

    gl.procs.genVertexArrays(1, &rp->gl.vao);

    gl.bindVertexArray(&rp->gl);

-    // Note: we don't update the vertex buffer bindings in the VAO just yet, we will do that
-    // later in draw() or bindRenderPrimitive(). At this point, the HwVertexBuffer might not
-    // have all its buffers set.
+    // update the VBO bindings in the VAO
+    updateVertexArrayObject(rp, eb);

    // this records the index buffer into the currently bound VAO
    gl.bindBuffer(GL_ELEMENT_ARRAY_BUFFER, ib->gl.buffer);
@@ -856,6 +847,7 @@ void OpenGLDriver::updateVertexArrayObject(GLRenderPrimitive* rp, GLVertexBuffer

    // NOTE: this is called from draw() and must be as efficient as possible.

+
    if (UTILS_LIKELY(gl.ext.OES_vertex_array_object)) {
        // The VAO for the given render primitive must already be bound.
 #ifndef NDEBUG
@@ -869,49 +861,42 @@ void OpenGLDriver::updateVertexArrayObject(GLRenderPrimitive* rp, GLVertexBuffer
        // that it's always reset in draw
    }

-    GLVertexBufferInfo const* const vbi = handle_cast<const GLVertexBufferInfo*>(vb->vbih);
-
-    for (size_t i = 0, n = vbi->attributes.size(); i < n; i++) {
-        const auto& attribute = vbi->attributes[i];
+    rp->maxVertexCount = vb->vertexCount;
+    for (size_t i = 0, n = vb->attributes.size(); i < n; i++) {
+        const auto& attribute = vb->attributes[i];
        const uint8_t bi = attribute.buffer;
-        if (bi != Attribute::BUFFER_UNUSED) {
-            // if a buffer is defined it must not be invalid.
-            assert_invariant(vb->gl.buffers[bi]);

-            // if w're on ES2, the user shouldn't use FLAG_INTEGER_TARGET
+        // Invoking glVertexAttribPointer without a bound VBO is an invalid operation, so we must
+        // take care to avoid it. This can occur when VertexBuffer is only partially populated with
+        // BufferObject items.
+        if (bi != Attribute::BUFFER_UNUSED && UTILS_LIKELY(vb->gl.buffers[bi] != 0)) {
+
            assert_invariant(!(gl.isES2() && (attribute.flags & Attribute::FLAG_INTEGER_TARGET)));

            gl.bindBuffer(GL_ARRAY_BUFFER, vb->gl.buffers[bi]);
-            GLuint const index = i;
-            GLint const size = getComponentCount(attribute.type);
-            GLenum const type = getComponentType(attribute.type);
-            GLboolean const normalized = getNormalization(attribute.flags & Attribute::FLAG_NORMALIZED);
-            GLsizei const stride = attribute.stride;
-            void const* pointer = reinterpret_cast<void const *>(attribute.offset);
-
 #ifndef FILAMENT_SILENCE_NOT_SUPPORTED_BY_ES2
            if (UTILS_UNLIKELY(attribute.flags & Attribute::FLAG_INTEGER_TARGET)) {
-                // integer attributes can't be floats
-                assert_invariant(type == GL_BYTE || type == GL_UNSIGNED_BYTE || type == GL_SHORT ||
-                        type == GL_UNSIGNED_SHORT || type == GL_INT || type == GL_UNSIGNED_INT);
-                glVertexAttribIPointer(index, size, type, stride, pointer);
+                glVertexAttribIPointer(GLuint(i),
+                        (GLint)getComponentCount(attribute.type),
+                        getComponentType(attribute.type),
+                        attribute.stride,
+                        (void*) uintptr_t(attribute.offset));
            } else
 #endif
            {
-                glVertexAttribPointer(index, size, type, normalized, stride, pointer);
+                glVertexAttribPointer(GLuint(i),
+                        (GLint)getComponentCount(attribute.type),
+                        getComponentType(attribute.type),
+                        getNormalization(attribute.flags & Attribute::FLAG_NORMALIZED),
+                        attribute.stride,
+                        (void*) uintptr_t(attribute.offset));
            }

            gl.enableVertexAttribArray(GLuint(i));
        } else {
-            // In some OpenGL implementations, we must supply a properly-typed placeholder for
-            // every integer input that is declared in the vertex shader.
-            // Note that the corresponding doesn't have to be enabled and in fact won't be. If it
-            // was enabled, it would indicate a user-error (providing the wrong type of array).
-            // With a disabled array, the vertex shader gets the attribute from glVertexAttrib,
-            // and must have the proper intergerness.
-            // But at this point, we don't know what the shader requirements are, and so we must
-            // rely on the attribute.

+            // In some OpenGL implementations, we must supply a properly-typed placeholder for
+            // every integer input that is declared in the vertex shader, even if disabled.
 #ifndef FILAMENT_SILENCE_NOT_SUPPORTED_BY_ES2
            if (UTILS_UNLIKELY(attribute.flags & Attribute::FLAG_INTEGER_TARGET)) {
                if (!gl.isES2()) {
@@ -1424,14 +1409,6 @@ void OpenGLDriver::createTimerQueryR(Handle<HwTimerQuery> tqh, int) {
 // Destroying driver objects
 // ------------------------------------------------------------------------------------------------

-void OpenGLDriver::destroyVertexBufferInfo(Handle<HwVertexBufferInfo> vbih) {
-    DEBUG_MARKER()
-    if (vbih) {
-        GLVertexBufferInfo const* vbi = handle_cast<const GLVertexBufferInfo*>(vbih);
-        destruct(vbih, vbi);
-    }
-}
-
 void OpenGLDriver::destroyVertexBuffer(Handle<HwVertexBuffer> vbh) {
    DEBUG_MARKER()
    if (vbh) {
@@ -1921,20 +1898,12 @@ bool OpenGLDriver::isSRGBSwapChainSupported() {
    return mPlatform.isSRGBSwapChainSupported();
 }

-bool OpenGLDriver::isStereoSupported(backend::StereoscopicType stereoscopicType) {
-    // Instanced-stereo requires instancing and EXT_clip_cull_distance.
-    // Multiview-stereo requires ES 3.0 and OVR_multiview2.
+bool OpenGLDriver::isStereoSupported() {
+    // Stereo requires instancing and EXT_clip_cull_distance.
    if (UTILS_UNLIKELY(mContext.isES2())) {
        return false;
    }
-    switch (stereoscopicType) {
-    case backend::StereoscopicType::INSTANCED:
-        return mContext.ext.EXT_clip_cull_distance;
-    case backend::StereoscopicType::MULTIVIEW:
-        return mContext.ext.OVR_multiview2;
-    default:
-        return false;
-    }
+    return mContext.ext.EXT_clip_cull_distance;
 }

 bool OpenGLDriver::isParallelShaderCompileSupported() {
@@ -3728,8 +3697,7 @@ void OpenGLDriver::updateTextureLodRange(GLTexture* texture, int8_t targetLevel)
 #endif
 }

-void OpenGLDriver::draw(PipelineState state, Handle<HwRenderPrimitive> rph,
-        uint32_t const indexOffset, uint32_t const indexCount, uint32_t const instanceCount) {
+void OpenGLDriver::draw(PipelineState state, Handle<HwRenderPrimitive> rph, uint32_t instanceCount) {
    DEBUG_MARKER()
    auto& gl = mContext;

@@ -3764,15 +3732,16 @@ void OpenGLDriver::draw(PipelineState state, Handle<HwRenderPrimitive> rph,

    gl.polygonOffset(state.polygonOffset.slope, state.polygonOffset.constant);

+    setScissor(state.scissor);
+
    if (UTILS_LIKELY(instanceCount <= 1)) {
-        glDrawElements(GLenum(rp->type), (GLsizei)indexCount, rp->gl.getIndicesType(),
-                reinterpret_cast<const void*>(indexOffset * rp->gl.indicesSize));
+        glDrawElements(GLenum(rp->type), (GLsizei)rp->count, rp->gl.getIndicesType(),
+                reinterpret_cast<const void*>(rp->offset));
    } else {
        assert_invariant(!mContext.isES2());
 #ifndef FILAMENT_SILENCE_NOT_SUPPORTED_BY_ES2
-        glDrawElementsInstanced(GLenum(rp->type), (GLsizei)indexCount,
-                rp->gl.getIndicesType(),
-                reinterpret_cast<const void*>(indexOffset * rp->gl.indicesSize),
+        glDrawElementsInstanced(GLenum(rp->type), (GLsizei)rp->count,
+                rp->gl.getIndicesType(), reinterpret_cast<const void*>(rp->offset),
                (GLsizei)instanceCount);
 #endif
    }
@@ -3784,10 +3753,6 @@ void OpenGLDriver::draw(PipelineState state, Handle<HwRenderPrimitive> rph,
 #endif
 }

-void OpenGLDriver::scissor(Viewport scissor) {
-    setScissor(scissor);
-}
-
 void OpenGLDriver::dispatchCompute(Handle<HwProgram> program, math::uint3 workGroupCount) {
    getShaderCompilerService().tick();

@@ -94,22 +94,8 @@ public:
        uint16_t age = 0;
    };

-    struct GLVertexBufferInfo : public HwVertexBufferInfo {
-        GLVertexBufferInfo() noexcept = default;
-        GLVertexBufferInfo(uint8_t bufferCount, uint8_t attributeCount,
-                AttributeArray const& attributes)
-                : HwVertexBufferInfo(bufferCount, attributeCount),
-                  attributes(attributes) {
-        }
-        AttributeArray attributes;
-    };
-
    struct GLVertexBuffer : public HwVertexBuffer {
-        GLVertexBuffer() noexcept = default;
-        GLVertexBuffer(uint32_t vertexCount, Handle<HwVertexBufferInfo> vbih)
-                : HwVertexBuffer(vertexCount), vbih(vbih) {
-        }
-        Handle<HwVertexBufferInfo> vbih;
+        using HwVertexBuffer::HwVertexBuffer;
        struct {
            // 4 * MAX_VERTEX_ATTRIBUTE_COUNT bytes
            std::array<GLuint, MAX_VERTEX_ATTRIBUTE_COUNT> buffers{};
@@ -414,9 +400,6 @@ private:
    // timer query implementation
    OpenGLTimerQueryInterface* mTimerQueryImpl = nullptr;

