Make some headers public so JobSystem.h can be used

certain material properties can only be set if supported by the 
material.

.github/workflows/release.yml

@@ -39,8 +39,8 @@ jobs:
         run: |
           REF=${RELEASE_TAG:-${GITHUB_REF}}
           TAG=${REF##*/}
-          echo ::set-output name=ref::${REF}
-          echo ::set-output name=tag::${TAG}
+          echo "ref=${REF}" >> $GITHUB_OUTPUT
+          echo "tag=${TAG}" >> $GITHUB_OUTPUT
       - uses: actions/checkout@v2.0.0
         with:
           ref: ${{ steps.git_ref.outputs.ref }}
@@ -69,8 +69,8 @@ jobs:
         run: |
           REF=${RELEASE_TAG:-${GITHUB_REF}}
           TAG=${REF##*/}
-          echo ::set-output name=ref::${REF}
-          echo ::set-output name=tag::${TAG}
+          echo "ref=${REF}" >> $GITHUB_OUTPUT
+          echo "tag=${TAG}" >> $GITHUB_OUTPUT
       - uses: actions/checkout@v2.0.0
         with:
           ref: ${{ steps.git_ref.outputs.ref }}
@@ -97,8 +97,8 @@ jobs:
         run: |
           REF=${RELEASE_TAG:-${GITHUB_REF}}
           TAG=${REF##*/}
-          echo ::set-output name=ref::${REF}
-          echo ::set-output name=tag::${TAG}
+          echo "ref=${REF}" >> $GITHUB_OUTPUT
+          echo "tag=${TAG}" >> $GITHUB_OUTPUT
       - uses: actions/checkout@v2.0.0
         with:
           ref: ${{ steps.git_ref.outputs.ref }}
@@ -143,8 +143,8 @@ jobs:
         run: |
           REF=${RELEASE_TAG:-${GITHUB_REF}}
           TAG=${REF##*/}
-          echo ::set-output name=ref::${REF}
-          echo ::set-output name=tag::${TAG}
+          echo "ref=${REF}" >> $GITHUB_OUTPUT
+          echo "tag=${TAG}" >> $GITHUB_OUTPUT
       - uses: actions/checkout@v2.0.0
         with:
           ref: ${{ steps.git_ref.outputs.ref }}
@@ -171,8 +171,8 @@ jobs:
         run: |
           REF=${RELEASE_TAG:-${GITHUB_REF}}
           TAG=${REF##*/}
-          echo ::set-output name=ref::${REF}
-          echo ::set-output name=tag::${TAG}
+          echo "ref=${REF}" >> $GITHUB_OUTPUT
+          echo "tag=${TAG}" >> $GITHUB_OUTPUT
         shell: bash
       - uses: actions/checkout@v2.0.0
         with:

CMakeLists.txt

@@ -337,19 +337,23 @@ endif()
 # ==================================================================================================
 # Release compiler flags
 # ==================================================================================================
-if (NOT MSVC)
+if (NOT MSVC AND NOT IOS)
+    # Omitting stack frame pointers prevents the generation of readable stack traces in crash reports on iOS
     set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} -fomit-frame-pointer")
 
     # These aren't compatible with -fembed-bitcode (and seem to have no effect on Apple platforms anyway)
-    if (NOT IOS)
-        set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} -ffunction-sections -fdata-sections")
-    endif()
+    set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} -ffunction-sections -fdata-sections")
 endif()
 
 # On Android RELEASE builds, we disable exceptions and RTTI to save some space (about 75 KiB
 # saved by -fno-exception and 10 KiB saved by -fno-rtti).
 if (ANDROID OR IOS OR WEBGL)
-    set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} -fno-exceptions -fno-unwind-tables -fno-asynchronous-unwind-tables -fno-rtti")
+    set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} -fno-exceptions -fno-rtti")
+
+    if (ANDROID OR WEBGL)
+        # Omitting unwind info prevents the generation of readable stack traces in crash reports on iOS
+        set(CMAKE_CXX_FLAGS_RELEASE "${CMAKE_CXX_FLAGS_RELEASE} -fno-unwind-tables -fno-asynchronous-unwind-tables")
+    endif()
 endif()
 
 # With WebGL, we disable RTTI even for debug builds because we pass emscripten::val back and forth
@@ -666,7 +670,7 @@ add_subdirectory(${EXTERNAL}/imgui/tnt)
 add_subdirectory(${EXTERNAL}/robin-map/tnt)
 add_subdirectory(${EXTERNAL}/smol-v/tnt)
 add_subdirectory(${EXTERNAL}/benchmark/tnt)
-add_subdirectory(${EXTERNAL}/meshoptimizer)
+add_subdirectory(${EXTERNAL}/meshoptimizer/tnt)
 add_subdirectory(${EXTERNAL}/cgltf/tnt)
 add_subdirectory(${EXTERNAL}/draco/tnt)
 add_subdirectory(${EXTERNAL}/jsmn/tnt)

README.md

@@ -31,7 +31,7 @@ repositories {
 }
 
 dependencies {
-    implementation 'com.google.android.filament:filament-android:1.28.1'
+    implementation 'com.google.android.filament:filament-android:1.28.2'
 }
 ```
 
@@ -51,7 +51,7 @@ Here are all the libraries available in the group `com.google.android.filament`:
 iOS projects can use CocoaPods to install the latest release:
 
 ```
-pod 'Filament', '~> 1.28.1'
+pod 'Filament', '~> 1.28.2'
 ```
 
 ### Snapshots
@@ -179,6 +179,7 @@ steps:
   - [x] KHR_mesh_quantization
   - [x] KHR_texture_basisu
   - [x] KHR_texture_transform
+  - [x] EXT_meshopt_compression
 
 
 ## Rendering with Filament

RELEASE_NOTES.md

@@ -5,6 +5,18 @@ A new header is inserted each time a *tag* is created.
 
 ## main branch
 
+- backend: add support for GGP platform
+- gltfio: calculate primitive's AABB correctly. 
+- gltfio: fix glTF breaking issue
+- gltfio: recompute bounding boxes with morph targets
+- engine: primitives with `CullingMode::FRONT_AND_BACK` are now skipped.
+- engine: add missing getters on `MaterialInstance`
+
+## v1.28.2
+
+- gltfio: support EXT_meshopt_compression
+- release packaging: fixed location of default IBL file
+
 ## v1.28.1
 
 - gltfio: fix reloading crash in ubershader mode
@@ -35,6 +47,7 @@ A new header is inserted each time a *tag* is created.
 - gltfio: fix regression with meshes that have no material
 - gltfio: fix regression with recomputeBoundingBoxes()
 - filamesh / matinfo: fix minor ASAN issues
+- engine: Added `Engine::resetBackendState()`, available only in WebGL builds
 
 ## v1.27.1
 

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

@@ -452,3 +452,84 @@ Java_com_google_android_filament_MaterialInstance_nDuplicate(JNIEnv* env, jclass
     }
     return (jlong)mi;
 }
+
+extern "C"
+JNIEXPORT jfloat JNICALL
+Java_com_google_android_filament_MaterialInstance_nGetMaskThreshold(JNIEnv* env, jclass clazz,
+        jlong nativeMaterialInstance) {
+    // TODO: implement nGetMaskThreshold()
+    MaterialInstance* instance = (MaterialInstance*)nativeMaterialInstance;
+    return instance->getMaskThreshold();
+}
+
+extern "C"
+JNIEXPORT jfloat JNICALL
+Java_com_google_android_filament_MaterialInstance_nGetSpecularAntiAliasingVariance(JNIEnv* env,
+        jclass clazz, jlong nativeMaterialInstance) {
+    // TODO: implement nGetSpecularAntiAliasingVariance()
+    MaterialInstance* instance = (MaterialInstance*)nativeMaterialInstance;
+    return instance->getSpecularAntiAliasingVariance();
+}
+
+extern "C"
+JNIEXPORT jfloat JNICALL
+Java_com_google_android_filament_MaterialInstance_nGetSpecularAntiAliasingThreshold(JNIEnv* env,
+        jclass clazz, jlong nativeMaterialInstance) {
+    // TODO: implement nGetSpecularAntiAliasingThreshold()
+    MaterialInstance* instance = (MaterialInstance*)nativeMaterialInstance;
+    return instance->getSpecularAntiAliasingThreshold();
+}
+
+extern "C"
+JNIEXPORT jboolean JNICALL
+Java_com_google_android_filament_MaterialInstance_nIsDoubleSided(JNIEnv* env, jclass clazz,
+        jlong nativeMaterialInstance) {
+    // TODO: implement nIsDoubleSided()
+    MaterialInstance* instance = (MaterialInstance*)nativeMaterialInstance;
+    return instance->isDoubleSided();
+}
+
+extern "C"
+JNIEXPORT jint JNICALL
+Java_com_google_android_filament_MaterialInstance_nGetCullingMode(JNIEnv* env, jclass clazz,
+        jlong nativeMaterialInstance) {
+    // TODO: implement nGetCullingMode()
+    MaterialInstance* instance = (MaterialInstance*)nativeMaterialInstance;
+    return (jint)instance->getCullingMode();
+}
+
+extern "C"
+JNIEXPORT jboolean JNICALL
+Java_com_google_android_filament_MaterialInstance_nIsColorWriteEnabled(JNIEnv* env, jclass clazz,
+        jlong nativeMaterialInstance) {
+    // TODO: implement nIsColorWriteEnabled()
+    MaterialInstance* instance = (MaterialInstance*)nativeMaterialInstance;
+    return instance->isColorWriteEnabled();
+}
+
+extern "C"
+JNIEXPORT jboolean JNICALL
+Java_com_google_android_filament_MaterialInstance_nIsDepthWriteEnabled(JNIEnv* env, jclass clazz,
+        jlong nativeMaterialInstance) {
+    // TODO: implement nIsDepthWriteEnabled()
+    MaterialInstance* instance = (MaterialInstance*)nativeMaterialInstance;
+    return instance->isDepthWriteEnabled();
+}
+
+extern "C"
+JNIEXPORT jboolean JNICALL
+Java_com_google_android_filament_MaterialInstance_nIsStencilWriteEnabled(JNIEnv* env, jclass clazz,
+        jlong nativeMaterialInstance) {
+    // TODO: implement nIsStencilWriteEnabled()
+    MaterialInstance* instance = (MaterialInstance*)nativeMaterialInstance;
+    return instance->isStencilWriteEnabled();
+}
+
+extern "C"
+JNIEXPORT jboolean JNICALL
+Java_com_google_android_filament_MaterialInstance_nIsDepthCullingEnabled(JNIEnv* env, jclass clazz,
+        jlong nativeMaterialInstance) {
+    // TODO: implement nIsDepthCullingEnabled()
+    MaterialInstance* instance = (MaterialInstance*)nativeMaterialInstance;
+    return instance->isDepthCullingEnabled();
+}

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

@@ -24,6 +24,7 @@ import com.google.android.filament.proguard.UsedByNative;
 
 @UsedByNative("AssetLoader.cpp")
 public class MaterialInstance {
+    private static final Material.CullingMode[] sCullingModeValues = Material.CullingMode.values();
     private Material mMaterial;
     private String mName;
     private long mNativeObject;
@@ -473,6 +474,14 @@ public class MaterialInstance {
         nSetMaskThreshold(getNativeObject(), threshold);
     }
 
+    /**
+     * Gets the minimum alpha value a fragment must have to not be discarded when the blend
+     * mode is MASKED
+     */
+    public float getMaskThreshold() {
+        return nGetMaskThreshold(getNativeObject());
+    }
+
     /**
      * Sets the screen space variance of the filter kernel used when applying specular
      * anti-aliasing. The default value is set to 0.15. The specified value should be between
@@ -486,6 +495,14 @@ public class MaterialInstance {
         nSetSpecularAntiAliasingVariance(getNativeObject(), variance);
     }
 
+    /**
+     * Gets the screen space variance of the filter kernel used when applying specular
+     * anti-aliasing.
+     */
+    public float getSpecularAntiAliasingVariance() {
+        return nGetSpecularAntiAliasingVariance(getNativeObject());
+    }
+
     /**
      * Sets the clamping threshold used to suppress estimation errors when applying specular
      * anti-aliasing. The default value is set to 0.2. The specified value should be between 0
@@ -499,6 +516,14 @@ public class MaterialInstance {
         nSetSpecularAntiAliasingThreshold(getNativeObject(), threshold);
     }
 
+    /**
+     * Gets the clamping threshold used to suppress estimation errors when applying specular
+     * anti-aliasing.