-    const Platform::DriverConfig mDriverConfig;
-    Platform::DriverConfig const& getDriverConfig() const noexcept { return mDriverConfig; }
-
    // for ES2 sRGB support
    GLSwapChain* mCurrentDrawSwapChain = nullptr;
    bool mRec709OutputColorspace = false;
@@ -18,14 +18,6 @@

 #include "OpenGLDriverFactory.h"

-#include <backend/AcquiredImage.h>
-#include <backend/DriverEnums.h>
-#include <backend/Platform.h>
-
-#include <utils/compiler.h>
-
-#include <stdint.h>
-
 namespace filament::backend {

 Driver* OpenGLPlatform::createDefaultDriver(OpenGLPlatform* platform,
@@ -151,12 +151,6 @@ bool ShaderCompilerService::isParallelShaderCompileSupported() const noexcept {
 }

 void ShaderCompilerService::init() noexcept {
-    if (UTILS_UNLIKELY(mDriver.getDriverConfig().disableParallelShaderCompile)) {
-        // user disabled parallel shader compile
-        mMode = Mode::SYNCHRONOUS;
-        return;
-    }
-
    // Here we decide which mode we'll be using. We always prefer our own thread-pool if
    // that mode is available because, we have no control on how the compilation queues are
    // handled if done by the driver (so at the very least we'd need to decode this per-driver).
@@ -261,11 +255,12 @@ ShaderCompilerService::program_token_t ShaderCompilerService::createProgram(
                [this, &gl, program = std::move(program), token]() mutable {
                    // compile the shaders
                    std::array<GLuint, Program::SHADER_TYPE_COUNT> shaders{};
+                    std::array<utils::CString, Program::SHADER_TYPE_COUNT> shaderSourceCode;
                    compileShaders(gl,
                            std::move(program.getShadersSource()),
                            program.getSpecializationConstants(),
                            shaders,
-                            token->shaderSourceCode);
+                            shaderSourceCode);

                    // link the program
                    GLuint const glProgram = linkProgram(gl, shaders, token->attributes);
@@ -109,7 +109,7 @@ private:
    };

    enum class Mode {
-        UNDEFINED,      // init() has not been called yet.
+        UNDEFINED,      // init() has not beed called yet.
        SYNCHRONOUS,    // synchronous shader compilation
        THREAD_POOL,    // asynchronous shader compilation using a thread-pool (most common)
        ASYNCHRONOUS    // asynchronous shader compilation using KHR_parallel_shader_compile
@@ -18,31 +18,16 @@

 #include "opengl/GLUtils.h"

-#include <backend/platforms/OpenGLPlatform.h>
-
-#include <backend/Platform.h>
-#include <backend/DriverEnums.h>
-
 #include <EGL/egl.h>
 #include <EGL/eglext.h>
-#if defined(__ANDROID__)

+#if defined(__ANDROID__)
 #include <sys/system_properties.h>
 #endif
+
 #include <utils/compiler.h>
-
-#include <utils/debug.h>
 #include <utils/Log.h>
-#include <utils/ostream.h>
-
-#include <algorithm>
-#include <new>
-#include <initializer_list>
-#include <utility>
-
-#include <stddef.h>
-#include <stdint.h>
-#include <stdlib.h>
+#include <utils/Panic.h>

 #ifndef EGL_CONTEXT_OPENGL_BACKWARDS_COMPATIBLE_ANGLE
 #   define EGL_CONTEXT_OPENGL_BACKWARDS_COMPATIBLE_ANGLE 0x3483
@@ -63,6 +48,7 @@ UTILS_PRIVATE PFNEGLDESTROYIMAGEKHRPROC eglDestroyImageKHR = {};
 }
 using namespace glext;

+
 // ---------------------------------------------------------------------------------------------
 // Utilities
 // ---------------------------------------------------------------------------------------------
@@ -151,7 +137,7 @@ Driver* PlatformEGL::createDriver(void* sharedContext, const Platform::DriverCon
    ext.egl.KHR_gl_colorspace = extensions.has("EGL_KHR_gl_colorspace");
    ext.egl.KHR_create_context = extensions.has("EGL_KHR_create_context");
    ext.egl.KHR_no_config_context = extensions.has("EGL_KHR_no_config_context");
-    ext.egl.KHR_surfaceless_context = extensions.has("EGL_KHR_surfaceless_context");
+    ext.egl.KHR_surfaceless_context = extensions.has("KHR_surfaceless_context");
    if (ext.egl.KHR_create_context) {
        // KHR_create_context implies KHR_surfaceless_context for ES3.x contexts
        ext.egl.KHR_surfaceless_context = true;
@@ -475,13 +461,7 @@ Platform::SwapChain* PlatformEGL::createSwapChain(
    // this is not fatal
    eglSurfaceAttrib(mEGLDisplay, sur, EGL_SWAP_BEHAVIOR, EGL_BUFFER_DESTROYED);

-    SwapChainEGL* const sc = new(std::nothrow) SwapChainEGL({
-        .sur = sur,
-        .attribs = std::move(attribs),
-        .nativeWindow = (EGLNativeWindowType)nativeWindow,
-        .config = config
-    });
-    return sc;
+    return (SwapChain*)sur;
 }

 Platform::SwapChain* PlatformEGL::createSwapChain(
@@ -515,42 +495,30 @@ Platform::SwapChain* PlatformEGL::createSwapChain(
        logEglError("eglCreatePbufferSurface");
        return nullptr;
    }
-
-    SwapChainEGL* const sc = new(std::nothrow) SwapChainEGL({
-            .sur = sur,
-            .attribs = std::move(attribs),
-            .config = config
-    });
-    return sc;
+    return (SwapChain*)sur;
 }

 void PlatformEGL::destroySwapChain(Platform::SwapChain* swapChain) noexcept {
-    if (swapChain) {
-        SwapChainEGL const* const sc = static_cast<SwapChainEGL const*>(swapChain);
-        if (sc->sur != EGL_NO_SURFACE) {
-            makeCurrent(mEGLDummySurface, mEGLDummySurface);
-            eglDestroySurface(mEGLDisplay, sc->sur);
-            delete sc;
-        }
+    EGLSurface sur = (EGLSurface) swapChain;
+    if (sur != EGL_NO_SURFACE) {
+        makeCurrent(mEGLDummySurface, mEGLDummySurface);
+        eglDestroySurface(mEGLDisplay, sur);
    }
 }

 void PlatformEGL::makeCurrent(Platform::SwapChain* drawSwapChain,
                              Platform::SwapChain* readSwapChain) noexcept {
-    SwapChainEGL const* const dsc = static_cast<SwapChainEGL const*>(drawSwapChain);
-    SwapChainEGL const* const rsc = static_cast<SwapChainEGL const*>(readSwapChain);
-    if (UTILS_UNLIKELY(dsc->sur == EGL_NO_SURFACE && rsc->sur == EGL_NO_SURFACE)) {
-        return;
+    EGLSurface drawSur = (EGLSurface) drawSwapChain;
+    EGLSurface readSur = (EGLSurface) readSwapChain;
+    if (drawSur != EGL_NO_SURFACE || readSur != EGL_NO_SURFACE) {
+        makeCurrent(drawSur, readSur);
    }
-    makeCurrent(dsc->sur, rsc->sur);
 }

 void PlatformEGL::commit(Platform::SwapChain* swapChain) noexcept {
-    if (swapChain) {
-        SwapChainEGL const* const sc = static_cast<SwapChainEGL const*>(swapChain);
-        if (sc->sur != EGL_NO_SURFACE) {
-            eglSwapBuffers(mEGLDisplay, sc->sur);
-        }
+    EGLSurface sur = (EGLSurface) swapChain;
+    if (sur != EGL_NO_SURFACE) {
+        eglSwapBuffers(mEGLDisplay, sur);
    }
 }

@@ -593,7 +561,7 @@ FenceStatus PlatformEGL::waitFence(
 }

 OpenGLPlatform::ExternalTexture* PlatformEGL::createExternalImageTexture() noexcept {
-    ExternalTexture* outTexture = new(std::nothrow) ExternalTexture{};
+    ExternalTexture* outTexture = new ExternalTexture{};
    glGenTextures(1, &outTexture->id);
    if (UTILS_LIKELY(ext.gl.OES_EGL_image_external_essl3)) {
        outTexture->target = GL_TEXTURE_EXTERNAL_OES;
@@ -39,12 +39,7 @@
 // not required for validation.

 // FVK is short for Filament Vulkan
-
-// Enables Android systrace
 #define FVK_DEBUG_SYSTRACE                0x00000001
-
-// Group markers are used to denote collection of GPU commands.  It is typically at the granualarity
-// of a renderpass.
 #define FVK_DEBUG_GROUP_MARKERS           0x00000002
 #define FVK_DEBUG_TEXTURE                 0x00000004
 #define FVK_DEBUG_LAYOUT_TRANSITION       0x00000008
@@ -60,9 +55,6 @@
 #define FVK_DEBUG_PIPELINE_CACHE          0x00002000
 #define FVK_DEBUG_ALLOCATION              0x00004000

-// Enable the debug utils extension if it is available.
-#define FVK_DEBUG_DEBUG_UTILS             0x00008000
-
 // Usefaul default combinations
 #define FVK_DEBUG_EVERYTHING              0xFFFFFFFF
 #define FVK_DEBUG_PERFORMANCE     \
@@ -80,7 +72,7 @@
    FVK_DEBUG_PRINT_GROUP_MARKERS

 #ifndef NDEBUG
-#define FVK_DEBUG_FLAGS (FVK_DEBUG_PERFORMANCE | FVK_DEBUG_DEBUG_UTILS | FVK_DEBUG_VALIDATION)
+#define FVK_DEBUG_FLAGS FVK_DEBUG_PERFORMANCE
 #else
 #define FVK_DEBUG_FLAGS 0
 #endif
@@ -88,23 +80,11 @@
 #define FVK_ENABLED(flags) ((FVK_DEBUG_FLAGS) & (flags))
 #define FVK_ENABLED_BOOL(flags) ((bool) FVK_ENABLED(flags))