+     */
+    public float getSpecularAntiAliasingThreshold() {
+        return nGetSpecularAntiAliasingThreshold(getNativeObject());
+    }
+
     /**
      * Enables or disables double-sided lighting if the parent Material has double-sided capability,
      * otherwise prints a warning. If double-sided lighting is enabled, backface culling is
@@ -512,6 +537,14 @@ public class MaterialInstance {
         nSetDoubleSided(getNativeObject(), doubleSided);
     }
 
+    /**
+     * Returns whether double-sided lighting is enabled when the parent Material has double-sided
+     * capability.
+     */
+    public boolean isDoubleSided() {
+        return nIsDoubleSided(getNativeObject());
+    }
+
     /**
      * Overrides the default triangle culling state that was set on the material.
      *
@@ -519,10 +552,18 @@ public class MaterialInstance {
      * <a href="https://google.github.io/filament/Materials.html#materialdefinitions/materialblock/rasterization:culling">
      * Rasterization: culling</a>
      */
-    public void setCullingMode(Material.CullingMode mode) {
+    public void setCullingMode(@NonNull Material.CullingMode mode) {
         nSetCullingMode(getNativeObject(), mode.ordinal());
     }
 
+    /**
+     * Returns the face culling mode.
+     */
+    @NonNull
+    public Material.CullingMode getCullingMode() {
+        return sCullingModeValues[nGetCullingMode(getNativeObject())];
+    }
+
     /**
      * Overrides the default color-buffer write state that was set on the material.
      *
@@ -534,6 +575,13 @@ public class MaterialInstance {
         nSetColorWrite(getNativeObject(), enable);
     }
 
+    /**
+     * Returns whether color write is enabled.
+     */
+    public boolean isColorWriteEnabled() {
+        return nIsColorWriteEnabled(getNativeObject());
+    }
+
     /**
      * Overrides the default depth-buffer write state that was set on the material.
      *
@@ -545,10 +593,27 @@ public class MaterialInstance {
         nSetDepthWrite(getNativeObject(), enable);
     }
 
+    /**
+     * Returns whether depth write is enabled.
+     */
+    public boolean isDepthWriteEnabled() {
+        return nIsDepthWriteEnabled(getNativeObject());
+    }
+
+    /**
+     * Enables or Disable stencil writes
+     */
     public void setStencilWrite(boolean enable) {
         nSetStencilWrite(getNativeObject(), enable);
     }
 
+    /**
+     * Returns whether stencil write is enabled.
+     */
+    public boolean isStencilWriteEnabled() {
+        return nIsStencilWriteEnabled(getNativeObject());
+    }
+
     /**
      * Overrides the default depth testing state that was set on the material.
      *
@@ -560,6 +625,13 @@ public class MaterialInstance {
         nSetDepthCulling(getNativeObject(), enable);
     }
 
+    /**
+     * Returns whether depth culling is enabled.
+     */
+    public boolean isDepthCullingEnabled() {
+        return nIsDepthCullingEnabled(getNativeObject());
+    }
+
     /**
      * Sets the stencil comparison function (default is {@link TextureSampler.CompareFunction#ALWAYS}).
      *
@@ -856,4 +928,15 @@ public class MaterialInstance {
     private static native long nGetMaterial(long nativeMaterialInstance);
 
     private static native long nDuplicate(long otherNativeMaterialInstance, String name);
+
+
+    private static native float nGetMaskThreshold(long nativeMaterialInstance);
+    private static native float nGetSpecularAntiAliasingVariance(long nativeMaterialInstance);
+    private static native float nGetSpecularAntiAliasingThreshold(long nativeMaterialInstance);
+    private static native boolean nIsDoubleSided(long nativeMaterialInstance);
+    private static native int nGetCullingMode(long nativeMaterialInstance);
+    private static native boolean nIsColorWriteEnabled(long nativeMaterialInstance);
+    private static native boolean nIsDepthWriteEnabled(long nativeMaterialInstance);
+    private static native boolean nIsStencilWriteEnabled(long nativeMaterialInstance);
+    private static native boolean nIsDepthCullingEnabled(long nativeMaterialInstance);
 }

android/gltfio-android/CMakeLists.txt

@@ -11,6 +11,10 @@ add_library(dracodec STATIC IMPORTED)
 set_target_properties(dracodec PROPERTIES IMPORTED_LOCATION
         ${FILAMENT_DIR}/lib/${ANDROID_ABI}/libdracodec.a)
 
+add_library(meshoptimizer STATIC IMPORTED)
+set_target_properties(meshoptimizer PROPERTIES IMPORTED_LOCATION
+        ${FILAMENT_DIR}/lib/${ANDROID_ABI}/libmeshoptimizer.a)
+
 add_library(ktxreader STATIC IMPORTED)
 set_target_properties(ktxreader PROPERTIES IMPORTED_LOCATION
         ${FILAMENT_DIR}/lib/${ANDROID_ABI}/libktxreader.a)
@@ -99,6 +103,7 @@ set(GLTFIO_INCLUDE_DIRS
         ../../libs/gltfio/include
         ../../third_party/basisu/zstd
         ../../third_party/cgltf
+        ../../third_party/meshoptimizer/src
         ../../third_party/robin-map
         ../../third_party/hat-trie
         ../../third_party/stb
@@ -111,7 +116,7 @@ target_include_directories(gltfio-jni PRIVATE ${GLTFIO_INCLUDE_DIRS})
 set_target_properties(gltfio-jni PROPERTIES LINK_DEPENDS ${CMAKE_CURRENT_SOURCE_DIR}/libgltfio-jni.symbols)
 set_target_properties(gltfio-jni PROPERTIES LINK_DEPENDS ${CMAKE_CURRENT_SOURCE_DIR}/libgltfio-jni.map)
 target_link_libraries(gltfio-jni filament-jni utils uberzlib log stb ktxreader basis_transcoder zstd uberarchive)
-target_link_libraries(gltfio-jni dracodec)
+target_link_libraries(gltfio-jni dracodec meshoptimizer)
 target_compile_definitions(gltfio-jni PUBLIC GLTFIO_DRACO_SUPPORTED=1)
 target_include_directories(gltfio-jni PRIVATE ${DRACO_DIR}/src)
 target_include_directories(gltfio-jni PRIVATE ${DRACO_DIR}/tnt)

android/gradle.properties

@@ -1,5 +1,5 @@
 GROUP=com.google.android.filament
-VERSION_NAME=1.28.1
+VERSION_NAME=1.28.2
 
 POM_DESCRIPTION=Real-time physically based rendering engine for Android.
 

build/common/check-metal-shaders.sh

@@ -0,0 +1,99 @@
+#!/usr/bin/env bash
+
+set -e
+
+function print_help {
+    local self_name=$(basename "$0")
+    echo "This tool verifies that every Metal shader in the given Filament material file(s) successfully compiles."
+    echo "Requires the matinfo tool and xcrun to be on the system PATH."
+    echo ""
+    echo "Usage:"
+    echo "    $self_name [options] [<material> ...]"
+    echo ""
+    echo "Options:"
+    echo "    -h"
+    echo "        Print this help message."
+    echo "    -w"
+    echo "        Pass the -w flag to the Metal compiler, to disable warnings."
+    echo ""
+}
+
+COMP_FLAGS=""
+while getopts ":hw" opt; do
+    case ${opt} in
+        h)
+            print_help
+            exit 1
+            ;;
+        w)
+            COMP_FLAGS="${COMP_FLAGS} -w"
+            ;;
+    esac
+done
+
+shift $((OPTIND - 1))
+
+if [[ "$#" == "0" ]]; then
+    print_help
+    exit 1
+fi
+
+# Make sure matinfo and xcrun are available.
+if [[ ! $(command -v matinfo) ]]; then
+    echo "Error: matinfo not on PATH."
+    exit 1
+fi
+if [[ ! $(command -v xcrun) ]]; then
+    echo "Error: xcrun not on PATH."
+    exit 1
+fi
+
+function check_material {
+    local material="$1"
+
+    tmpdir="$(mktemp -d /tmp/check_metal_shaders.XXXXX)"
+
+    # First check that the material is valid.
+    if [[ ! $(matinfo "${material}") ]]; then
+        echo "Invalid material file: ${material}"
+        exit 1
+    fi
+
+    echo "Checking that Metal shaders compile: ${material}"
+
+    set +e
+    local i=0
+    while true; do
+        # The file must have a .metal extension.
+        metalFile="${tmpdir}/${i}.metal"
+
+        # Extract shader i. matinfo will return a non-zero exit code if the shader doesn't exist,
+        # which means we're finished with this material.
+        matinfo --print-metal=${i} "${material}" > "${metalFile}" 2> /dev/null
+
+        if [[ "$?" -ne 0 ]]; then
+            break
+        fi
+
+        echo "Testing Metal shader ${i}"
+
+        # Attempt to compile the Metal shader.
+        xcrun -sdk macosx metal ${COMP_FLAGS} -c "${metalFile}" -o /dev/null
+
+        if [[ "$?" -ne 0 ]]; then
+            echo "Error compiling Metal shader: ${metalFile}"
+            exit 1
+        fi
+
+        ((i++))
+    done
+    set -e
+
+    rm -r "${tmpdir}"
+}
+
+for material in "$@"
+do
+    check_material "${material}"
+done
+

docs/Materials.html

@@ -1432,12 +1432,13 @@ non-shader data.
 
 </p><dl><table><tbody><tr valign="top"><td><dt>Type</dt></td><td><dd><p>     <code>number</code>
 
-</p></dd></td></tr><tr valign="top"><td><dt>Value</dt></td><td><dd><p>      An integer value, either 1 or 2. Defaults to 1.
+</p></dd></td></tr><tr valign="top"><td><dt>Value</dt></td><td><dd><p>      An integer value, either 1, 2 or 3. Defaults to 1.
 
 </p></dd></td></tr></tbody></table></dl><div class="table">
 <table class="table"><tbody><tr><th style="text-align:left"> Feature Level </th><th style="text-align:left"> Guaranteed features </th></tr>
 <tr><td style="text-align:left"> 1 </td><td style="text-align:left"> 9 textures per material </td></tr>
-<tr><td style="text-align:left"> 2 </td><td style="text-align:left"> 12 textures per material, cubemap arrays, ESSL 3.10 </td></tr>
+<tr><td style="text-align:left"> 2 </td><td style="text-align:left"> 9 textures per material, cubemap arrays, ESSL 3.10 </td></tr>
+<tr><td style="text-align:left"> 3 </td><td style="text-align:left"> 12 textures per material, cubemap arrays, ESSL 3.10 </td></tr>
 </tbody></table><center><div class="tablecaption"><a class="target" name="table_featurelevels">&nbsp;</a><b style="font-style:normal;">Table&nbsp;13:</b> Feature levels</div></center></div>
 
 <p></p><p>
@@ -2036,23 +2037,21 @@ and sorting issues are minimized or eliminated</span></center></div></center>
 
 </p></dd><dt>Value</dt><dd><p>      <code>true</code> or <code>false</code>. Defaults to <code>false</code>.
 
-</p></dd><dt>Description</dt><dd><p>      Only available in the <code>unlit</code> shading model. If this property is enabled, the final color
-      computed by the material is multiplied by the shadowing factor (or visibility). This allows to
-      create transparent shadow-receiving objects (for instance an invisible ground plane in AR).
-      This is only supported with shadows from directional lights.
+</p></dd><dt>Description</dt><dd><p>      Enables two-sided rendering and its capability to be toggled at run time. When set to <code>true</code>,
+      <code>culling</code> is automatically set to <code>none</code>; if the triangle is back-facing, the triangle's
+      normal is flipped to become front-facing. When explicitly set to <code>false</code>, this allows the
+      double-sidedness to be toggled at run time.