-
-// Group marker only works only if validation or debug utils is enabled since it uses
-// vkCmd(Begin/End)DebugUtilsLabelEXT or vkCmdDebugMarker(Begin/End)EXT
-#if FVK_ENABLED(FVK_DEBUG_PRINT_GROUP_MARKERS)
-static_assert(FVK_ENABLED(FVK_DEBUG_DEBUG_UTILS) || FVK_ENABLED(FVK_DEBUG_VALIDATION));
-#endif
-
 // Ensure dependencies are met between debug options
 #if FVK_ENABLED(FVK_DEBUG_PRINT_GROUP_MARKERS)
 static_assert(FVK_ENABLED(FVK_DEBUG_GROUP_MARKERS));
 #endif

-// Only enable debug utils if validation is enabled.
-#if FVK_ENABLED(FVK_DEBUG_DEBUG_UTILS)
-static_assert(FVK_ENABLED(FVK_DEBUG_VALIDATION));
-#endif
-
 // end dependcy checks

 #if FVK_ENABLED(FVK_DEBUG_SYSTRACE)
@@ -24,7 +24,6 @@
 #include "VulkanHandles.h"
 #include "VulkanImageUtility.h"
 #include "VulkanMemory.h"
-#include "VulkanTexture.h"

 #include <backend/platforms/VulkanPlatform.h>

@@ -118,9 +117,7 @@ VKAPI_ATTR VkBool32 VKAPI_CALL debugReportCallback(VkDebugReportFlagsEXT flags,
    utils::slog.e << utils::io::endl;
    return VK_FALSE;
 }
-#endif // FVK_EANBLED(FVK_DEBUG_VALIDATION)

-#if FVK_ENABLED(FVK_DEBUG_DEBUG_UTILS)
 VKAPI_ATTR VkBool32 VKAPI_CALL debugUtilsCallback(VkDebugUtilsMessageSeverityFlagBitsEXT severity,
        VkDebugUtilsMessageTypeFlagsEXT types, const VkDebugUtilsMessengerCallbackDataEXT* cbdata,
        void* pUserData) {
@@ -139,63 +136,9 @@ VKAPI_ATTR VkBool32 VKAPI_CALL debugUtilsCallback(VkDebugUtilsMessageSeverityFla
    utils::slog.e << utils::io::endl;
    return VK_FALSE;
 }
-#endif // FVK_EANBLED(FVK_DEBUG_DEBUG_UTILS)
-
-}// anonymous namespace
-
-#if FVK_ENABLED(FVK_DEBUG_DEBUG_UTILS)
-using DebugUtils = VulkanDriver::DebugUtils;
-DebugUtils* DebugUtils::mSingleton = nullptr;
-
-DebugUtils::DebugUtils(VkInstance instance, VkDevice device, VulkanContext const* context)
-    : mInstance(instance), mDevice(device), mEnabled(context->isDebugUtilsSupported()) {
-
-#if FVK_ENABLED(FVK_DEBUG_VALIDATION)
-    // Also initialize debug utils messenger here
-    if (mEnabled) {
-        VkDebugUtilsMessengerCreateInfoEXT const createInfo = {
-                .sType = VK_STRUCTURE_TYPE_DEBUG_UTILS_MESSENGER_CREATE_INFO_EXT,
-                .pNext = nullptr,
-                .flags = 0,
-                .messageSeverity = VK_DEBUG_UTILS_MESSAGE_SEVERITY_WARNING_BIT_EXT
-                                   | VK_DEBUG_UTILS_MESSAGE_SEVERITY_ERROR_BIT_EXT,
-                .messageType = VK_DEBUG_UTILS_MESSAGE_TYPE_GENERAL_BIT_EXT
-                               | VK_DEBUG_UTILS_MESSAGE_TYPE_VALIDATION_BIT_EXT,
-                .pfnUserCallback = debugUtilsCallback,
-        };
-        VkResult result = vkCreateDebugUtilsMessengerEXT(instance, &createInfo,
-                VKALLOC, &mDebugMessenger);
-        ASSERT_POSTCONDITION(result == VK_SUCCESS, "Unable to create Vulkan debug messenger.");
-    }
 #endif // FVK_EANBLED(FVK_DEBUG_VALIDATION)
-}

-DebugUtils* DebugUtils::get() {
-    assert_invariant(DebugUtils::mSingleton);
-    return DebugUtils::mSingleton;
-}
-
-DebugUtils::~DebugUtils() {
-    if (mDebugMessenger) {
-        vkDestroyDebugUtilsMessengerEXT(mInstance, mDebugMessenger, VKALLOC);
-    }
-}
-
-void DebugUtils::setName(VkObjectType type, uint64_t handle, char const* name) {
-    auto impl = DebugUtils::get();
-    if (!impl->mEnabled) {
-        return;
-    }
-    VkDebugUtilsObjectNameInfoEXT const info = {
-            .sType = VK_STRUCTURE_TYPE_DEBUG_UTILS_OBJECT_NAME_INFO_EXT,
-            .pNext = nullptr,
-            .objectType = type,
-            .objectHandle = handle,
-            .pObjectName = name,
-    };
-    vkSetDebugUtilsObjectNameEXT(impl->mDevice, &info);
-}
-#endif // FVK_EANBLED(FVK_DEBUG_DEBUG_UTILS)
+} // anonymous namespace

 using ImgUtil = VulkanImageUtility;

@@ -216,26 +159,35 @@ VulkanDriver::VulkanDriver(VulkanPlatform* platform, VulkanContext const& contex
      mReadPixels(mPlatform->getDevice()),
      mIsSRGBSwapChainSupported(mPlatform->getCustomization().isSRGBSwapChainSupported) {

-#if FVK_ENABLED(FVK_DEBUG_DEBUG_UTILS)
-    DebugUtils::mSingleton =
-            new DebugUtils(mPlatform->getInstance(), mPlatform->getDevice(), &context);
-#endif
-
 #if FVK_ENABLED(FVK_DEBUG_VALIDATION)
    UTILS_UNUSED const PFN_vkCreateDebugReportCallbackEXT createDebugReportCallback
            = vkCreateDebugReportCallbackEXT;
-    if (!context.isDebugUtilsSupported() && createDebugReportCallback) {
+    VkResult result;
+    if (mContext.isDebugUtilsSupported()) {
+        VkDebugUtilsMessengerCreateInfoEXT const createInfo = {
+                .sType = VK_STRUCTURE_TYPE_DEBUG_UTILS_MESSENGER_CREATE_INFO_EXT,
+                .pNext = nullptr,
+                .flags = 0,
+                .messageSeverity = VK_DEBUG_UTILS_MESSAGE_SEVERITY_WARNING_BIT_EXT
+                                   | VK_DEBUG_UTILS_MESSAGE_SEVERITY_ERROR_BIT_EXT,
+                .messageType = VK_DEBUG_UTILS_MESSAGE_TYPE_GENERAL_BIT_EXT
+                               | VK_DEBUG_UTILS_MESSAGE_TYPE_VALIDATION_BIT_EXT,
+                .pfnUserCallback = debugUtilsCallback,
+        };
+        result = vkCreateDebugUtilsMessengerEXT(mPlatform->getInstance(), &createInfo, VKALLOC,
+                &mDebugMessenger);
+        ASSERT_POSTCONDITION(result == VK_SUCCESS, "Unable to create Vulkan debug messenger.");
+    } else if (createDebugReportCallback) {
        VkDebugReportCallbackCreateInfoEXT const cbinfo = {
                .sType = VK_STRUCTURE_TYPE_DEBUG_REPORT_CALLBACK_CREATE_INFO_EXT,
                .flags = VK_DEBUG_REPORT_WARNING_BIT_EXT | VK_DEBUG_REPORT_ERROR_BIT_EXT,
                .pfnCallback = debugReportCallback,
        };
-        VkResult result = createDebugReportCallback(mPlatform->getInstance(), &cbinfo, VKALLOC,
+        result = createDebugReportCallback(mPlatform->getInstance(), &cbinfo, VKALLOC,
                &mDebugCallback);
        ASSERT_POSTCONDITION(result == VK_SUCCESS, "Unable to create Vulkan debug callback.");
    }
 #endif
-
    mTimestamps = std::make_unique<VulkanTimestamps>(mPlatform->getDevice());
    mCommands = std::make_unique<VulkanCommands>(mPlatform->getDevice(),
            mPlatform->getGraphicsQueue(), mPlatform->getGraphicsQueueFamilyIndex(), &mContext,
@@ -261,41 +213,6 @@ VulkanDriver::~VulkanDriver() noexcept = default;
 UTILS_NOINLINE
 Driver* VulkanDriver::create(VulkanPlatform* platform, VulkanContext const& context,
         Platform::DriverConfig const& driverConfig) noexcept {
-#if 0
-    // this is useful for development, but too verbose even for debug builds
-    // For reference on a 64-bits machine in Release mode:
-    //    VulkanSamplerGroup            :  24       few
-    //    HwStream                      :  24       few
-    //    VulkanFence                   :  40       few
-    //    VulkanProgram                 :  40       moderate
-    //    VulkanIndexBuffer             :  72       moderate
-    //    VulkanBufferObject            :  72       many
-    // -- less than or equal 80 bytes
-    //    VulkanRenderPrimitive         : 104       many
-    //    VulkanSwapChain               : 112       few
-    //    VulkanTimerQuery              : 168       few
-    // -- less than or equal 176 bytes
-    //    VulkanTexture                 : 232       moderate
-    //    VulkanVertexBuffer            : 312       moderate
-    //    VulkanRenderTarget            : 320       few
-    // -- less than or equal to 320 bytes
-
-    utils::slog.d
-           << "\nVulkanSwapChain: " << sizeof(VulkanSwapChain)
-           << "\nVulkanBufferObject: " << sizeof(VulkanBufferObject)
-           << "\nVulkanVertexBuffer: " << sizeof(VulkanVertexBuffer)
-           << "\nVulkanIndexBuffer: " << sizeof(VulkanIndexBuffer)
-           << "\nVulkanSamplerGroup: " << sizeof(VulkanSamplerGroup)
-           << "\nVulkanRenderPrimitive: " << sizeof(VulkanRenderPrimitive)
-           << "\nVulkanTexture: " << sizeof(VulkanTexture)
-           << "\nVulkanTimerQuery: " << sizeof(VulkanTimerQuery)
-           << "\nHwStream: " << sizeof(HwStream)
-           << "\nVulkanRenderTarget: " << sizeof(VulkanRenderTarget)
-           << "\nVulkanFence: " << sizeof(VulkanFence)
-           << "\nVulkanProgram: " << sizeof(VulkanProgram)
-           << utils::io::endl;
-#endif
-
    assert_invariant(platform);
    size_t defaultSize = FVK_HANDLE_ARENA_SIZE_IN_MB * 1024U * 1024U;
    Platform::DriverConfig validConfig {driverConfig};
@@ -334,12 +251,9 @@ void VulkanDriver::terminate() {
    if (mDebugCallback) {
        vkDestroyDebugReportCallbackEXT(mPlatform->getInstance(), mDebugCallback, VKALLOC);
    }
-
-#if FVK_ENABLED(FVK_DEBUG_DEBUG_UTILS)
-    assert_invariant(DebugUtils::mSingleton);
-    delete DebugUtils::mSingleton;
-#endif
-
+    if (mDebugMessenger) {
+        vkDestroyDebugUtilsMessengerEXT(mPlatform->getInstance(), mDebugMessenger, VKALLOC);
+    }
    mPlatform->terminate();
 }