 
 </p></dd></dl><p></p><pre class="listing tilde"><code><span class="line">material {</span>
-<span class="line">    name : <span class="hljs-string">"Invisible shadow plane"</span>,</span>
-<span class="line">    shadingModel : unlit,</span>
-<span class="line">    shadowMultiplier : <span class="hljs-literal">true</span>,</span>
-<span class="line">    blending : transparent</span>
+<span class="line">    name : <span class="hljs-string">"Double sided material"</span>,</span>
+<span class="line">    shadingModel : lit,</span>
+<span class="line">    doubleSided : <span class="hljs-literal">true</span></span>
 <span class="line">}</span>
 <span class="line"></span>
 <span class="line">fragment {</span>
 <span class="line">    void material(inout MaterialInputs material) {</span>
 <span class="line">        prepareMaterial(material);</span>
-<span class="line">        <span class="hljs-comment">// baseColor defines the color and opacity of the final shadow</span></span>
-<span class="line">        material.baseColor = vec4(0.0, 0.0, 0.0, 0.7);</span>
+<span class="line">        material.baseColor = materialParams.albedo;</span>
 <span class="line">    }</span>
 <span class="line">}</span></code></pre>
 <a class="target" name="lighting:reflections">&nbsp;</a><a class="target" name="materialdefinitions/materialblock/lighting:reflections">&nbsp;</a><a class="target" name="toc4.2.25">&nbsp;</a><h3>Lighting: reflections</h3>

docs/Materials.md.html

@@ -1582,24 +1582,22 @@ Value
 :     `true` or `false`. Defaults to `false`.
 
 Description
-:     Only available in the `unlit` shading model. If this property is enabled, the final color
-      computed by the material is multiplied by the shadowing factor (or visibility). This allows to
-      create transparent shadow-receiving objects (for instance an invisible ground plane in AR).
-      This is only supported with shadows from directional lights.
+:     Enables two-sided rendering and its capability to be toggled at run time. When set to `true`,
+      `culling` is automatically set to `none`; if the triangle is back-facing, the triangle's
+      normal is flipped to become front-facing. When explicitly set to `false`, this allows the
+      double-sidedness to be toggled at run time.
 
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ JSON
 material {
-    name : "Invisible shadow plane",
-    shadingModel : unlit,
-    shadowMultiplier : true,
-    blending : transparent
+    name : "Double sided material",
+    shadingModel : lit,
+    doubleSided : true
 }
 
 fragment {
     void material(inout MaterialInputs material) {
         prepareMaterial(material);
-        // baseColor defines the color and opacity of the final shadow
-        material.baseColor = vec4(0.0, 0.0, 0.0, 0.7);
+        material.baseColor = materialParams.albedo;
     }
 }
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

filament/CMakeLists.txt

@@ -48,6 +48,7 @@ set(PUBLIC_HDRS
 )
 
 set(SRCS
+        src/AtlasAllocator.cpp
         src/Box.cpp
         src/BufferObject.cpp
         src/Camera.cpp
@@ -72,6 +73,7 @@ set(SRCS
         src/MaterialParser.cpp
         src/MorphTargetBuffer.cpp
         src/PerViewUniforms.cpp
+        src/PerShadowMapUniforms.cpp
         src/PostProcessManager.cpp
         src/RenderPass.cpp
         src/RenderPrimitive.cpp
@@ -145,6 +147,7 @@ set(PRIVATE_HDRS
         src/Intersections.h
         src/MaterialParser.h
         src/PerViewUniforms.h
+        src/PerShadowMapUniforms.h
         src/PIDController.h
         src/PostProcessManager.h
         src/RendererUtils.h

filament/backend/include/backend/DriverEnums.h

@@ -220,8 +220,7 @@ enum class PrimitiveType : uint8_t {
     LINES          = 1,    //!< lines
     LINE_STRIP     = 3,    //!< line strip
     TRIANGLES      = 4,    //!< triangles
-    TRIANGLE_STRIP = 5,    //!< triangle strip
-    NONE           = 0xFF
+    TRIANGLE_STRIP = 5     //!< triangle strip
 };
 
 /**

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

@@ -158,6 +158,9 @@ DECL_DRIVER_API_0(flush)
 // flush and wait for the effects to be done
 DECL_DRIVER_API_0(finish)
 
+// reset state tracking, if the driver does any state tracking (e.g. GL)
+DECL_DRIVER_API_0(resetState)
+
 /*
  * Creating driver objects
  * -----------------------

filament/backend/src/Platform.cpp

@@ -42,7 +42,9 @@
         #include "vulkan/PlatformVkCocoa.h"
     #endif
 #elif defined(__linux__)
-    #if defined(FILAMENT_SUPPORTS_WAYLAND)
+    #if defined(FILAMENT_SUPPORTS_GGP)
+        #include "vulkan/PlatformVkLinuxGGP.h"
+    #elif defined(FILAMENT_SUPPORTS_WAYLAND)
         #if defined (FILAMENT_DRIVER_SUPPORTS_VULKAN)
             #include "vulkan/PlatformVkLinuxWayland.h"
         #endif
@@ -120,7 +122,9 @@ DefaultPlatform* DefaultPlatform::create(Backend* backend) noexcept {
             #elif defined(IOS)
                 return new PlatformVkCocoaTouch();
             #elif defined(__linux__)
-                #if defined(FILAMENT_SUPPORTS_WAYLAND)
+                #if defined(FILAMENT_SUPPORTS_GGP)
+                    return new PlatformVkLinuxGGP();
+                #elif defined(FILAMENT_SUPPORTS_WAYLAND)
                     return new PlatformVkLinuxWayland();
                 #elif defined(FILAMENT_SUPPORTS_X11)
                     return new PlatformVkLinuxX11();

filament/backend/src/metal/MetalBlitter.mm

@@ -91,12 +91,14 @@ blitterFrag(VertexOut in [[stage_in]],
 {
     FragmentOut out = {};
 
+#if defined(BLIT_COLOR) || defined(BLIT_DEPTH)
     // These coordinates match the Vulkan vkCmdBlitImage spec:
     // https://www.khronos.org/registry/vulkan/specs/1.2-extensions/man/html/vkCmdBlitImage.html
     float2 uvbase = in.position.xy; // unnormalized coordinates at center of texel: (1.5, 2.5, etc)
     float2 uvoffset = uvbase - args->dstOffset;
     float2 uvscaled = uvoffset * args->scale;
     float2 uv = uvscaled + args->srcOffset;
+#endif
 
 #ifdef BLIT_COLOR
 #ifdef MSAA_COLOR_SOURCE

filament/backend/src/metal/MetalBuffer.h

@@ -34,7 +34,8 @@ namespace filament::backend {
 class MetalBuffer {
 public:
 
-    MetalBuffer(MetalContext& context, BufferUsage usage, size_t size, bool forceGpuBuffer = false);
+    MetalBuffer(MetalContext& context, BufferObjectBinding bindingType, BufferUsage usage,
+         size_t size, bool forceGpuBuffer = false);
     ~MetalBuffer();
 
     MetalBuffer(const MetalBuffer& rhs) = delete;
@@ -61,18 +62,21 @@ public:
 
     void* getCpuBuffer() const noexcept { return mCpuBuffer; }
 
-    enum Stage {
-        VERTEX = 1,
-        FRAGMENT = 2
+    enum Stage : uint8_t {
+        VERTEX      = 1u << 0u,
+        FRAGMENT    = 1u << 1u,
+        COMPUTE     = 1u << 2u
     };
 
     /**
      * Bind multiple buffers to pipeline stages.
      *
-     * bindBuffers binds an array of buffers to the given stage(s) of a MTLRenderCommandEncoder's
-     * pipeline.
+     * bindBuffers binds an array of buffers to the given stage(s) of a MTLCommandEncoders's
+     * pipeline. The encoder must be either a MTLRenderCommandEncoder or a MTLComputeCommandEncoder.
+     * For MTLRenderCommandEncoders, only the VERTEX and FRAGMENT stages may be specified.
+     * For MTLComputeCommandEncoders, only the COMPUTE stage may be specified.
      */
-    static void bindBuffers(id<MTLCommandBuffer> cmdBuffer, id<MTLRenderCommandEncoder> encoder,
+    static void bindBuffers(id<MTLCommandBuffer> cmdBuffer, id<MTLCommandEncoder> encoder,
             size_t bufferStart, uint8_t stages, MetalBuffer* const* buffers, size_t const* offsets,
             size_t count);
 

filament/backend/src/metal/MetalBuffer.mm

@@ -21,15 +21,17 @@
 namespace filament {
 namespace backend {
 
-MetalBuffer::MetalBuffer(MetalContext& context, BufferUsage usage, size_t size, bool forceGpuBuffer)
-        : mBufferSize(size), mContext(context) {
+MetalBuffer::MetalBuffer(MetalContext& context, BufferObjectBinding bindingType, BufferUsage usage,
+        size_t size, bool forceGpuBuffer) : mBufferSize(size), mContext(context) {
     // If the buffer is less than 4K in size and is updated frequently, we don't use an explicit
     // buffer. Instead, we use immediate command encoder methods like setVertexBytes:length:atIndex:.
-   if (size <= 4 * 1024 && usage == BufferUsage::DYNAMIC && !forceGpuBuffer) {
-       mBuffer = nil;
-       mCpuBuffer = malloc(size);
-       return;
-   }
+    // This won't work for SSBOs, since they are read/write.
+    if (size <= 4 * 1024 && bindingType != BufferObjectBinding::SHADER_STORAGE &&
+            usage == BufferUsage::DYNAMIC && !forceGpuBuffer) {
+        mBuffer = nil;
+        mCpuBuffer = malloc(size);
+        return;
+    }
 
     // Otherwise, we allocate a private GPU buffer.
     mBuffer = [context.device newBufferWithLength:size options:MTLResourceStorageModePrivate];
@@ -93,9 +95,24 @@ id<MTLBuffer> MetalBuffer::getGpuBufferForDraw(id<MTLCommandBuffer> cmdBuffer) n
     return mBuffer;
 }
 
-void MetalBuffer::bindBuffers(id<MTLCommandBuffer> cmdBuffer, id<MTLRenderCommandEncoder> encoder,
+void MetalBuffer::bindBuffers(id<MTLCommandBuffer> cmdBuffer, id<MTLCommandEncoder> encoder,
         size_t bufferStart, uint8_t stages, MetalBuffer* const* buffers, size_t const* offsets,
         size_t count) {
+    // Ensure we were given the correct type of encoder:
+    // either a MTLRenderCommandEncoder or a MTLComputeCommandEncoder.
+    if (stages & MetalBuffer::Stage::VERTEX || stages & MetalBuffer::Stage::FRAGMENT) {
+        assert_invariant([encoder respondsToSelector:@selector(setVertexBuffers:offsets:withRange:)]);
+        assert_invariant([encoder respondsToSelector:@selector(setFragmentBuffers:offsets:withRange:)]);
+        assert_invariant([encoder respondsToSelector:@selector(setVertexBytes:length:atIndex:)]);
+        assert_invariant([encoder respondsToSelector:@selector(setFragmentBytes:length:atIndex:)]);
+        assert_invariant(!(stages & MetalBuffer::Stage::COMPUTE));
+    }
+    if (stages & MetalBuffer::Stage::COMPUTE) {
+        assert_invariant([encoder respondsToSelector:@selector(setBuffers:offsets:withRange:)]);
+        assert_invariant([encoder respondsToSelector:@selector(setBytes:length:atIndex:)]);
+        assert_invariant(!(stages & (MetalBuffer::Stage::FRAGMENT | MetalBuffer::Stage::VERTEX)));
+    }
+
     const NSRange bufferRange = NSMakeRange(bufferStart, count);
 
     constexpr size_t MAX_BUFFERS = 16;
@@ -120,14 +137,19 @@ void MetalBuffer::bindBuffers(id<MTLCommandBuffer> cmdBuffer, id<MTLRenderComman
     }
 
     if (stages & Stage::VERTEX) {
-        [encoder setVertexBuffers:metalBuffers.data()
-                          offsets:metalOffsets.data()
-                        withRange:bufferRange];
+        [(id<MTLRenderCommandEncoder>) encoder setVertexBuffers:metalBuffers.data()
+                                                        offsets:metalOffsets.data()
+                                                      withRange:bufferRange];
     }
     if (stages & Stage::FRAGMENT) {
-        [encoder setFragmentBuffers:metalBuffers.data()
-                            offsets:metalOffsets.data()
-                          withRange:bufferRange];
+        [(id<MTLRenderCommandEncoder>) encoder setFragmentBuffers:metalBuffers.data()
+                                                          offsets:metalOffsets.data()
+                                                        withRange:bufferRange];
+    }
+    if (stages & Stage::COMPUTE) {
+        [(id<MTLComputeCommandEncoder>) encoder setBuffers:metalBuffers.data()
+                                                   offsets:metalOffsets.data()
+                                                 withRange:bufferRange];
     }
 
     for (size_t b = 0; b < count; b++) {
@@ -146,14 +168,22 @@ void MetalBuffer::bindBuffers(id<MTLCommandBuffer> cmdBuffer, id<MTLRenderComman
         auto* bytes = static_cast<const uint8_t*>(cpuBuffer);
 
         if (stages & Stage::VERTEX) {
-            [encoder setVertexBytes:(bytes + offset)
-                             length:(buffer->getSize() - offset)
-                            atIndex:bufferIndex];
+            [(id<MTLRenderCommandEncoder>) encoder setVertexBytes:(bytes + offset)
+                                                           length:(buffer->getSize() - offset)
+                                                          atIndex:bufferIndex];
         }
         if (stages & Stage::FRAGMENT) {
-            [encoder setFragmentBytes:(bytes + offset)
-                               length:(buffer->getSize() - offset)
-                              atIndex:bufferIndex];
+            [(id<MTLRenderCommandEncoder>) encoder setFragmentBytes:(bytes + offset)
+                                                             length:(buffer->getSize() - offset)
+                                                            atIndex:bufferIndex];
+        }
+        if (stages & Stage::COMPUTE) {
+            // TODO: using setBytes means the data is read-only, which currently isn't enforced.
+            // In practice this won't be an issue since MetalBuffer ensures all SSBOs are realized
+            // through actual id<MTLBuffer> allocations.
+            [(id<MTLComputeCommandEncoder>) encoder setBytes:(bytes + offset)
+                                                      length:(buffer->getSize() - offset)
+                                                     atIndex:bufferIndex];
         }
     }
 }

filament/backend/src/metal/MetalContext.h

@@ -64,6 +64,7 @@ struct MetalContext {
 
     // Supported features.
     bool supportsTextureSwizzling = false;
+    bool supportsAutoDepthResolve = false;
     bool supportsMemorylessRenderTargets = false;
     uint8_t maxColorRenderTargets = 4;
     struct {
@@ -84,7 +85,8 @@ struct MetalContext {
     // State trackers.
     PipelineStateTracker pipelineState;
     DepthStencilStateTracker depthStencilState;
-    UniformBufferState uniformState[Program::UNIFORM_BINDING_COUNT];
+    std::array<BufferState, Program::UNIFORM_BINDING_COUNT> uniformState;
+    std::array<BufferState, MAX_SSBO_COUNT> ssboState;
     CullModeStateTracker cullModeState;
     WindingStateTracker windingState;
 

filament/backend/src/metal/MetalDriver.h

@@ -39,7 +39,7 @@ class MetalTexture;
 struct MetalUniformBuffer;
 struct MetalContext;
 struct MetalProgram;
-struct UniformBufferState;
+struct BufferState;
 
 #ifndef FILAMENT_METAL_HANDLE_ARENA_SIZE_IN_MB
 #define FILAMENT_METAL_HANDLE_ARENA_SIZE_IN_MB 8
@@ -129,8 +129,8 @@ private:
             uint32_t offset, uint32_t minIndex, uint32_t maxIndex, uint32_t count);
 
     void finalizeSamplerGroup(MetalSamplerGroup* sg);
-    void enumerateBoundUniformBuffers(const std::function<void(const UniformBufferState&,
-            MetalBuffer*, uint32_t)>& f);
+    void enumerateBoundBuffers(BufferObjectBinding bindingType,
+            const std::function<void(const BufferState&, MetalBuffer*, uint32_t)>& f);
 
 };
 

filament/backend/src/metal/MetalDriver.mm

@@ -92,6 +92,13 @@ MetalDriver::MetalDriver(MetalPlatform* platform, const Platform::DriverConfig&
             mContext->highestSupportedGpuFamily.mac   >= 2;     // newer macOS GPUs
     }
 
+    // In order to support resolve store action on depth attachment, the GPU needs to support it.
+    // Note that support for depth resolve implies support for stencil resolve using .sample0 resolve filter.
+    // (Other resolve filters are supported starting .apple5 and .mac2 families).
+    mContext->supportsAutoDepthResolve =
+        mContext->highestSupportedGpuFamily.apple >= 3 ||
+        mContext->highestSupportedGpuFamily.mac   >= 2;
+
     // In order to support memoryless render targets, an Apple GPU is needed.
     // Available starting macOS 11.0, the first version to run on Apple GPUs.
     // On iOS, it's available on all OS versions.
@@ -243,7 +250,7 @@ void MetalDriver::createIndexBufferR(Handle<HwIndexBuffer> ibh, ElementType elem
 
 void MetalDriver::createBufferObjectR(Handle<HwBufferObject> boh, uint32_t byteCount,
         BufferObjectBinding bindingType, BufferUsage usage) {
-    construct_handle<MetalBufferObject>(boh, *mContext, usage, byteCount);
+    construct_handle<MetalBufferObject>(boh, *mContext, bindingType, usage, byteCount);
 }
 
 void MetalDriver::createTextureR(Handle<HwTexture> th, SamplerType target, uint8_t levels,
@@ -481,10 +488,20 @@ void MetalDriver::destroyBufferObject(Handle<HwBufferObject> boh) {
         return;
     }
     auto* bo = handle_cast<MetalBufferObject>(boh);
-    // Unbind this buffer object from any uniform slots it's still bound to.
+    // Unbind this buffer object from any uniform / SSBO slots it's still bound to.
     bo->boundUniformBuffers.forEachSetBit([this](size_t index) {
-        mContext->uniformState[index].buffer = nullptr;
-        mContext->uniformState[index].bound = false;
+        mContext->uniformState[index] = BufferState {
+                .buffer = nullptr,
+                .offset = 0,
+                .bound = false
+        };
+    });
+    bo->boundSsbos.forEachSetBit([this](size_t index) {
+        mContext->ssboState[index] = BufferState {
+                .buffer = nullptr,
+                .offset = 0,
+                .bound = false
+        };
     });
     destruct_handle<MetalBufferObject>(boh);
 }
@@ -686,7 +703,7 @@ bool MetalDriver::isFrameTimeSupported() {
 }
 
 bool MetalDriver::isAutoDepthResolveSupported() {
-    return true;
+    return mContext->supportsAutoDepthResolve;
 }
 
 bool MetalDriver::isWorkaroundNeeded(Workaround workaround) {
@@ -1051,7 +1068,7 @@ void MetalDriver::bindUniformBuffer(uint32_t index, Handle<HwBufferObject> boh)
         currentBo->boundUniformBuffers.unset(index);
     }
     bo->boundUniformBuffers.set(index);
-    mContext->uniformState[index] = UniformBufferState{
+    mContext->uniformState[index] = BufferState{
             .buffer = bo,
             .offset = 0,
             .bound = true
@@ -1064,24 +1081,81 @@ void MetalDriver::bindBufferRange(BufferObjectBinding bindingType, uint32_t inde
     assert_invariant(bindingType == BufferObjectBinding::SHADER_STORAGE ||
                      bindingType == BufferObjectBinding::UNIFORM);
 
-    // TODO: implement BufferObjectBinding::SHADER_STORAGE case
-
-    assert_invariant(index < Program::UNIFORM_BINDING_COUNT);
     auto* bo = handle_cast<MetalBufferObject>(boh);
-    auto* currentBo = mContext->uniformState[index].buffer;
-    if (currentBo) {
-        currentBo->boundUniformBuffers.unset(index);
+
+    switch (bindingType) {
+        default:
+        case BufferObjectBinding::UNIFORM: {
+            assert_invariant(index < Program::UNIFORM_BINDING_COUNT);
+            auto* currentBo = mContext->uniformState[index].buffer;
+            if (currentBo) {
+                currentBo->boundUniformBuffers.unset(index);
+            }
+            bo->boundUniformBuffers.set(index);
+            mContext->uniformState[index] = BufferState {
+                    .buffer = bo,
+                    .offset = offset,
+                    .bound = true
+            };
+
+            break;
+        }
+
+        case BufferObjectBinding::SHADER_STORAGE: {
+            assert_invariant(index < MAX_SSBO_COUNT);
+            auto* currentBo = mContext->ssboState[index].buffer;
+            if (currentBo) {
+                currentBo->boundSsbos.unset(index);
+            }
+            bo->boundSsbos.set(index);
+            mContext->ssboState[index] = BufferState {
+                    .buffer = bo,
+                    .offset = offset,
+                    .bound = true
+            };
+
+            break;
+        }
     }
-    bo->boundUniformBuffers.set(index);
-    mContext->uniformState[index] = UniformBufferState{
-            .buffer = bo,
-            .offset = offset,
-            .bound = true
-    };
 }
 
 void MetalDriver::unbindBuffer(BufferObjectBinding bindingType, uint32_t index) {
-    // TODO: implement unbindBuffer()
+
+    assert_invariant(bindingType == BufferObjectBinding::SHADER_STORAGE ||
+                     bindingType == BufferObjectBinding::UNIFORM);
+
+    switch (bindingType) {
+        default:
+        case BufferObjectBinding::UNIFORM: {
+            assert_invariant(index < Program::UNIFORM_BINDING_COUNT);
+            auto* currentBo = mContext->uniformState[index].buffer;
+            if (currentBo) {
+                currentBo->boundUniformBuffers.unset(index);
+            }
+            mContext->uniformState[index] = BufferState {
+                    .buffer = nullptr,
+                    .offset = 0,
+                    .bound = false
+            };
+
+            break;
+        }
+
+        case BufferObjectBinding::SHADER_STORAGE: {
+            assert_invariant(index < MAX_SSBO_COUNT);
+            auto* currentBo = mContext->ssboState[index].buffer;
+            if (currentBo) {
+                currentBo->boundSsbos.unset(index);
+            }
+            mContext->ssboState[index] = BufferState {
+                    .buffer = nullptr,
+                    .offset = 0,
+                    .bound = false
+            };
+
+            break;
+        }
+    }
 }
 
 void MetalDriver::bindSamplers(uint32_t index, Handle<HwSamplerGroup> sbh) {
@@ -1552,8 +1626,9 @@ void MetalDriver::draw(PipelineState ps, Handle<HwRenderPrimitive> rph, uint32_t
     MetalBuffer* uniformsToBind[Program::UNIFORM_BINDING_COUNT] = { nil };
     NSUInteger offsets[Program::UNIFORM_BINDING_COUNT] = { 0 };
 
-    enumerateBoundUniformBuffers([&uniformsToBind, &offsets](const UniformBufferState& state,
-            MetalBuffer* buffer, uint32_t index) {
+    enumerateBoundBuffers(BufferObjectBinding::UNIFORM,
+            [&uniformsToBind, &offsets](const BufferState& state, MetalBuffer* buffer,
+                    uint32_t index) {
         uniformsToBind[index] = buffer;
         offsets[index] = state.offset;
     });
@@ -1635,7 +1710,67 @@ void MetalDriver::draw(PipelineState ps, Handle<HwRenderPrimitive> rph, uint32_t
 }
 
 void MetalDriver::dispatchCompute(Handle<HwProgram> program, math::uint3 workGroupCount) {
-    // FIXME: implement me
+    ASSERT_PRECONDITION(!isInRenderPass(mContext),
+            "dispatchCompute must be called outside of a render pass.");
+
+    auto mtlProgram = handle_cast<MetalProgram>(program);
+
+    // If the material debugger is enabled, avoid fatal (or cascading) errors and that can occur
+    // during the draw call when the program is invalid. The shader compile error has already been
+    // dumped to the console at this point, so it's fine to simply return early.
+    if (FILAMENT_ENABLE_MATDBG && UTILS_UNLIKELY(!mtlProgram->isValid)) {
+        return;
+    }
+
+    assert_invariant(mtlProgram->isValid && mtlProgram->computeFunction);
+
+    id<MTLComputeCommandEncoder> computeEncoder =
+            [getPendingCommandBuffer(mContext) computeCommandEncoder];
+
+    NSError* error = nil;
+    id<MTLComputePipelineState> computePipelineState =
+            [mContext->device newComputePipelineStateWithFunction:mtlProgram->computeFunction
+                                                            error:&error];
+    if (error) {
+        auto description = [error.localizedDescription cStringUsingEncoding:NSUTF8StringEncoding];
+        utils::slog.e << description << utils::io::endl;
+    }
+    assert_invariant(!error);
+
+    // Bind uniform buffers.