@@ -410,10 +324,12 @@ void VulkanDriver::createSamplerGroupR(Handle<HwSamplerGroup> sbh, uint32_t coun

 void VulkanDriver::createRenderPrimitiveR(Handle<HwRenderPrimitive> rph,
        Handle<HwVertexBuffer> vbh, Handle<HwIndexBuffer> ibh,
-        PrimitiveType pt) {
+        PrimitiveType pt, uint32_t offset,
+        uint32_t minIndex, uint32_t maxIndex, uint32_t count) {
    auto rp = mResourceAllocator.construct<VulkanRenderPrimitive>(rph, &mResourceAllocator);
    mResourceManager.acquire(rp);
-    VulkanDriver::setRenderPrimitiveBuffer(rph, pt, vbh, ibh);
+    VulkanDriver::setRenderPrimitiveBuffer(rph, vbh, ibh);
+    VulkanDriver::setRenderPrimitiveRange(rph, pt, offset, minIndex, maxIndex, count);
 }

 void VulkanDriver::destroyRenderPrimitive(Handle<HwRenderPrimitive> rph) {
@@ -424,26 +340,10 @@ void VulkanDriver::destroyRenderPrimitive(Handle<HwRenderPrimitive> rph) {
    mResourceManager.release(rp);
 }

-void VulkanDriver::createVertexBufferInfoR(Handle<HwVertexBufferInfo> vbih, uint8_t bufferCount,
-        uint8_t attributeCount, AttributeArray attributes) {
-    auto vbi = mResourceAllocator.construct<VulkanVertexBufferInfo>(vbih,
-            bufferCount, attributeCount, attributes);
-    mResourceManager.acquire(vbi);
-}
-
-void VulkanDriver::destroyVertexBufferInfo(Handle<HwVertexBufferInfo> vbih) {
-    if (!vbih) {
-        return;
-    }
-    auto vbi = mResourceAllocator.handle_cast<VulkanVertexBufferInfo*>(vbih);
-    mResourceManager.release(vbi);
-}
-
-
-void VulkanDriver::createVertexBufferR(Handle<HwVertexBuffer> vbh,
-        uint32_t vertexCount, Handle<HwVertexBufferInfo> vbih) {
-    auto vertexBuffer = mResourceAllocator.construct<VulkanVertexBuffer>(vbh,
-            mContext, mStagePool, &mResourceAllocator, vertexCount, vbih);
+void VulkanDriver::createVertexBufferR(Handle<HwVertexBuffer> vbh, uint8_t bufferCount,
+        uint8_t attributeCount, uint32_t elementCount, AttributeArray attributes) {
+    auto vertexBuffer = mResourceAllocator.construct<VulkanVertexBuffer>(vbh, mContext, mStagePool,
+            &mResourceAllocator, bufferCount, attributeCount, elementCount, attributes);
    mResourceManager.acquire(vertexBuffer);
 }

@@ -650,10 +550,6 @@ void VulkanDriver::createTimerQueryR(Handle<HwTimerQuery> tqh, int) {
    // nothing to do, timer query was constructed in createTimerQueryS
 }

-Handle<HwVertexBufferInfo> VulkanDriver::createVertexBufferInfoS() noexcept {
-    return mResourceAllocator.allocHandle<VulkanVertexBufferInfo>();
-}
-
 Handle<HwVertexBuffer> VulkanDriver::createVertexBufferS() noexcept {
    return mResourceAllocator.allocHandle<VulkanVertexBuffer>();
 }
@@ -868,14 +764,8 @@ bool VulkanDriver::isSRGBSwapChainSupported() {
    return mIsSRGBSwapChainSupported;
 }

-bool VulkanDriver::isStereoSupported(backend::StereoscopicType stereoscopicType) {
-    switch (stereoscopicType) {
-    case backend::StereoscopicType::INSTANCED:
-        return true;
-    case backend::StereoscopicType::MULTIVIEW:
-        // TODO: implement multiview feature in Vulkan.
-        return false;
-    }
+bool VulkanDriver::isStereoSupported() {
+    return true;
 }

 bool VulkanDriver::isParallelShaderCompileSupported() {
@@ -974,7 +864,7 @@ void VulkanDriver::setVertexBufferObject(Handle<HwVertexBuffer> vbh, uint32_t in
    auto bo = mResourceAllocator.handle_cast<VulkanBufferObject*>(boh);
    assert_invariant(bo->bindingType == BufferObjectBinding::VERTEX);

-    vb->setBuffer(mResourceAllocator, bo, index);
+    vb->setBuffer(bo, index);
 }

 void VulkanDriver::updateIndexBuffer(Handle<HwIndexBuffer> ibh, BufferDescriptor&& p,
@@ -1138,7 +1028,7 @@ void VulkanDriver::beginRenderPass(Handle<HwRenderTarget> rth, const RenderPassP

    VulkanRenderTarget* const rt = mResourceAllocator.handle_cast<VulkanRenderTarget*>(rth);
    const VkExtent2D extent = rt->getExtent();
-    assert_invariant(rt == mDefaultRenderTarget || extent.width > 0 && extent.height > 0);
+    assert_invariant(extent.width > 0 && extent.height > 0);

    // Filament has the expectation that the contents of the swap chain are not preserved on the
    // first render pass. Note however that its contents are often preserved on subsequent render
@@ -1290,14 +1180,22 @@ void VulkanDriver::beginRenderPass(Handle<HwRenderTarget> rth, const RenderPassP
    }
    VkFramebuffer vkfb = mFramebufferCache.getFramebuffer(fbkey);

-// Assign a label to the framebuffer for debugging purposes.
-#if FVK_ENABLED(FVK_DEBUG_GROUP_MARKERS | FVK_DEBUG_DEBUG_UTILS)
-    auto const topMarker = mCommands->getTopGroupMarker();
-    if (!topMarker.empty()) {
-        DebugUtils::setName(VK_OBJECT_TYPE_FRAMEBUFFER, reinterpret_cast<uint64_t>(vkfb),
-                topMarker.c_str());
+    // Assign a label to the framebuffer for debugging purposes.
+    #if FVK_ENABLED(FVK_DEBUG_GROUP_MARKERS)
+    if (UTILS_UNLIKELY(mContext.isDebugUtilsSupported())) {
+        auto const topMarker = mCommands->getTopGroupMarker();
+        if (!topMarker.empty()) {
+            const VkDebugUtilsObjectNameInfoEXT info = {
+                VK_STRUCTURE_TYPE_DEBUG_UTILS_OBJECT_NAME_INFO_EXT,
+                nullptr,
+                VK_OBJECT_TYPE_FRAMEBUFFER,
+                reinterpret_cast<uint64_t>(vkfb),
+                topMarker.c_str(),
+            };
+            vkSetDebugUtilsObjectNameEXT(mPlatform->getDevice(), &info);
+        }
    }
-#endif
+    #endif

    // The current command buffer now owns a reference to the render target and its attachments.
    // Note that we must acquire parent textures, not sidecars.
@@ -1451,12 +1349,22 @@ void VulkanDriver::nextSubpass(int) {
    }
 }

-void VulkanDriver::setRenderPrimitiveBuffer(Handle<HwRenderPrimitive> rph, PrimitiveType pt,
+void VulkanDriver::setRenderPrimitiveBuffer(Handle<HwRenderPrimitive> rph,
        Handle<HwVertexBuffer> vbh, Handle<HwIndexBuffer> ibh) {
    auto primitive = mResourceAllocator.handle_cast<VulkanRenderPrimitive*>(rph);
    primitive->setBuffers(mResourceAllocator.handle_cast<VulkanVertexBuffer*>(vbh),
            mResourceAllocator.handle_cast<VulkanIndexBuffer*>(ibh));
-    primitive->setPrimitiveType(pt);
+}
+
+void VulkanDriver::setRenderPrimitiveRange(Handle<HwRenderPrimitive> rph,
+        PrimitiveType pt, uint32_t offset,
+        uint32_t minIndex, uint32_t maxIndex, uint32_t count) {
+    auto& primitive = *mResourceAllocator.handle_cast<VulkanRenderPrimitive*>(rph);
+    primitive.setPrimitiveType(pt);
+    primitive.offset = offset * primitive.indexBuffer->elementSize;
+    primitive.count = count;
+    primitive.minIndex = minIndex;
+    primitive.maxIndex = maxIndex > minIndex ? maxIndex : primitive.maxVertexCount - 1;
 }

 void VulkanDriver::makeCurrent(Handle<HwSwapChain> drawSch, Handle<HwSwapChain> readSch) {
@@ -1705,7 +1613,7 @@ void VulkanDriver::blitDEPRECATED(TargetBufferFlags buffers,
 }

 void VulkanDriver::draw(PipelineState pipelineState, Handle<HwRenderPrimitive> rph,
-        uint32_t const indexOffset, uint32_t const indexCount, uint32_t const instanceCount) {
+        const uint32_t instanceCount) {
    FVK_SYSTRACE_CONTEXT();
    FVK_SYSTRACE_START("draw");