+    MetalBuffer* uniformsToBind[Program::UNIFORM_BINDING_COUNT] = { nil };
+    NSUInteger uniformOffsets[Program::UNIFORM_BINDING_COUNT] = { 0 };
+    enumerateBoundBuffers(BufferObjectBinding::UNIFORM,
+            [&uniformsToBind, &uniformOffsets](const BufferState& state, MetalBuffer* buffer,
+                    uint32_t index) {
+        uniformsToBind[index] = buffer;
+        uniformOffsets[index] = state.offset;
+    });
+    MetalBuffer::bindBuffers(getPendingCommandBuffer(mContext), computeEncoder,
+            UNIFORM_BUFFER_BINDING_START, MetalBuffer::Stage::COMPUTE, uniformsToBind,
+            uniformOffsets, Program::UNIFORM_BINDING_COUNT);
+
+    // Bind SSBOs.
+    MetalBuffer* ssbosToBind[MAX_SSBO_COUNT] = { nil };
+    NSUInteger ssboOffsets[MAX_SSBO_COUNT] = { 0 };
+    enumerateBoundBuffers(BufferObjectBinding::SHADER_STORAGE,
+            [&ssbosToBind, &ssboOffsets](const BufferState& state, MetalBuffer* buffer,
+                    uint32_t index) {
+        ssbosToBind[index] = buffer;
+        ssboOffsets[index] = state.offset;
+    });
+    MetalBuffer::bindBuffers(getPendingCommandBuffer(mContext), computeEncoder, SSBO_BINDING_START,
+            MetalBuffer::Stage::COMPUTE, ssbosToBind, ssboOffsets, MAX_SSBO_COUNT);
+
+    [computeEncoder setComputePipelineState:computePipelineState];
+
+    MTLSize threadgroupsPerGrid = MTLSizeMake(workGroupCount.x, workGroupCount.y, workGroupCount.z);
+    // FIXME: the threadgroup size should be specified in the Program
+    MTLSize threadsPerThreadgroup = MTLSizeMake(16u, 1u, 1u);
+    [computeEncoder dispatchThreadgroups:threadgroupsPerGrid
+                   threadsPerThreadgroup:threadsPerThreadgroup];
+
+    [computeEncoder endEncoding];
 }
 
 void MetalDriver::beginTimerQuery(Handle<HwTimerQuery> tqh) {
@@ -1652,17 +1787,35 @@ void MetalDriver::endTimerQuery(Handle<HwTimerQuery> tqh) {
     mContext->timerQueryImpl->endTimeElapsedQuery(tq);
 }
 
-void MetalDriver::enumerateBoundUniformBuffers(
-        const std::function<void(const UniformBufferState&, MetalBuffer*, uint32_t)>& f) {
-    for (uint32_t i = 0; i < Program::UNIFORM_BINDING_COUNT; i++) {
-        auto& thisUniform = mContext->uniformState[i];
-        if (!thisUniform.bound) {
-            continue;
+void MetalDriver::enumerateBoundBuffers(BufferObjectBinding bindingType,
+        const std::function<void(const BufferState&, MetalBuffer*, uint32_t)>& f) {
+    assert_invariant(bindingType == BufferObjectBinding::UNIFORM ||
+            bindingType == BufferObjectBinding::SHADER_STORAGE);
+
+    auto enumerate = [&](auto arrayType){
+        for (auto i = 0u; i < arrayType.size(); i++) {
+            const auto& thisBuffer = arrayType[i];
+            if (!thisBuffer.bound) {
+                continue;
+            }
+            f(thisBuffer, thisBuffer.buffer->getBuffer(), i);
         }
-        f(thisUniform, thisUniform.buffer->getBuffer(), i);
+    };
+
+    switch (bindingType) {
+        default:
+        case (BufferObjectBinding::UNIFORM):
+            enumerate(mContext->uniformState);
+            break;
+        case (BufferObjectBinding::SHADER_STORAGE):
+            enumerate(mContext->ssboState);
+            break;
     }
 }
 
+void MetalDriver::resetState(int) {
+}
+
 // explicit instantiation of the Dispatcher
 template class ConcreteDispatcher<MetalDriver>;
 

filament/backend/src/metal/MetalEnums.h

@@ -337,8 +337,6 @@ constexpr inline MTLPrimitiveType getMetalPrimitiveType(PrimitiveType type) noex
         case PrimitiveType::LINE_STRIP: return MTLPrimitiveTypeLineStrip;
         case PrimitiveType::TRIANGLES: return MTLPrimitiveTypeTriangle;
         case PrimitiveType::TRIANGLE_STRIP: return MTLPrimitiveTypeTriangleStrip;
-        case PrimitiveType::NONE:
-            ASSERT_POSTCONDITION(false, "NONE is not a valid primitive type.");
     }
 }
 

filament/backend/src/metal/MetalHandles.h

@@ -118,15 +118,18 @@ private:
 
 class MetalBufferObject : public HwBufferObject {
 public:
-    MetalBufferObject(MetalContext& context, BufferUsage usage, uint32_t byteCount);
+    MetalBufferObject(MetalContext& context, BufferObjectBinding bindingType, BufferUsage usage,
+         uint32_t byteCount);
 
     void updateBuffer(void* data, size_t size, uint32_t byteOffset);
     void updateBufferUnsynchronized(void* data, size_t size, uint32_t byteOffset);
     MetalBuffer* getBuffer() { return &buffer; }
 
-    // Tracks which uniform buffers this buffer object is bound into.
+    // Tracks which uniform/ssbo buffers this buffer object is bound into.
     static_assert(Program::UNIFORM_BINDING_COUNT <= 32);
+    static_assert(MAX_SSBO_COUNT <= 32);
     utils::bitset32 boundUniformBuffers;
+    utils::bitset32 boundSsbos;
 
 private:
     MetalBuffer buffer;
@@ -163,6 +166,7 @@ struct MetalProgram : public HwProgram {
 
     id<MTLFunction> vertexFunction;
     id<MTLFunction> fragmentFunction;
+    id<MTLFunction> computeFunction;
 
     Program::SamplerGroupInfo samplerGroupInfo;
 

filament/backend/src/metal/MetalHandles.mm

@@ -26,6 +26,7 @@
 
 #include "private/backend/BackendUtils.h"
 
+#include <utils/compiler.h>
 #include <utils/Panic.h>
 #include <utils/trap.h>
 #include <utils/debug.h>
@@ -270,8 +271,9 @@ void MetalSwapChain::scheduleFrameCompletedCallback() {
     }];
 }
 
-MetalBufferObject::MetalBufferObject(MetalContext& context, BufferUsage usage, uint32_t byteCount)
-        : HwBufferObject(byteCount), buffer(context, usage, byteCount) {}
+MetalBufferObject::MetalBufferObject(MetalContext& context, BufferObjectBinding bindingType,
+        BufferUsage usage, uint32_t byteCount)
+        : HwBufferObject(byteCount), buffer(context, bindingType, usage, byteCount) {}
 
 void MetalBufferObject::updateBuffer(void* data, size_t size, uint32_t byteOffset) {
     buffer.copyIntoBuffer(data, size, byteOffset);
@@ -287,7 +289,7 @@ MetalVertexBuffer::MetalVertexBuffer(MetalContext& context, uint8_t bufferCount,
 
 MetalIndexBuffer::MetalIndexBuffer(MetalContext& context, BufferUsage usage, uint8_t elementSize,
         uint32_t indexCount) : HwIndexBuffer(elementSize, indexCount),
-        buffer(context, usage, elementSize * indexCount, true) { }
+        buffer(context, BufferObjectBinding::VERTEX, usage, elementSize * indexCount, true) { }
 
 void MetalRenderPrimitive::setBuffers(MetalVertexBuffer* vertexBuffer, MetalIndexBuffer*
         indexBuffer) {
@@ -338,13 +340,12 @@ void MetalRenderPrimitive::setBuffers(MetalVertexBuffer* vertexBuffer, MetalInde
 
 MetalProgram::MetalProgram(id<MTLDevice> device, const Program& program) noexcept
     : HwProgram(program.getName()), vertexFunction(nil), fragmentFunction(nil),
-        isValid(false) {
+            computeFunction(nil), isValid(false) {
 
     using MetalFunctionPtr = __strong id<MTLFunction>*;
 
-    // FIXME: rework this to handle compute
-    //static_assert(Program::SHADER_TYPE_COUNT == 2, "Only vertex and fragment shaders expected.");
-    MetalFunctionPtr shaderFunctions[2] = { &vertexFunction, &fragmentFunction };
+    static_assert(Program::SHADER_TYPE_COUNT == 3, "Only vertex, fragment, and/or compute shaders expected.");
+    MetalFunctionPtr shaderFunctions[3] = { &vertexFunction, &fragmentFunction, &computeFunction };
 
     const auto& sources = program.getShadersSource();
     for (size_t i = 0; i < Program::SHADER_TYPE_COUNT; i++) {
@@ -387,12 +388,19 @@ MetalProgram::MetalProgram(id<MTLDevice> device, const Program& program) noexcep
             }, sc.value);
         }
 
-        *shaderFunctions[i] = [library newFunctionWithName:@"main0"
-                                            constantValues:constants
-                                                     error:&error];
-
+        id<MTLFunction> function = [library newFunctionWithName:@"main0"
+                                                 constantValues:constants
+                                                          error:&error];
+        assert_invariant(function);
+        *shaderFunctions[i] = function;
     }
 
+    UTILS_UNUSED_IN_RELEASE const bool isRasterizationProgram =
+            vertexFunction != nil && fragmentFunction != nil;
+    UTILS_UNUSED_IN_RELEASE const bool isComputeProgram = computeFunction != nil;
+    // The program must be either a rasterization program XOR a compute program.
+    assert_invariant(isRasterizationProgram != isComputeProgram);
+
     // All stages of the program have compiled successfully, this is a valid program.
     isValid = true;
 
@@ -1046,6 +1054,7 @@ void MetalRenderTarget::setUpRenderPassAttachments(MTLRenderPassDescriptor* desc
         descriptor.depthAttachment.slice = 0;
         const bool discard = any(discardFlags & TargetBufferFlags::DEPTH);
         if (!discard) {
+            assert_invariant(context->supportsAutoDepthResolve);
             descriptor.depthAttachment.resolveTexture = depthAttachment.getTexture();
             descriptor.depthAttachment.resolveLevel = depthAttachment.level;
             descriptor.depthAttachment.resolveSlice = depthAttachment.layer;
@@ -1078,10 +1087,14 @@ void MetalRenderTarget::setUpRenderPassAttachments(MTLRenderPassDescriptor* desc
         descriptor.stencilAttachment.slice = 0;
         const bool discard = any(discardFlags & TargetBufferFlags::STENCIL);
         if (!discard) {
+            assert_invariant(context->supportsAutoDepthResolve);
             descriptor.stencilAttachment.resolveTexture = stencilAttachment.getTexture();
             descriptor.stencilAttachment.resolveLevel = stencilAttachment.level;
             descriptor.stencilAttachment.resolveSlice = stencilAttachment.layer;
             descriptor.stencilAttachment.storeAction = MTLStoreActionMultisampleResolve;
+            if (@available(iOS 12.0, *)) {
+                descriptor.stencilAttachment.stencilResolveFilter = MTLMultisampleStencilResolveFilterSample0;
+            }
         }
     }
 }

filament/backend/src/metal/MetalState.h

@@ -37,6 +37,8 @@ inline bool operator==(const SamplerParams& lhs, const SamplerParams& rhs) {
     return lhs.u == rhs.u;
 }
 
+//   Rasterization Bindings
+//   ----------------------
 //   Bindings    Buffer name                          Count
 //   ------------------------------------------------------
 //   0           Zero buffer (placeholder vertex buffer)  1
@@ -46,6 +48,16 @@ inline bool operator==(const SamplerParams& lhs, const SamplerParams& rhs) {
 //
 //   Total                                               31
 
+//   Compute Bindings
+//   ----------------------
+//   Bindings    Buffer name                          Count
+//   ------------------------------------------------------
+//   0-3         SSBO buffers                             4   MAX_SSBO_COUNT
+//   17-26       Uniform buffers                         10   Program::UNIFORM_BINDING_COUNT
+//   27-30       Sampler groups (argument buffers)        4   Program::SAMPLER_BINDING_COUNT
+//
+//   Total                                               18
+
 // The total number of vertex buffer "slots" that the Metal backend can bind.
 // + 1 to account for the zero buffer, a placeholder buffer used internally by the Metal backend.
 // MAX_VERTEX_BUFFER_COUNT represents the max number of vertex buffers Filament can bind.