@@ -1716,6 +1624,7 @@ void VulkanDriver::draw(PipelineState pipelineState, Handle<HwRenderPrimitive> r
    Handle<HwProgram> programHandle = pipelineState.program;
    RasterState rasterState = pipelineState.rasterState;
    PolygonOffset depthOffset = pipelineState.polygonOffset;
+    Viewport const& scissorBox = pipelineState.scissor;

    auto* program = mResourceAllocator.handle_cast<VulkanProgram*>(programHandle);
    commands->acquire(program);
@@ -1732,31 +1641,29 @@ void VulkanDriver::draw(PipelineState pipelineState, Handle<HwRenderPrimitive> r
    // Update the VK raster state.
    const VulkanRenderTarget* rt = mCurrentRenderPass.renderTarget;

-    VulkanPipelineCache::RasterState const vulkanRasterState{
-        .cullMode = getCullMode(rasterState.culling),
-        .frontFace = getFrontFace(rasterState.inverseFrontFaces),
-        .depthBiasEnable = (depthOffset.constant || depthOffset.slope) ? true : false,
-        .blendEnable = rasterState.hasBlending(),
-        .depthWriteEnable = rasterState.depthWrite,
-        .alphaToCoverageEnable = rasterState.alphaToCoverage,
-        .srcColorBlendFactor = getBlendFactor(rasterState.blendFunctionSrcRGB),
-        .dstColorBlendFactor = getBlendFactor(rasterState.blendFunctionDstRGB),
-        .srcAlphaBlendFactor = getBlendFactor(rasterState.blendFunctionSrcAlpha),
-        .dstAlphaBlendFactor = getBlendFactor(rasterState.blendFunctionDstAlpha),
-        .colorWriteMask = (VkColorComponentFlags) (rasterState.colorWrite ? 0xf : 0x0),
-        .rasterizationSamples = rt->getSamples(),
-        .colorTargetCount = rt->getColorTargetCount(mCurrentRenderPass),
-        .colorBlendOp = rasterState.blendEquationRGB,
-        .alphaBlendOp =  rasterState.blendEquationAlpha,
-        .depthCompareOp = rasterState.depthFunc,
-        .depthBiasConstantFactor = depthOffset.constant,
-        .depthBiasSlopeFactor = depthOffset.slope
-    };
+    auto vkraster = mPipelineCache.getCurrentRasterState();
+    vkraster.cullMode = getCullMode(rasterState.culling);
+    vkraster.frontFace = getFrontFace(rasterState.inverseFrontFaces);
+    vkraster.depthBiasEnable = (depthOffset.constant || depthOffset.slope) ? true : false;
+    vkraster.depthBiasConstantFactor = depthOffset.constant;
+    vkraster.depthBiasSlopeFactor = depthOffset.slope;
+    vkraster.blendEnable = rasterState.hasBlending();
+    vkraster.srcColorBlendFactor = getBlendFactor(rasterState.blendFunctionSrcRGB);
+    vkraster.dstColorBlendFactor = getBlendFactor(rasterState.blendFunctionDstRGB);
+    vkraster.colorBlendOp = rasterState.blendEquationRGB;
+    vkraster.srcAlphaBlendFactor = getBlendFactor(rasterState.blendFunctionSrcAlpha);
+    vkraster.dstAlphaBlendFactor = getBlendFactor(rasterState.blendFunctionDstAlpha);
+    vkraster.alphaBlendOp =  rasterState.blendEquationAlpha;
+    vkraster.colorWriteMask = (VkColorComponentFlags) (rasterState.colorWrite ? 0xf : 0x0);
+    vkraster.depthWriteEnable = rasterState.depthWrite;
+    vkraster.depthCompareOp = rasterState.depthFunc;
+    vkraster.rasterizationSamples = rt->getSamples();
+    vkraster.alphaToCoverageEnable = rasterState.alphaToCoverage;
+    vkraster.colorTargetCount = rt->getColorTargetCount(mCurrentRenderPass);
+    mPipelineCache.setCurrentRasterState(vkraster);

    // Declare fixed-size arrays that get passed to the pipeCache and to vkCmdBindVertexBuffers.
-    VulkanVertexBufferInfo const* const vbi =
-            mResourceAllocator.handle_cast<VulkanVertexBufferInfo*>(prim.vertexBuffer->vbih);
-    uint32_t const bufferCount = vbi->attributes.size();
+    uint32_t const bufferCount = prim.vertexBuffer->attributes.size();
    VkVertexInputAttributeDescription const* attribDesc = prim.vertexBuffer->getAttribDescriptions();
    VkVertexInputBindingDescription const* bufferDesc =  prim.vertexBuffer->getBufferDescriptions();
    VkBuffer const* buffers = prim.vertexBuffer->getVkBuffers();
@@ -1764,7 +1671,7 @@ void VulkanDriver::draw(PipelineState pipelineState, Handle<HwRenderPrimitive> r

    // Push state changes to the VulkanPipelineCache instance. This is fast and does not make VK calls.
    mPipelineCache.bindProgram(program);
-    mPipelineCache.bindRasterState(vulkanRasterState);
+    mPipelineCache.bindRasterState(mPipelineCache.getCurrentRasterState());
    mPipelineCache.bindPrimitiveTopology(prim.primitiveTopology);
    mPipelineCache.bindVertexArray(attribDesc, bufferDesc, bufferCount);

@@ -1849,37 +1756,6 @@ void VulkanDriver::draw(PipelineState pipelineState, Handle<HwRenderPrimitive> r
        return;
    }

-    // Bind a new pipeline if the pipeline state changed.
-    // If allocation failed, skip the draw call and bail. We do not emit an error since the
-    // validation layer will already do so.
-    if (!mPipelineCache.bindPipeline(commands)) {
-        return;
-    }
-
-    // Next bind the vertex buffers and index buffer. One potential performance improvement is to
-    // avoid rebinding these if they are already bound, but since we do not (yet) support subranges
-    // it would be rare for a client to make consecutive draw calls with the same render primitive.
-    vkCmdBindVertexBuffers(cmdbuffer, 0, bufferCount, buffers, offsets);
-    vkCmdBindIndexBuffer(cmdbuffer, prim.indexBuffer->buffer.getGpuBuffer(), 0,
-            prim.indexBuffer->indexType);
-
-    // Finally, make the actual draw call. TODO: support subranges
-    const uint32_t firstIndex = indexOffset;
-    const int32_t vertexOffset = 0;
-    const uint32_t firstInstId = 0;
-
-    vkCmdDrawIndexed(cmdbuffer, indexCount, instanceCount, firstIndex, vertexOffset, firstInstId);
-    FVK_SYSTRACE_END();
-}
-
-void VulkanDriver::dispatchCompute(Handle<HwProgram> program, math::uint3 workGroupCount) {
-    // FIXME: implement me
-}
-
-void VulkanDriver::scissor(Viewport scissorBox) {
-    VulkanCommandBuffer* commands = &mCommands->get();
-    VkCommandBuffer cmdbuffer = commands->buffer();
-
    // Set scissoring.
    // clamp left-bottom to 0,0 and avoid overflows
    constexpr int32_t maxvali  = std::numeric_limits<int32_t>::max();
@@ -1898,9 +1774,35 @@ void VulkanDriver::scissor(Viewport scissorBox) {
            .extent = { uint32_t(r - l), uint32_t(t - b) }
    };

-    const VulkanRenderTarget* rt = mCurrentRenderPass.renderTarget;
    rt->transformClientRectToPlatform(&scissor);
    mPipelineCache.bindScissor(cmdbuffer, scissor);
+
+    // Bind a new pipeline if the pipeline state changed.
+    // If allocation failed, skip the draw call and bail. We do not emit an error since the
+    // validation layer will already do so.
+    if (!mPipelineCache.bindPipeline(commands)) {
+        return;
+    }
+
+    // Next bind the vertex buffers and index buffer. One potential performance improvement is to
+    // avoid rebinding these if they are already bound, but since we do not (yet) support subranges
+    // it would be rare for a client to make consecutive draw calls with the same render primitive.
+    vkCmdBindVertexBuffers(cmdbuffer, 0, bufferCount, buffers, offsets);
+    vkCmdBindIndexBuffer(cmdbuffer, prim.indexBuffer->buffer.getGpuBuffer(), 0,
+            prim.indexBuffer->indexType);
+
+    // Finally, make the actual draw call. TODO: support subranges
+    const uint32_t indexCount = prim.count;
+    const uint32_t firstIndex = prim.offset / prim.indexBuffer->elementSize;
+    const int32_t vertexOffset = 0;
+    const uint32_t firstInstId = 0;
+
+    vkCmdDrawIndexed(cmdbuffer, indexCount, instanceCount, firstIndex, vertexOffset, firstInstId);
+    FVK_SYSTRACE_END();
+}
+
+void VulkanDriver::dispatchCompute(Handle<HwProgram> program, math::uint3 workGroupCount) {
+    // FIXME: implement me
 }

 void VulkanDriver::beginTimerQuery(Handle<HwTimerQuery> tqh) {
@@ -45,30 +45,6 @@ public:
    static Driver* create(VulkanPlatform* platform, VulkanContext const& context,
            Platform::DriverConfig const& driverConfig) noexcept;

-#if FVK_ENABLED(FVK_DEBUG_DEBUG_UTILS)
-    // Encapsulates the VK_EXT_debug_utils extension.  In particular, we use
-    // vkSetDebugUtilsObjectNameEXT and vkCreateDebugUtilsMessengerEXT
-    class DebugUtils {
-    public:
-        static void setName(VkObjectType type, uint64_t handle, char const* name);
-
-    private:
-        static DebugUtils* get();
-
-        DebugUtils(VkInstance instance, VkDevice device, VulkanContext const* context);
-        ~DebugUtils();
-
-        VkInstance const mInstance;
-        VkDevice const mDevice;
-        bool const mEnabled;
-        VkDebugUtilsMessengerEXT mDebugMessenger = VK_NULL_HANDLE;
-
-        static DebugUtils* mSingleton;
-
-        friend class VulkanDriver;
-    };
-#endif // FVK_ENABLED(FVK_DEBUG_DEBUG_UTILS)
-
 private:
    void debugCommandBegin(CommandStream* cmds, bool synchronous,
            const char* methodName) noexcept override;
@@ -101,10 +77,12 @@ private:
    VulkanDriver& operator=(VulkanDriver const&) = delete;

 private:
-    inline void setRenderPrimitiveBuffer(Handle<HwRenderPrimitive> rph, PrimitiveType pt,
-            Handle<HwVertexBuffer> vbh,
+    inline void setRenderPrimitiveBuffer(Handle<HwRenderPrimitive> rph, Handle<HwVertexBuffer> vbh,
            Handle<HwIndexBuffer> ibh);