@@ -59,6 +71,7 @@ static constexpr uint32_t USER_VERTEX_BUFFER_BINDING_START = 1u;
 
 // These constants must match the equivalent in CodeGenerator.h.
 static constexpr uint32_t UNIFORM_BUFFER_BINDING_START = 17u;
+static constexpr uint32_t SSBO_BINDING_START = 0u;
 static constexpr uint32_t SAMPLER_GROUP_BINDING_START = 27u;
 
 // Forward declarations necessary here, definitions at end of file.
@@ -335,27 +348,12 @@ using DepthStencilStateCache = StateCache<DepthStencilState, id<MTLDepthStencilS
 
 class MetalBufferObject;
 
-struct UniformBufferState {
+struct BufferState {
     MetalBufferObject* buffer = nullptr;  // 8 bytes
     uint32_t offset = 0;                  // 4 bytes
     bool bound = false;                   // 1 byte
-    char padding[3] = { 0 };              // 3 bytes
-
-    bool operator==(const UniformBufferState& rhs) const noexcept {
-        return this->bound == rhs.bound &&
-               this->buffer == rhs.buffer &&
-               this->offset == rhs.offset;
-    }
-
-    bool operator!=(const UniformBufferState& rhs) const noexcept {
-        return !operator==(rhs);
-    }
 };
 
-static_assert(sizeof(UniformBufferState) == 16, "UniformBufferState unexpected size.");
-
-using UniformBufferStateTracker = StateTracker<UniformBufferState>;
-
 // Sampler states
 
 struct SamplerState {

filament/backend/src/noop/NoopDriver.cpp

@@ -328,4 +328,7 @@ void NoopDriver::beginTimerQuery(Handle<HwTimerQuery> tqh) {
 void NoopDriver::endTimerQuery(Handle<HwTimerQuery> tqh) {
 }
 
+void NoopDriver::resetState(int) {
+}
+
 } // namespace filament

filament/backend/src/opengl/OpenGLContext.cpp

@@ -61,6 +61,8 @@ OpenGLContext::OpenGLContext() noexcept {
     glGetIntegerv(GL_MAX_DRAW_BUFFERS, &gets.max_draw_buffers);
     glGetIntegerv(GL_MAX_TEXTURE_IMAGE_UNITS, &gets.max_texture_image_units);
     glGetIntegerv(GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS, &gets.max_combined_texture_image_units);
+    glGetIntegerv(GL_MAX_TRANSFORM_FEEDBACK_SEPARATE_ATTRIBS, &gets.max_transform_feedback_separate_attribs);
+    glGetIntegerv(GL_MAX_UNIFORM_BUFFER_BINDINGS, &gets.max_uniform_buffer_bindings);
     glGetIntegerv(GL_UNIFORM_BUFFER_OFFSET_ALIGNMENT, &gets.uniform_buffer_offset_alignment);
 
     constexpr auto const caps3 = FEATURE_LEVEL_CAPS[+FeatureLevel::FEATURE_LEVEL_3];
@@ -247,31 +249,7 @@ OpenGLContext::OpenGLContext() noexcept {
     flush(slog.v);
 #endif
 
-    /*
-     * Set our default state
-     */
-
-    // We need to make sure our internal state matches the GL state when we start.
-    // (some of these calls may be unneeded as they might be the gl defaults)
-    glDisable(GL_BLEND);
-    glDisable(GL_CULL_FACE);
-    glDisable(GL_SCISSOR_TEST);
-    glDisable(GL_DEPTH_TEST);
-    glDisable(GL_STENCIL_TEST);
-    glDisable(GL_DITHER);
-    glDisable(GL_SAMPLE_ALPHA_TO_COVERAGE);
-    glDisable(GL_SAMPLE_COVERAGE);
-    glDisable(GL_POLYGON_OFFSET_FILL);
-
-    // Point sprite size and seamless cubemap filtering are disabled by default in desktop GL.
-    // In OpenGL ES, these flags do not exist because they are always on.
-#if BACKEND_OPENGL_VERSION == BACKEND_OPENGL_VERSION_GL
-    enable(GL_PROGRAM_POINT_SIZE);
-#endif
-
-#ifdef GL_ARB_seamless_cube_map
-    enable(GL_TEXTURE_CUBE_MAP_SEAMLESS);
-#endif
+    setDefaultState();
 
 #ifdef GL_EXT_texture_filter_anisotropic
     if (ext.EXT_texture_filter_anisotropic) {
@@ -290,10 +268,6 @@ OpenGLContext::OpenGLContext() noexcept {
     }
 #endif
 
-#ifdef GL_FRAGMENT_SHADER_DERIVATIVE_HINT
-    glHint(GL_FRAGMENT_SHADER_DERIVATIVE_HINT, GL_NICEST);
-#endif
-
 #if !defined(NDEBUG) && defined(GL_KHR_debug)
     if (ext.KHR_debug) {
         auto cb = [](GLenum source, GLenum type, GLuint id, GLenum severity, GLsizei length,
@@ -325,6 +299,50 @@ OpenGLContext::OpenGLContext() noexcept {
         glDebugMessageCallback(cb, nullptr);
     }
 #endif
+}
+
+void OpenGLContext::setDefaultState() noexcept {
+    // We need to make sure our internal state matches the GL state when we start.
+    // (some of these calls may be unneeded as they might be the gl defaults)
+    GLenum caps[] = {
+        GL_BLEND,
+        GL_CULL_FACE,
+        GL_SCISSOR_TEST,
+        GL_DEPTH_TEST,
+        GL_STENCIL_TEST,
+        GL_DITHER,
+        GL_SAMPLE_ALPHA_TO_COVERAGE,
+        GL_SAMPLE_COVERAGE,
+        GL_POLYGON_OFFSET_FILL,  
+    };
+    size_t capsCount = sizeof(caps) / sizeof(caps[0]);
+
+    UTILS_NOUNROLL
+    for (GLenum capi = 0; capi < capsCount; ++capi) {
+        size_t capIndex = getIndexForCap(caps[capi]);
+        GLenum cap = caps[capi];
+        if (state.enables.caps[capIndex]) {
+            glEnable(cap);
+        } else {
+            glDisable(cap);
+        }
+    }
+
+    // Point sprite size and seamless cubemap filtering are disabled by default in desktop GL.
+    // In OpenGL ES, these flags do not exist because they are always on.
+#if BACKEND_OPENGL_VERSION == BACKEND_OPENGL_VERSION_GL
+    glEnable(GL_PROGRAM_POINT_SIZE);
+    enable(GL_PROGRAM_POINT_SIZE);
+#endif
+
+#ifdef GL_ARB_seamless_cube_map
+    glEnable(GL_TEXTURE_CUBE_MAP_SEAMLESS);
+    enable(GL_TEXTURE_CUBE_MAP_SEAMLESS);
+#endif
+
+#ifdef GL_FRAGMENT_SHADER_DERIVATIVE_HINT
+    glHint(GL_FRAGMENT_SHADER_DERIVATIVE_HINT, GL_NICEST);
+#endif
 
 #if defined(GL_EXT_clip_control) || defined(GL_ARB_clip_control) || defined(GL_VERSION_4_5)
     if (ext.EXT_clip_control) {
@@ -521,4 +539,161 @@ void OpenGLContext::deleteVertexArrays(GLsizei n, const GLuint* arrays) noexcept
     }
 }
 
+void OpenGLContext::resetState() noexcept {
+    // Force GL state to match the Filament state
+    glBindFramebuffer(GL_DRAW_FRAMEBUFFER, state.draw_fbo);
+    glBindFramebuffer(GL_READ_FRAMEBUFFER, state.read_fbo);
+
+    // state.program
+    glUseProgram(state.program.use);
+
+    // state.vao
+    if (state.vao.p) {
+        glBindVertexArray(state.vao.p->vao);
+    } else {
+        bindVertexArray(nullptr);
+    }
+
+    // state.raster
+    glFrontFace(state.raster.frontFace);
+    glCullFace(state.raster.cullFace);
+    glBlendEquationSeparate(state.raster.blendEquationRGB, state.raster.blendEquationA);
+    glBlendFuncSeparate(
+        state.raster.blendFunctionSrcRGB, 
+        state.raster.blendFunctionDstRGB,
+        state.raster.blendFunctionSrcA,
+        state.raster.blendFunctionDstA
+    );
+    glColorMask(
+        state.raster.colorMask, 
+        state.raster.colorMask, 
+        state.raster.colorMask, 
+        state.raster.colorMask
+    );
+    glDepthMask(state.raster.depthMask);
+    glDepthFunc(state.raster.depthFunc);
+    
+    // state.stencil
+    glStencilFuncSeparate(
+        GL_FRONT, 
+        state.stencil.front.func.func, 
+        state.stencil.front.func.ref, 
+        state.stencil.front.func.mask
+    );
+    glStencilFuncSeparate(
+        GL_BACK, 
+        state.stencil.back.func.func, 
+        state.stencil.back.func.ref, 
+        state.stencil.back.func.mask
+    );
+    glStencilOpSeparate(
+        GL_FRONT, 
+        state.stencil.front.op.sfail,
+        state.stencil.front.op.dpfail,
+        state.stencil.front.op.dppass
+    );
+    glStencilOpSeparate(
+        GL_BACK, 
+        state.stencil.back.op.sfail,
+        state.stencil.back.op.dpfail,
+        state.stencil.back.op.dppass
+    );
+    glStencilMaskSeparate(GL_FRONT, state.stencil.front.stencilMask);
+    glStencilMaskSeparate(GL_BACK, state.stencil.back.stencilMask);
+
+    // state.polygonOffset
+    glPolygonOffset(state.polygonOffset.factor, state.polygonOffset.units);
+
+    // state.enables
+    setDefaultState();
+
+    // state.buffers
+    // Reset state.buffers to its default state to avoid the complexity and error-prone
+    // nature of resetting the GL state to its existing state
+    state.buffers = {};
+    GLenum bufferTargets[] = {
+        GL_UNIFORM_BUFFER,
+        GL_TRANSFORM_FEEDBACK_BUFFER,
+#if !defined(__EMSCRIPTEN__)
+        GL_SHADER_STORAGE_BUFFER,
+#endif
+        GL_ARRAY_BUFFER,
+        GL_COPY_READ_BUFFER,
+        GL_COPY_WRITE_BUFFER,
+        GL_ELEMENT_ARRAY_BUFFER,
+        GL_PIXEL_PACK_BUFFER,
+        GL_PIXEL_UNPACK_BUFFER,
+    };
+    size_t bufferTargetCount = sizeof(bufferTargets) / sizeof(bufferTargets[0]);
+    for (size_t targetArrayIndex = 0; targetArrayIndex < bufferTargetCount; ++targetArrayIndex) {
+        GLenum target = bufferTargets[targetArrayIndex];
+        glBindBuffer(target, 0);
+    }
+
+    for (size_t bufferIndex = 0; bufferIndex < MAX_BUFFER_BINDINGS; ++bufferIndex) {
+        if (bufferIndex < (size_t)gets.max_uniform_buffer_bindings) {
+            glBindBufferBase(GL_UNIFORM_BUFFER, bufferIndex, 0);
+        }
+
+        if (bufferIndex < (size_t)gets.max_transform_feedback_separate_attribs) {
+            glBindBufferBase(GL_TRANSFORM_FEEDBACK_BUFFER, bufferIndex, 0);
+        }
+    }
+
+    // state.textures
+    // Reset state.textures to its default state to avoid the complexity and error-prone
+    // nature of resetting the GL state to its existing state
+    state.textures = {};
+    const GLuint textureTargets[] = {
+        GL_TEXTURE_2D,
+        GL_TEXTURE_2D_ARRAY,
+        GL_TEXTURE_CUBE_MAP,
+        GL_TEXTURE_3D,
+#if !defined(__EMSCRIPTEN__)
+        GL_TEXTURE_2D_MULTISAMPLE,
+        GL_TEXTURE_EXTERNAL_OES,
+        GL_TEXTURE_CUBE_MAP_ARRAY,
+#endif
+    };
+    const size_t textureTargetCount = sizeof(textureTargets) / sizeof(textureTargets[0]);
+    for (GLint unit = 0; unit < gets.max_combined_texture_image_units; ++unit) {
+        glActiveTexture(GL_TEXTURE0 + unit);
+        glBindSampler(unit, 0);
+
+        for (size_t textureTargetArrayIndex = 0; textureTargetArrayIndex < textureTargetCount; ++textureTargetArrayIndex) {
+            GLuint target = textureTargets[textureTargetArrayIndex];
+            glBindTexture(target, 0);
+        }
+    }
+    glActiveTexture(GL_TEXTURE0 + state.textures.active);
+
+    // state.unpack
+    glPixelStorei(GL_UNPACK_ALIGNMENT, state.unpack.alignment);
+    glPixelStorei(GL_UNPACK_ROW_LENGTH, state.unpack.row_length);
+    glPixelStorei(GL_UNPACK_SKIP_PIXELS, state.unpack.skip_pixels);
+    glPixelStorei(GL_UNPACK_SKIP_ROWS, state.unpack.skip_row);
+
+    // state.pack
+    glPixelStorei(GL_PACK_ALIGNMENT, state.pack.alignment);
+    glPixelStorei(GL_PACK_ROW_LENGTH, state.pack.row_length);
+    glPixelStorei(GL_PACK_SKIP_PIXELS, state.pack.skip_pixels);
+    glPixelStorei(GL_PACK_SKIP_ROWS, state.pack.skip_row);
+
+    // state.window
+    glScissor(
+        state.window.scissor.x, 
+        state.window.scissor.y, 
+        state.window.scissor.z, 
+        state.window.scissor.w
+    );
+    glViewport(
+        state.window.viewport.x,
+        state.window.viewport.y,
+        state.window.viewport.z,
+        state.window.viewport.w
+    );
+    glDepthRangef(state.window.depthRange.x, state.window.depthRange.y);
+    
+}
+
 } // namesapce filament

filament/backend/src/opengl/OpenGLContext.h

@@ -73,6 +73,7 @@ public:
 
     ShaderModel getShaderModel() const noexcept { return mShaderModel; }
 
+    void resetState() noexcept;
 
     inline void useProgram(GLuint program) noexcept;
 
@@ -129,6 +130,8 @@ public:
         GLint max_uniform_block_size;
         GLint max_texture_image_units;
         GLint max_combined_texture_image_units;
+        GLint max_transform_feedback_separate_attribs;
+        GLint max_uniform_buffer_bindings; 
         GLint uniform_buffer_offset_alignment;
     } gets = {};
 
@@ -409,6 +412,8 @@ private:
         }
     }
 
+    void setDefaultState() noexcept;
+
     static constexpr const size_t TEXTURE_TARGET_COUNT =
             sizeof(state.textures.units[0].targets) / sizeof(state.textures.units[0].targets[0]);
 

filament/backend/src/opengl/OpenGLDriver.cpp

@@ -235,6 +235,9 @@ ShaderModel OpenGLDriver::getShaderModel() const noexcept {
 // ------------------------------------------------------------------------------------------------
 // Change and track GL state
 // ------------------------------------------------------------------------------------------------
+void OpenGLDriver::resetState(int) {
+    mContext.resetState();
+}
 
 void OpenGLDriver::bindSampler(GLuint unit, GLuint sampler) noexcept {
     mContext.bindSampler(unit, sampler);

filament/backend/src/ostream.cpp

@@ -54,7 +54,6 @@ io::ostream& operator<<(io::ostream& out, PrimitiveType type) {
         CASE(PrimitiveType, LINES)
         CASE(PrimitiveType, LINE_STRIP)
         CASE(PrimitiveType, POINTS)
-        CASE(PrimitiveType, NONE)
     }
     return out;
 }

filament/backend/src/vulkan/PlatformVkLinuxGGP.cpp

@@ -0,0 +1,74 @@
+/*
+ * Copyright (C) 2018 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "vulkan/PlatformVkLinuxGGP.h"
+
+#include <bluevk/BlueVK.h>
+#if defined(FILAMENT_SUPPORTS_GGP)
+#include <ggp_c/ggp.h>
+#endif
+#include <utils/Panic.h>
+
+#include "VulkanDriverFactory.h"
+
+using namespace bluevk;
+
+namespace filament::backend {
+
+Driver* PlatformVkLinuxGGP::createDriver(
+    void* const sharedContext,
+    const Platform::DriverConfig& driverConfig) noexcept {
+#if defined(FILAMENT_SUPPORTS_GGP)
+  ASSERT_PRECONDITION(sharedContext == nullptr,
+                      "Vulkan does not support shared contexts.");
+  const char* requiredInstanceExtensions[] = {
+      VK_GGP_STREAM_DESCRIPTOR_SURFACE_EXTENSION_NAME};
+  return VulkanDriverFactory::create(this, requiredInstanceExtensions,
+                                     sizeof(requiredInstanceExtensions) /
+                                         sizeof(requiredInstanceExtensions[0]),
+                                     driverConfig);
+#else
+  PANIC_PRECONDITION("Filament does not support GGP.");
+  return nullptr;
+#endif
+}
+
+void* PlatformVkLinuxGGP::createVkSurfaceKHR(void* nativeWindow, void* instance,
+                                             uint64_t flags) noexcept {
+#if defined(FILAMENT_SUPPORTS_GGP)
+  VkStreamDescriptorSurfaceCreateInfoGGP surface_create_info = {
+      VK_STRUCTURE_TYPE_STREAM_DESCRIPTOR_SURFACE_CREATE_INFO_GGP};
+  surface_create_info.streamDescriptor = kGgpPrimaryStreamDescriptor;
+  PFN_vkCreateStreamDescriptorSurfaceGGP fpCreateStreamDescriptorSurfaceGGP =
+      reinterpret_cast<PFN_vkCreateStreamDescriptorSurfaceGGP>(
+          vkGetInstanceProcAddr(static_cast<VkInstance>(instance),
+                                "vkCreateStreamDescriptorSurfaceGGP"));
+  ASSERT_PRECONDITION(fpCreateStreamDescriptorSurfaceGGP != nullptr,
+                      "Error getting VkInstance "
+                      "function vkCreateStreamDescriptorSurfaceGGP");
+  VkSurfaceKHR surface = nullptr;
+  VkResult res = fpCreateStreamDescriptorSurfaceGGP(
+      static_cast<VkInstance>(instance), &surface_create_info, nullptr,
+      &surface);
+  ASSERT_PRECONDITION(res == VK_SUCCESS, "Error in vulkan: %d", res);
+  return surface;
+#else
+  PANIC_PRECONDITION("Filament does not support GGP.");
+  return nullptr;
+#endif
+}
+
+}  // namespace filament::backend

filament/backend/src/vulkan/PlatformVkLinuxGGP.h

@@ -0,0 +1,41 @@
+/*
+ * Copyright (C) 2018 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef TNT_FILAMENT_BACKEND_VULKAN_PLATFORM_VK_LINUX_GGP_H
+#define TNT_FILAMENT_BACKEND_VULKAN_PLATFORM_VK_LINUX_GGP_H
+
+#include <backend/DriverEnums.h>
+#include <stdint.h>
+
+#include "VulkanPlatform.h"
+
+namespace filament::backend {
+
+class PlatformVkLinuxGGP final : public VulkanPlatform {
+ public:
+  Driver* createDriver(
+      void* const sharedContext,
+      const Platform::DriverConfig& driverConfig) noexcept override;
+
+  void* createVkSurfaceKHR(void* nativeWindow, void* instance,
+                           uint64_t flags) noexcept override;
+
+  int getOSVersion() const noexcept override { return 0; }
+};
+
+}  // namespace filament::backend
+
+#endif  // TNT_FILAMENT_BACKEND_VULKAN_PLATFORM_VK_LINUX_GGP_H

filament/backend/src/vulkan/VulkanDriver.cpp

@@ -2012,6 +2012,9 @@ void VulkanDriver::debugCommandBegin(CommandStream* cmds, bool synchronous, cons
 #endif
 }
 
+void VulkanDriver::resetState(int) {
+}
+
 // explicit instantiation of the Dispatcher
 template class ConcreteDispatcher<VulkanDriver>;
 

filament/backend/src/vulkan/VulkanHandles.cpp

@@ -263,7 +263,6 @@ VulkanBufferObject::VulkanBufferObject(VulkanContext& context, VulkanStagePool&
 void VulkanRenderPrimitive::setPrimitiveType(PrimitiveType pt) {
     this->type = pt;
     switch (pt) {
-        case PrimitiveType::NONE:
         case PrimitiveType::POINTS:
             primitiveTopology = VK_PRIMITIVE_TOPOLOGY_POINT_LIST;
             break;

filament/backend/test/test_ComputeBasic.cpp

@@ -22,25 +22,29 @@ using namespace filament::backend;
 TEST_F(ComputeTest, basic) {
     auto& driver = getDriverApi();
 
-    std::string_view shader_gles310 = {R"(
+    std::string shader_gles310 = {R"(
 #version 310 es
 layout(local_size_x = 16) in;
 void main() {
 }
 )"};
 
-    std::string_view shader_gl450 = {R"(
+    std::string shader_gl450 = {R"(
 #version 450 core
 layout(local_size_x = 16) in;
 void main() {
 }
 )"};
 
-    std::string_view shader_msl = {R"(
-// TODO: msl test
+    std::string shader_msl = {R"(
+#include <simd/simd.h>
+#include <metal_stdlib>
+using namespace metal;
+constant uint3 WorkGroupSize [[maybe_unused]] = uint3(16u, 1u, 1u);
+kernel void main0() {}
 )"};
 
-    std::string_view shader_spirv = {R"(
+    std::string shader_spirv = {R"(
 // TODO: spirv test
 )"};
 
@@ -54,7 +58,7 @@ void main() {
     }
 
     Program program;
-    program.shader(ShaderStage::COMPUTE, shader.data(), shader.size());
+    program.shader(ShaderStage::COMPUTE, shader.data(), shader.size() + 1);
 
     Handle<HwProgram> ph = driver.createProgram(std::move(program));
     driver.dispatchCompute(ph, { 1, 1, 1 });
@@ -68,7 +72,7 @@ void main() {
 TEST_F(ComputeTest, copy) {
     auto& driver = getDriverApi();
 
-    std::string_view shader_gles310 = {R"(
+    std::string shader_gles310 = {R"(
 #version 310 es
 layout(local_size_x = 16) in;
 layout(std430) buffer;
@@ -80,7 +84,7 @@ void main() {
 }
 )"};
 
-    std::string_view shader_gl450 = {R"(
+    std::string shader_gl450 = {R"(
 #version 450 core
 layout(local_size_x = 16) in;
 layout(std430) buffer;
@@ -92,11 +96,25 @@ void main() {
 }
 )"};
 
-    std::string_view shader_msl = {R"(
-// TODO: msl test
+    std::string shader_msl = {R"(
+#include <simd/simd.h>
+#include <metal_stdlib>
+using namespace metal;
+struct Output_data {
+    float elements[1];
+};
+struct Input_data {
+    float elements[1];
+};
+constant uint3 WorkGroupSize [[maybe_unused]] = uint3(16u, 1u, 1u);
+kernel void main0(device Output_data& output_data [[buffer(0)]],
+        device Input_data& input_data [[buffer(1)]],
+        uint3 GlobalInvocationID [[thread_position_in_grid]]) {
+    output_data.elements[GlobalInvocationID.x] = input_data.elements[GlobalInvocationID.x];
+}
 )"};
 
-    std::string_view shader_spirv = {R"(
+    std::string shader_spirv = {R"(
 // TODO: spirv test
 )"};
 
@@ -119,12 +137,14 @@ void main() {
         return v;
     });
 
+    driver.startCapture(0);
+
     auto output_data = driver.createBufferObject(size, BufferObjectBinding::SHADER_STORAGE, BufferUsage::STATIC);
     auto input_data = driver.createBufferObject(size, BufferObjectBinding::SHADER_STORAGE, BufferUsage::STATIC);
     driver.updateBufferObject(input_data, { data.data(), size }, 0);
 
     Program program;
-    program.shader(ShaderStage::COMPUTE, shader.data(), shader.size());
+    program.shader(ShaderStage::COMPUTE, shader.data(), shader.size() + 1);
     Handle<HwProgram> ph = driver.createProgram(std::move(program));
 
 
@@ -144,6 +164,8 @@ void main() {
     driver.destroyBufferObject(output_data);
     driver.finish();
 
+    driver.stopCapture(0);
+
     executeCommands();
     getDriver().purge();
 

filament/docs/Vulkan.md.html

@@ -11,7 +11,7 @@
 </style>
 
 **Vulkan Backend**
-          *Last updated: 11 June 2021*
+          *Last updated: 26 October 2022*
 
 ![](../../docs/images/filament_logo.png)
 
@@ -47,7 +47,7 @@ single `VkDevice` and `VkInstance`.