+    inline void setRenderPrimitiveRange(Handle<HwRenderPrimitive> rph, PrimitiveType pt,
+            uint32_t offset, uint32_t minIndex, uint32_t maxIndex, uint32_t count);
+
    void collectGarbage();

    VulkanPlatform* mPlatform = nullptr;
@@ -117,6 +95,7 @@ private:
    VulkanRenderPass mCurrentRenderPass = {};
    VmaAllocator mAllocator = VK_NULL_HANDLE;
    VkDebugReportCallbackEXT mDebugCallback = VK_NULL_HANDLE;
+    VkDebugUtilsMessengerEXT mDebugMessenger = VK_NULL_HANDLE;

    VulkanContext mContext = {};
    VulkanResourceAllocator mResourceAllocator;
@@ -16,9 +16,7 @@

 #include "VulkanHandles.h"

-#include "VulkanDriver.h"
 #include "VulkanConstants.h"
-#include "VulkanDriver.h"
 #include "VulkanMemory.h"
 #include "spirv/VulkanSpirvUtils.h"

@@ -80,23 +78,6 @@ VulkanProgram::VulkanProgram(VkDevice device, Program const& builder) noexcept
        };
        VkResult result = vkCreateShaderModule(mDevice, &moduleInfo, VKALLOC, &module);
        ASSERT_POSTCONDITION(result == VK_SUCCESS, "Unable to create shader module.");
-
-#if FVK_ENABLED(FVK_DEBUG_DEBUG_UTILS)
-        std::string name{ builder.getName().c_str(), builder.getName().size() };
-        switch (static_cast<ShaderStage>(i)) {
-            case ShaderStage::VERTEX:
-                name += "_vs";
-                break;
-            case ShaderStage::FRAGMENT:
-                name += "_fs";
-                break;
-            default:
-                PANIC_POSTCONDITION("Unexpected stage");
-                break;
-        }
-        VulkanDriver::DebugUtils::setName(VK_OBJECT_TYPE_SHADER_MODULE,
-                reinterpret_cast<uint64_t>(module), name.c_str());
-#endif
    }

    auto& groupInfo = builder.getSamplerGroupInfo();
@@ -112,10 +93,10 @@ VulkanProgram::VulkanProgram(VkDevice device, Program const& builder) noexcept
        }
    }

-#if FVK_ENABLED(FVK_DEBUG_SHADER_MODULE)
-    utils::slog.d << "Created VulkanProgram " << builder << ", shaders = (" << bundle.vertex
-                  << ", " << bundle.fragment << ")" << utils::io::endl;
-#endif
+    #if FVK_ENABLED(FVK_DEBUG_SHADER_MODULE)
+        utils::slog.d << "Created VulkanProgram " << builder << ", shaders = (" << bundle.vertex
+                      << ", " << bundle.fragment << ")" << utils::io::endl;
+    #endif
 }

 VulkanProgram::VulkanProgram(VkDevice device, VkShaderModule vs, VkShaderModule fs,
@@ -281,23 +262,13 @@ uint8_t VulkanRenderTarget::getColorTargetCount(const VulkanRenderPass& pass) co
    return count;
 }

-VulkanVertexBufferInfo::VulkanVertexBufferInfo(uint8_t bufferCount, uint8_t attributeCount,
-        AttributeArray const& attributes)
-        : HwVertexBufferInfo(bufferCount, attributeCount),
-          VulkanResource(VulkanResourceType::VERTEX_BUFFER_INFO),
-          attributes(attributes) {
-}
-
 VulkanVertexBuffer::VulkanVertexBuffer(VulkanContext& context, VulkanStagePool& stagePool,
-        VulkanResourceAllocator* allocator, uint32_t vertexCount, Handle<HwVertexBufferInfo> vbih)
-    : HwVertexBuffer(vertexCount),
+        VulkanResourceAllocator* allocator, uint8_t bufferCount, uint8_t attributeCount,
+        uint32_t elementCount, AttributeArray const& attribs)
+    : HwVertexBuffer(bufferCount, attributeCount, elementCount, attribs),
      VulkanResource(VulkanResourceType::VERTEX_BUFFER),
-      vbih(vbih),
+      mInfo(new PipelineInfo(attribs.size())),
      mResources(allocator) {
-
-    VulkanVertexBufferInfo const* const vbi = allocator->handle_cast<VulkanVertexBufferInfo*>(vbih);
-    mInfo = new PipelineInfo(vbi->attributes.size());
-
    auto attribDesc = mInfo->mSoa.data<PipelineInfo::ATTRIBUTE_DESCRIPTION>();
    auto bufferDesc = mInfo->mSoa.data<PipelineInfo::BUFFER_DESCRIPTION>();
    auto offsets = mInfo->mSoa.data<PipelineInfo::OFFSETS>();
@@ -305,8 +276,8 @@ VulkanVertexBuffer::VulkanVertexBuffer(VulkanContext& context, VulkanStagePool&
    std::fill(mInfo->mSoa.begin<PipelineInfo::ATTRIBUTE_TO_BUFFER_INDEX>(),
            mInfo->mSoa.end<PipelineInfo::ATTRIBUTE_TO_BUFFER_INDEX>(), -1);

-    for (uint32_t attribIndex = 0; attribIndex < vbi->attributes.size(); attribIndex++) {
-        Attribute attrib = vbi->attributes[attribIndex];
+    for (uint32_t attribIndex = 0; attribIndex < attribs.size(); attribIndex++) {
+        Attribute attrib = attribs[attribIndex];
        bool const isInteger = attrib.flags & Attribute::FLAG_INTEGER_TARGET;
        bool const isNormalized = attrib.flags & Attribute::FLAG_NORMALIZED;
        VkFormat vkformat = getVkFormat(attrib.type, isNormalized, isInteger);
@@ -318,7 +289,7 @@ VulkanVertexBuffer::VulkanVertexBuffer(VulkanContext& context, VulkanStagePool&
        // expects to receive floats or ints.
        if (attrib.buffer == Attribute::BUFFER_UNUSED) {
            vkformat = isInteger ? VK_FORMAT_R8G8B8A8_UINT : VK_FORMAT_R8G8B8A8_SNORM;
-            attrib = vbi->attributes[0];
+            attrib = attribs[0];
        }
        offsets[attribIndex] = attrib.offset;
        attribDesc[attribIndex] = {
@@ -338,12 +309,8 @@ VulkanVertexBuffer::~VulkanVertexBuffer() {
    delete mInfo;
 }

-void VulkanVertexBuffer::setBuffer(VulkanResourceAllocator const& allocator,
-        VulkanBufferObject* bufferObject, uint32_t index) {
-
-    VulkanVertexBufferInfo const* const vbi =
-            const_cast<VulkanResourceAllocator&>(allocator).handle_cast<VulkanVertexBufferInfo*>(vbih);
-    size_t count = vbi->attributes.size();
+void VulkanVertexBuffer::setBuffer(VulkanBufferObject* bufferObject, uint32_t index) {
+    size_t count = attributes.size();
    auto vkbuffers = mInfo->mSoa.data<PipelineInfo::VK_BUFFER>();
    auto attribToBuffer = mInfo->mSoa.data<PipelineInfo::ATTRIBUTE_TO_BUFFER_INDEX>();
    for (uint8_t attribIndex = 0; attribIndex < count; attribIndex++) {
@@ -129,21 +129,14 @@ private:

 struct VulkanBufferObject;

-struct VulkanVertexBufferInfo : public HwVertexBufferInfo, VulkanResource {
-    VulkanVertexBufferInfo(uint8_t bufferCount, uint8_t attributeCount,
-            AttributeArray const& attributes);
-    AttributeArray attributes;
-};
-
 struct VulkanVertexBuffer : public HwVertexBuffer, VulkanResource {
    VulkanVertexBuffer(VulkanContext& context, VulkanStagePool& stagePool,
-            VulkanResourceAllocator* allocator,
-            uint32_t vertexCount, Handle<HwVertexBufferInfo> vbih);
+            VulkanResourceAllocator* allocator, uint8_t bufferCount, uint8_t attributeCount,
+            uint32_t elementCount, AttributeArray const& attributes);

    ~VulkanVertexBuffer();

-    void setBuffer(VulkanResourceAllocator const& allocator,
-            VulkanBufferObject* bufferObject, uint32_t index);
+    void setBuffer(VulkanBufferObject* bufferObject, uint32_t index);

    inline VkVertexInputAttributeDescription const* getAttribDescriptions() {
        return mInfo->mSoa.data<PipelineInfo::ATTRIBUTE_DESCRIPTION>();
@@ -161,8 +154,6 @@ struct VulkanVertexBuffer : public HwVertexBuffer, VulkanResource {
        return mInfo->mSoa.data<PipelineInfo::OFFSETS>();
    }

-    Handle<HwVertexBufferInfo> vbih;
-
 private:
    struct PipelineInfo {
        PipelineInfo(size_t size)
@@ -170,7 +161,7 @@ private:
            mSoa.resize(size);
        }

-        // These correspond to the index of the element in the SoA
+        // These corresponds to the index of the element in the SoA
        static constexpr uint8_t ATTRIBUTE_DESCRIPTION = 0;
        static constexpr uint8_t BUFFER_DESCRIPTION = 1;
        static constexpr uint8_t VK_BUFFER = 2;
@@ -34,6 +34,8 @@ using namespace bluevk;

 namespace filament::backend {

+static VulkanPipelineCache::RasterState createDefaultRasterState();
+
 static VkShaderStageFlags getShaderStageFlags(VulkanPipelineCache::UsageFlags key, uint16_t binding) {
    // NOTE: if you modify this function, you also need to modify getUsageFlags.
    assert_invariant(binding < MAX_SAMPLER_COUNT);
@@ -71,7 +73,8 @@ VulkanPipelineCache::UsageFlags VulkanPipelineCache::disableUsageFlags(uint16_t
 }

 VulkanPipelineCache::VulkanPipelineCache(VulkanResourceAllocator* allocator)
-    : mResourceAllocator(allocator),
+    : mCurrentRasterState(createDefaultRasterState()),
+      mResourceAllocator(allocator),
      mPipelineBoundResources(allocator) {
    mDummyBufferWriteInfo.sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET;
    mDummyBufferWriteInfo.pNext = nullptr;
@@ -566,7 +569,7 @@ void VulkanPipelineCache::bindProgram(VulkanProgram* program) noexcept {
 }

 void VulkanPipelineCache::bindRasterState(const RasterState& rasterState) noexcept {
-    mPipelineRequirements.rasterState = rasterState;
+    mPipelineRequirements.rasterState = mCurrentRasterState = rasterState;
 }

 void VulkanPipelineCache::bindRenderPass(VkRenderPass renderPass, int subpassIndex) noexcept {
@@ -914,6 +917,23 @@ bool VulkanPipelineCache::DescEqual::operator()(const DescriptorKey& k1,
    return true;
 }

+static VulkanPipelineCache::RasterState createDefaultRasterState() {
+    return VulkanPipelineCache::RasterState {
+        .cullMode = VK_CULL_MODE_NONE,
+        .frontFace = VK_FRONT_FACE_COUNTER_CLOCKWISE,
+        .depthBiasEnable = VK_FALSE,
+        .blendEnable = VK_FALSE,
+        .depthWriteEnable = VK_TRUE,
+        .alphaToCoverageEnable = true,
+        .colorWriteMask = 0xf,
+        .rasterizationSamples = VK_SAMPLE_COUNT_1_BIT,
+        .colorTargetCount = 1,
+        .depthCompareOp = SamplerCompareFunc::LE,
+        .depthBiasConstantFactor = 0.0f,
+        .depthBiasSlopeFactor = 0.0f,
+    };
+}
+
 } // namespace filament::backend

 #pragma clang diagnostic pop
@@ -199,6 +199,15 @@ public:
        mPipelineBoundResources.acquire(resource);
    }

+    inline RasterState getCurrentRasterState() const noexcept {
+        return mCurrentRasterState;
+    }
+
+    // We need to update this outside of bindRasterState due to VulkanDriver::draw.
+    inline void setCurrentRasterState(RasterState const& rasterState) noexcept {
+        mCurrentRasterState = rasterState;
+    }
+
 private:
    // PIPELINE LAYOUT CACHE KEY
    // -------------------------
@@ -404,6 +413,7 @@ private:
    VmaAllocator mAllocator = VK_NULL_HANDLE;

    // Current requirements for the pipeline layout, pipeline, and descriptor sets.
+    RasterState mCurrentRasterState;
    PipelineKey mPipelineRequirements = {};
    DescriptorKey mDescriptorRequirements = {};

@@ -53,9 +53,6 @@ void deallocateResource(VulkanResourceAllocator* allocator, VulkanResourceType t
        case VulkanResourceType::TIMER_QUERY:
            allocator->destruct<VulkanTimerQuery>(Handle<HwTimerQuery>(id));
            break;
-        case VulkanResourceType::VERTEX_BUFFER_INFO:
-            allocator->destruct<VulkanVertexBufferInfo>(Handle<HwVertexBufferInfo>(id));
-            break;
        case VulkanResourceType::VERTEX_BUFFER:
            allocator->destruct<VulkanVertexBuffer>(Handle<HwVertexBuffer>(id));
            break;
@@ -43,7 +43,6 @@ enum class VulkanResourceType : uint8_t {
    TEXTURE,
    TIMER_QUERY,
    VERTEX_BUFFER,
-    VERTEX_BUFFER_INFO,

    // Below are resources that are managed manually (i.e. not ref counted).
    FENCE,
@@ -63,16 +62,15 @@ struct VulkanResourceBase {
 protected:
    explicit VulkanResourceBase(VulkanResourceType type)
        : mRefCount(IS_HEAP_ALLOC_TYPE(type) ? 1 : 0),
-          mType(uint32_t(type)),
-          mHandleId(0) {
-    }
+          mType(type),
+          mHandleId(0) {}

 private:
-    inline VulkanResourceType getType() const noexcept {
-        return VulkanResourceType(mType);
+    inline VulkanResourceType getType() {
+        return mType;
    }

-    inline HandleBase::HandleId getId() const noexcept {
+    inline HandleBase::HandleId getId() {
        return mHandleId;
    }

@@ -81,27 +79,25 @@ private:
    }

    inline void ref() noexcept {
-        if (IS_HEAP_ALLOC_TYPE(getType())) {
+        if (IS_HEAP_ALLOC_TYPE(mType)) {
            return;
        }
-        assert_invariant(mRefCount < ((1<<24) - 1));
        ++mRefCount;
    }

    inline void deref() noexcept {
-        if (IS_HEAP_ALLOC_TYPE(getType())) {
+        if (IS_HEAP_ALLOC_TYPE(mType)) {
            return;
        }
-        assert_invariant(mRefCount > 0);
        --mRefCount;
    }

-    inline size_t refcount() const noexcept {
+    inline size_t refcount() noexcept {
        return mRefCount;
    }

-    uint32_t mRefCount : 24; // 16M is enough for the refcount
-    uint32_t mType : 8; // must be uint32_t or MSVC doesn't pack it. no codegen impact w/ clang.
+    size_t mRefCount = 0;
+    VulkanResourceType mType = VulkanResourceType::BUFFER_OBJECT;
    HandleBase::HandleId mHandleId;

    friend struct VulkanThreadSafeResource;
@@ -111,8 +107,6 @@ private:
    friend class VulkanResourceManagerImpl;
 };

-static_assert(sizeof(VulkanResourceBase) == 8, "VulkanResourceBase should be 8 bytes");
-
 struct VulkanThreadSafeResource {
 protected:
    explicit VulkanThreadSafeResource(VulkanResourceType type)
@@ -23,7 +23,6 @@

 #include <bluevk/BlueVK.h>
 #include <utils/PrivateImplementation-impl.h>
-#include <utils/Panic.h>

 #define SWAPCHAIN_RET_FUNC(func, handle, ...)                                                      \
    if (mImpl->mSurfaceSwapChains.find(handle) != mImpl->mSurfaceSwapChains.end()) {               \
@@ -159,9 +158,7 @@ ExtensionSet getInstanceExtensions() {
            VK_KHR_GET_PHYSICAL_DEVICE_PROPERTIES_2_EXTENSION_NAME,

            // Request these if available.
-#if FVK_ENABLED(FVK_DEBUG_DEBUG_UTILS)
            VK_EXT_DEBUG_UTILS_EXTENSION_NAME,
-#endif
            VK_KHR_PORTABILITY_ENUMERATION_EXTENSION_NAME,

 #if FVK_ENABLED(FVK_DEBUG_VALIDATION)
@@ -184,9 +181,7 @@ ExtensionSet getInstanceExtensions() {

 ExtensionSet getDeviceExtensions(VkPhysicalDevice device) {
    std::string_view const TARGET_EXTS[] = {
-#if FVK_ENABLED(FVK_DEBUG_DEBUG_UTILS)
            VK_EXT_DEBUG_MARKER_EXTENSION_NAME,
-#endif
            VK_KHR_PORTABILITY_SUBSET_EXTENSION_NAME,
            VK_KHR_MAINTENANCE1_EXTENSION_NAME,
            VK_KHR_MAINTENANCE2_EXTENSION_NAME,
@@ -342,22 +337,36 @@ std::tuple<ExtensionSet, ExtensionSet> pruneExtensions(VkPhysicalDevice device,
        ExtensionSet const& instExts, ExtensionSet const& deviceExts) {
    ExtensionSet newInstExts = instExts;
    ExtensionSet newDeviceExts = deviceExts;
+    if (vkGetPhysicalDeviceProperties2KHR) {
+        VkPhysicalDeviceDriverProperties driverProperties = {
+                .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_DRIVER_PROPERTIES,
+        };
+        VkPhysicalDeviceProperties2 physicalDeviceProperties2 = {
+                .sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROPERTIES_2,
+                .pNext = &driverProperties,
+        };
+        vkGetPhysicalDeviceProperties2KHR(device, &physicalDeviceProperties2);
+        char* driverInfo = driverProperties.driverInfo;
+
+        if (newInstExts.find(VK_EXT_DEBUG_UTILS_EXTENSION_NAME) != newInstExts.end()) {
+            // Workaround for Mesa drivers. See issue #6192
+            if ((driverInfo && strstr(driverInfo, "Mesa"))) {
+                newInstExts.erase(VK_EXT_DEBUG_UTILS_EXTENSION_NAME);
+            }
+        }
+    }

-#if FVK_ENABLED(FVK_DEBUG_DEBUG_UTILS)    
    // debugUtils and debugMarkers extensions are used mutually exclusively.
    if (newInstExts.find(VK_EXT_DEBUG_UTILS_EXTENSION_NAME) != newInstExts.end()
            && newDeviceExts.find(VK_EXT_DEBUG_MARKER_EXTENSION_NAME) != newDeviceExts.end()) {
        newDeviceExts.erase(VK_EXT_DEBUG_MARKER_EXTENSION_NAME);
    }
-#endif    

-#if FVK_ENABLED(FVK_DEBUG_VALIDATION)
    // debugMarker must also request debugReport the instance extension. So check if that's present.
    if (newDeviceExts.find(VK_EXT_DEBUG_MARKER_EXTENSION_NAME) != newDeviceExts.end()
            && newInstExts.find(VK_EXT_DEBUG_REPORT_EXTENSION_NAME) == newInstExts.end()) {
        newDeviceExts.erase(VK_EXT_DEBUG_MARKER_EXTENSION_NAME);
    }
-#endif

    return std::tuple(newInstExts, newDeviceExts);
 }
@@ -662,12 +671,6 @@ Driver* VulkanPlatform::createDriver(void* sharedContext,
    context.mDebugMarkersSupported
            = deviceExts.find(VK_EXT_DEBUG_MARKER_EXTENSION_NAME) != deviceExts.end();

-#ifdef NDEBUG
-    // If we are in release build, we should not have turned on debug extensions
-    ASSERT_POSTCONDITION(!context.mDebugUtilsSupported && !context.mDebugMarkersSupported,
-            "Debug utils should not be enabled in release build.");
-#endif
-
    context.mDepthFormats = findAttachmentDepthFormats(mImpl->mPhysicalDevice);

    assert_invariant(context.mDepthFormats.size() > 0);
@@ -15,7 +15,6 @@
 */

 #include <backend/platforms/VulkanPlatform.h>
-#include <backend/DriverEnums.h>

 #include "vulkan/VulkanConstants.h"
 #include "vulkan/VulkanDriverFactory.h"
@@ -24,11 +23,6 @@

 #include <bluevk/BlueVK.h>

-#include <tuple>
-
-#include <stdint.h>
-#include <stddef.h>
-
 #if defined(__linux__) || defined(__FreeBSD__)
 #define LINUX_OR_FREEBSD 1
 #endif
@@ -143,7 +143,7 @@ void BackendTest::renderTriangle(Handle<HwRenderTarget> renderTarget,
    state.rasterState.depthFunc = RasterState::DepthFunc::A;
    state.rasterState.culling = CullingMode::NONE;

-    api.draw(state, triangle.getRenderPrimitive(), 0, 3, 1);
+    api.draw(state, triangle.getRenderPrimitive(), 1);

    api.endRenderPass();
 }
@@ -48,9 +48,7 @@ TrianglePrimitive::TrianglePrimitive(filament::backend::DriverApi& driverApi,

    const size_t size = sizeof(math::float2) * 3;
    mBufferObject = mDriverApi.createBufferObject(size, BufferObjectBinding::VERTEX, BufferUsage::STATIC);
-    mVertexBufferInfo = mDriverApi.createVertexBufferInfo(1, 1, attributes);
-    mVertexBuffer = mDriverApi.createVertexBuffer(mVertexCount, mVertexBufferInfo);
-
+    mVertexBuffer = mDriverApi.createVertexBuffer(1, 1, mVertexCount, attributes);
    mDriverApi.setVertexBufferObject(mVertexBuffer, 0, mBufferObject);
    BufferDescriptor vertexBufferDesc(gVertices, size);
    mDriverApi.updateBufferObject(mBufferObject, std::move(vertexBufferDesc), 0);
@@ -62,7 +60,7 @@ TrianglePrimitive::TrianglePrimitive(filament::backend::DriverApi& driverApi,
    mDriverApi.updateIndexBuffer(mIndexBuffer, std::move(indexBufferDesc), 0);

    mRenderPrimitive = mDriverApi.createRenderPrimitive(
-            mVertexBuffer, mIndexBuffer, PrimitiveType::TRIANGLES);
+            mVertexBuffer, mIndexBuffer, PrimitiveType::TRIANGLES, 0, 0, 2, 3);
 }

 void TrianglePrimitive::updateVertices(const filament::math::float2 vertices[3]) noexcept {
@@ -104,7 +102,6 @@ void TrianglePrimitive::updateIndices(const index_type* indices, int count, int
 TrianglePrimitive::~TrianglePrimitive() {
    mDriverApi.destroyBufferObject(mBufferObject);
    mDriverApi.destroyVertexBuffer(mVertexBuffer);
-    mDriverApi.destroyVertexBufferInfo(mVertexBufferInfo);
    mDriverApi.destroyIndexBuffer(mIndexBuffer);
    mDriverApi.destroyRenderPrimitive(mRenderPrimitive);
 }
@@ -40,7 +40,6 @@ public:

    using PrimitiveHandle = filament::backend::Handle<filament::backend::HwRenderPrimitive>;
    using BufferObjectHandle = filament::backend::Handle<filament::backend::HwBufferObject>;
-    using VertexInfoHandle = filament::backend::Handle<filament::backend::HwVertexBufferInfo>;
    using VertexHandle = filament::backend::Handle<filament::backend::HwVertexBuffer>;
    using IndexHandle = filament::backend::Handle<filament::backend::HwIndexBuffer>;

@@ -64,7 +63,6 @@ private:

    PrimitiveHandle mRenderPrimitive;
    BufferObjectHandle mBufferObject;
-    VertexInfoHandle mVertexBufferInfo;
    VertexHandle mVertexBuffer;
    IndexHandle mIndexBuffer;

@@ -405,7 +405,7 @@ TEST_F(BackendTest, ColorResolve) {
    api.beginFrame(0, 0);
        api.beginRenderPass(srcRenderTarget, params);
            api.bindUniformBuffer(0, ubuffer);
-            api.draw(state, triangle.getRenderPrimitive(), 0, 3, 1);
+            api.draw(state, triangle.getRenderPrimitive(), 1);
        api.endRenderPass();
    api.endFrame(0);

@@ -172,7 +172,7 @@ TEST_F(BackendTest, VertexBufferUpdate) {
            }

            getDriverApi().beginRenderPass(defaultRenderTarget, params);
-            getDriverApi().draw(state, triangle.getRenderPrimitive(), 0, 3, 1);
+            getDriverApi().draw(state, triangle.getRenderPrimitive(), 1);
            getDriverApi().endRenderPass();

            triangleIndex++;
@@ -218,7 +218,7 @@ TEST_F(BackendTest, FeedbackLoops) {
                    .sourceLevel = float(sourceLevel),
                });
                api.beginRenderPass(renderTargets[targetLevel], params);
-                api.draw(state, triangle.getRenderPrimitive(), 0, 3, 1);
+                api.draw(state, triangle.getRenderPrimitive(), 1);
                api.endRenderPass();
            }

@@ -237,7 +237,7 @@ TEST_F(BackendTest, FeedbackLoops) {
                    .sourceLevel = float(sourceLevel),
                });
                api.beginRenderPass(renderTargets[targetLevel], params);
-                api.draw(state, triangle.getRenderPrimitive(), 0, 3, 1);
+                api.draw(state, triangle.getRenderPrimitive(), 1);
                api.endRenderPass();
            }

@@ -130,7 +130,7 @@ TEST_F(BackendTest, MRT) {

        // Draw a triangle.
        getDriverApi().beginRenderPass(renderTarget, params);
-        getDriverApi().draw(state, triangle.getRenderPrimitive(), 0, 3, 1);
+        getDriverApi().draw(state, triangle.getRenderPrimitive(), 1);
        getDriverApi().endRenderPass();

        getDriverApi().flush();
@@ -168,7 +168,7 @@ TEST_F(BackendTest, SetMinMaxLevel) {
            ps.rasterState.colorWrite = true;
            ps.rasterState.depthWrite = false;
            api.beginRenderPass(renderTarget, params);
-            api.draw(ps, triangle.getRenderPrimitive(), 0, 3, 1);
+            api.draw(ps, triangle.getRenderPrimitive(), 1);
            api.endRenderPass();
        }

@@ -182,6 +182,7 @@ TEST_F(BackendTest, SetMinMaxLevel) {
        params.flags.discardEnd = TargetBufferFlags::NONE;

        PipelineState state;
+        state.scissor = params.viewport;
        state.program = textureProgram;
        state.rasterState.colorWrite = true;
        state.rasterState.depthWrite = false;
@@ -202,8 +203,7 @@ TEST_F(BackendTest, SetMinMaxLevel) {
        // Because the min level is 1, the result color should be the white triangle drawn in the
        // previous pass.
        api.beginRenderPass(defaultRenderTarget, params);
-        api.scissor(params.viewport);
-        api.draw(state, triangle.getRenderPrimitive(), 0, 3, 1);
+        api.draw(state, triangle.getRenderPrimitive(), 1);
        api.endRenderPass();

        // Adjust the base mip to 2.
@@ -221,8 +221,7 @@ TEST_F(BackendTest, SetMinMaxLevel) {
        params.flags.clear = TargetBufferFlags::NONE;
        params.flags.discardStart = TargetBufferFlags::NONE;
        api.beginRenderPass(defaultRenderTarget, params);
-        api.scissor(params.viewport);
-        api.draw(state, triangle.getRenderPrimitive(), 0, 3, 1);
+        api.draw(state, triangle.getRenderPrimitive(), 1);
        api.endRenderPass();

        api.commit(swapChain);
@@ -104,7 +104,7 @@ TEST_F(BackendTest, MissingRequiredAttributes) {

        // Render a triangle.
        getDriverApi().beginRenderPass(defaultRenderTarget, params);
-        getDriverApi().draw(state, triangle.getRenderPrimitive(), 0, 3, 1);
+        getDriverApi().draw(state, triangle.getRenderPrimitive(), 1);
        getDriverApi().endRenderPass();

        getDriverApi().flush();
@@ -309,7 +309,7 @@ TEST_F(ReadPixelsTest, ReadPixels) {
        state.rasterState.depthWrite = false;
        state.rasterState.depthFunc = RasterState::DepthFunc::A;
        state.rasterState.culling = CullingMode::NONE;
-        getDriverApi().draw(state, triangle.getRenderPrimitive(), 0, 3, 1);
+        getDriverApi().draw(state, triangle.getRenderPrimitive(), 1);

        getDriverApi().endRenderPass();

@@ -438,7 +438,7 @@ TEST_F(ReadPixelsTest, ReadPixelsPerformance) {

        // Render some content, just so we don't read back uninitialized data.
        getDriverApi().beginRenderPass(renderTarget, params);
-        getDriverApi().draw(state, triangle.getRenderPrimitive(), 0, 3, 1);
+        getDriverApi().draw(state, triangle.getRenderPrimitive(), 1);
        getDriverApi().endRenderPass();

        PixelBufferDescriptor descriptor(buffer, renderTargetSize * renderTargetSize * 4,
@@ -120,7 +120,7 @@ TEST_F(BackendTest, RenderExternalImageWithoutSet) {

    // Render a triangle.
    getDriverApi().beginRenderPass(defaultRenderTarget, params);
-    getDriverApi().draw(state, triangle.getRenderPrimitive(), 0, 3, 1);
+    getDriverApi().draw(state, triangle.getRenderPrimitive(), 1);
    getDriverApi().endRenderPass();

    getDriverApi().flush();
@@ -242,7 +242,7 @@ TEST_F(BackendTest, RenderExternalImage) {

    // Render a triangle.
    getDriverApi().beginRenderPass(defaultRenderTarget, params);
-    getDriverApi().draw(state, triangle.getRenderPrimitive(), 0, 3, 1);
+    getDriverApi().draw(state, triangle.getRenderPrimitive(), 1);
    getDriverApi().endRenderPass();

    getDriverApi().flush();
--- a/Show More
+++ b/Show More