 *  |                                          |   |                                         |  *
 *  |------------------------------------------|   |-----------------------------------------|  *
 *  |                                          |   |                                         |  *
-*  |    Helpers            VulkanContext      |   |    SwapContext       SwapContext        |  *
+*  |    Helpers            VulkanContext      |   |    VulkanTexture    VulkanTexture       |  *
 *  |  .--------------.    .---------------.   |   |   .----------------+----------------.   |  *
 *  |  | PipelineCache|   | VkDevice        |  |   |  | VkImage         | VkImage         |  |  *
 *  |  +--------------+   | VkInstance      |  |   |  | VkImageView     | VkImageView     |  |  *
@@ -91,13 +91,14 @@ that are defined in `DriverBase.h`.
 
 Most of the handle types do not contain much logic, but they often do more than mere wrapping of
 Vk objects. Notably, `VulkanRenderTarget` does not have a straightforward mapping to any one Vk
-object. See its class comment for details.
+object. See its class comment for details. Also, `VulkanTexture` and `VulkanSwapChain` have their
+own cpps file because of complexity around layouts and subresources.
 
 VulkanProgram
 : Descends from `HwProgram`, owns a pair of `VkShaderModule`.
 
 VulkanRenderTarget
-: Descends from `HwRenderTarget`, contains references to `VkImage` objects.
+: Descends from `HwRenderTarget`, contains references to `VulkanTexture` objects.
 
 VulkanSwapChain
 : Descends from `HwSwapChain`, has a reference to `VkSurfaceKHR` and `VkSwapchainKHR`.
@@ -127,14 +128,16 @@ they do follow a common set of conventions. For example, most of them implement
 that is called once per frame to free unused objects.
 
 VulkanPipelineCache
-: Cache of VkDescriptorSet and VkPipeline.
+: Cache of VkDescriptorSet, VkPipeline, and VkPipelineLayout.
 
 VulkanDisposer
 : Holds reference counts to Vulkan objects in order to defer their destruction.
 
 VulkanFboCache
 : Cache of VkRenderPass and VkFramebuffer. Contains weak references to VkImage handles, but does not
-  own any pixel data.
+  own any pixel data. The cache key for render passes is quite complex because it needs to include
+  image layout information in a compact way. For details see the comments in the `RenderPassKey`
+  declaration.
 
 VulkanStagePool
 : Pool of VkBuffer objects that can be memory mapped (shared between host and device).
@@ -149,6 +152,71 @@ no subfolders. All backend-specific types and functions live directly in the `fi
 namespace, and we simply prefix our names with "Vulkan" when necessary to avoid ambiguation with
 OpenGL.
 
+# Image Layout Strategy
+
+Adhering to a coherent strategy for image layout transitions is crucial. Without this, validation
+errors (and actual black screens with some GPU vendors) occur far too easily and can feel like
+whack-a-mole.
+
+Vulkan provides several mechanisms for transitioning layout (render passes, barriers) so the
+strategy outlined below specifies which mechanism should be used in various scenarios.
+
+The current image layouts are tracked at the subresource level in `VulkanTexture` using a
+specialized STL-ish container that we designed for this purpose:
+[RangeMap](https://github.com/google/filament/blob/main/libs/utils/include/utils/RangeMap.h).
+
+In the past we tried to avoid tracking since it theoritically duplicates state that exists in the
+GPU vendor's driver. However in practice this was impossible to manage, because complex logic was
+required to deduce the current layout. Also, tracking lets us use `assert_invariant` in many places
+to ensure that the behavior described in this document is implemented properly.  It is much easier
+to catch errors with asserts than with validation.
+
+As of this writing, the current rules are:
+
+(1) **Read-Only Color Textures**
+
+These should be created in an UNDEFINED state and stay that way until data gets uploaded into them,
+at which point the appropriate subresources are transitioned to TRANSFER_DST_OPTIMAL for the upload,
+then to SHADER_READ_ONLY_OPTIMAL, both times using a barrier.  Note that a subresource might be
+updated several times; each time it undergoes these two transitions.
+
+If a shader samples from a texture whose mipmaps are only partially loaded, we might see validation
+warnings about how some subresources are in an UNDEFINED layout.  However if we are properly using
+the `setMinMaxLevels` driver API, then the Vulkan backend will not bind those particular
+subresources, so validation should not complain.
+
+(2) **Writeable Color Textures**
+
+During construction of the HwTexture, these are transitioned to GENERAL using a barrier.
+
+GENERAL layout is not recommended by AMD best practices but in our case we wish to avoid thrashing
+the layout between COLOR_ATTACHMENT_OPTIMAL and SHADER_READ_ONLY_OPTIMAL for post-processing
+effects.  Moreover we do not have a priori knowledge about blitting and read pixels usage.
+
+(3) **Depth Textures**
+
+During construction of the HwTexture, these are transitioned to GENERAL using a barrier.
+
+Note that some post-processing passes (e.g. SSAO) sample from the depth buffer while also enabling
+depth testing. Using GENERAL makes this legal.
+
+(4) **Swap Chain Textures**
+
+During construction, these are transitioned to GENERAL using a barrier. They are then transitioned
+to PRESENT at the end of the frame using a barrier, and transitioned back to GENERAL using the first
+render pass of the frame.
+
+Note that we transition to PRESENT using a barrier, but with some GPU vendors it might be preferable
+to transition them using the last render pass. However we have no way of knowing ahead of time which
+pass is last.
+
+Note that Filament allows swap chains to be used as blit targets and blit sources (e.g. for
+ReadPixels). Using GENERAL makes this easier to manage.
+
+(5) **Headless Swap Chain Textures**
+
+These follow the same rules as writeable color texture and depth textures.
+
 # VulkanPipelineCache
 
 Uniforms, samplers, and shaders are not directly bindable in Vulkan; instead they are bound through
@@ -161,8 +229,6 @@ public methods. See its header file for details.
 
 The pipeline cache also manages the *layout objects*, which constrain the number of uniform buffers
 and samplers that can be bound simultaneously (see `VkDescriptorSetLayout` and `VkPipelineLayout`).
-Here we take advantage of Filament's opinionated shading model in order to keep things simple, which
-allows us to create only one layout for the entire lifetime of the app.
 
 # Memory Allocation
 
@@ -196,7 +262,8 @@ Here we list some of the most significant issues and to-be-done items for the Vu
 
 - Improve the UBO strategy
     - Currently we use staging buffers and perform lots of copies (host-to-host, then host-to-device)
-    - Consider an approach similar to MetalDriver, which effectively implements double-buffering using a pool of shared mem buffers
+    - Consider an approach similar to MetalDriver, which effectively implements double-buffering
+      using a pool of shared mem buffers
 - Consider using subpasses for the depth prepass
 - We lazily create pipeline objects
         1. This is notoriously slow in Vulkan

filament/include/filament/Engine.h

@@ -725,6 +725,22 @@ public:
       */
     utils::JobSystem& getJobSystem() noexcept;
 
+#if defined(__EMSCRIPTEN__)
+    /**
+      * WebGL only: Tells the driver to reset any internal state tracking if necessary.
+      * 
+      * This is only useful when integrating an external renderer into Filament on platforms 
+      * like WebGL, where share contexts do not exist. Filament keeps track of the GL
+      * state it has set (like which texture is bound), and does not re-set that state if
+      * it does not think it needs to. However, if an external renderer has set different
+      * state in the mean time, Filament will use that new state unknowingly.
+      * 
+      * If you are in this situation, call this function - ideally only once per frame, 
+      * immediately after calling Engine::execute().
+      */
+    void resetBackendState() noexcept;
+#endif
+
     DebugRegistry& getDebugRegistry() noexcept;
 
 protected:

filament/include/filament/MaterialInstance.h

@@ -277,6 +277,12 @@ public:
      */
     void setMaskThreshold(float threshold) noexcept;
 
+    /**
+     * Gets the minimum alpha value a fragment must have to not be discarded when the blend
+     * mode is MASKED
+     */
+    float getMaskThreshold() const noexcept;
+
     /**
      * Sets the screen space variance of the filter kernel used when applying specular
      * anti-aliasing. The default value is set to 0.15. The specified value should be between
@@ -284,6 +290,12 @@ public:
      */
     void setSpecularAntiAliasingVariance(float variance) noexcept;
 
+    /**
+     * Gets the screen space variance of the filter kernel used when applying specular
+     * anti-aliasing.
+     */
+    float getSpecularAntiAliasingVariance() const noexcept;
+
     /**
      * Sets the clamping threshold used to suppress estimation errors when applying specular
      * anti-aliasing. The default value is set to 0.2. The specified value should be between 0
@@ -291,6 +303,12 @@ public:
      */
     void setSpecularAntiAliasingThreshold(float threshold) noexcept;
 
+    /**
+     * Gets the clamping threshold used to suppress estimation errors when applying specular
+     * anti-aliasing.
+     */
+    float getSpecularAntiAliasingThreshold() const noexcept;
+
     /**
      * Enables or disables double-sided lighting if the parent Material has double-sided capability,
      * otherwise prints a warning. If double-sided lighting is enabled, backface culling is
@@ -298,36 +316,72 @@ public:
      */
     void setDoubleSided(bool doubleSided) noexcept;
 
+    /**
+     * Returns whether double-sided lighting is enabled when the parent Material has double-sided
+     * capability.
+     */
+    bool isDoubleSided() const noexcept;
+
     /**
      * Specifies how transparent objects should be rendered (default is DEFAULT).
      */
     void setTransparencyMode(TransparencyMode mode) noexcept;
 
+    /**
+     * Returns the transparency mode.
+     */
+    TransparencyMode getTransparencyMode() const noexcept;
+
     /**
      * Overrides the default triangle culling state that was set on the material.
      */
     void setCullingMode(CullingMode culling) noexcept;
 
+    /**
+     * Returns the face culling mode.
+     */
+    CullingMode getCullingMode() const noexcept;
+
     /**
      * Overrides the default color-buffer write state that was set on the material.
      */
     void setColorWrite(bool enable) noexcept;
 
+    /**
+     * Returns whether color write is enabled.
+     */
+    bool isColorWriteEnabled() const noexcept;
+
     /**
      * Overrides the default depth-buffer write state that was set on the material.
      */
     void setDepthWrite(bool enable) noexcept;
 
+    /**
+     * Returns whether depth write is enabled.
+     */
+    bool isDepthWriteEnabled() const noexcept;
+
     /**
      * Overrides the default depth testing state that was set on the material.
      */
     void setDepthCulling(bool enable) noexcept;
 
+    /**
+     * Returns whether depth culling is enabled.
+     */
+    bool isDepthCullingEnabled() const noexcept;
+
     /**
      * Overrides the default stencil-buffer write state that was set on the material.
      */
     void setStencilWrite(bool enable) noexcept;
 
+    /**
+     * Returns whether stencil write is enabled.
+     */
+    bool isStencilWriteEnabled() const noexcept;
+
     /**
      * Sets the stencil comparison function (default is StencilCompareFunc::A).
      *

filament/src/AtlasAllocator.cpp

@@ -0,0 +1,161 @@
+/*
+ * Copyright (C) 2022 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "AtlasAllocator.h"
+
+#include <utils/debug.h>
+
+namespace filament {
+
+using namespace utils;
+
+static inline constexpr std::pair<uint8_t, uint8_t> unmorton(uint16_t m) noexcept {
+    uint32_t r = (m | (uint32_t(m) << 15u)) & 0x55555555u;
+    r = (r | (r >> 1u)) & 0x33333333u;
+    r = (r | (r >> 2u)) & 0x0f0f0f0fu;
+    r =  r | (r >> 4u);
+    return { uint8_t(r), uint8_t(r >> 16u) };
+}
+
+AtlasAllocator::AtlasAllocator(size_t maxTextureSize) noexcept {
+    // round to power-of-two immediately inferior or equal to the size specified.
+    mMaxTextureSizePot = (sizeof(maxTextureSize) * 8 - 1u) - utils::clz(maxTextureSize);
+}
+
+Viewport AtlasAllocator::allocate(size_t textureSize) noexcept {
+    Viewport result{};
+    const size_t powerOfTwo = (sizeof(textureSize) * 8 - 1u) - utils::clz(textureSize);
+
+    // asked for a texture size too large
+    if (UTILS_UNLIKELY(powerOfTwo > mMaxTextureSizePot)) {
+        return result;
+    }
+
+    // asked for a texture size too small
+    if (UTILS_UNLIKELY(mMaxTextureSizePot - powerOfTwo >= QuadTree::height())) {
+        return result;
+    }
+
+    const size_t layer = mMaxTextureSizePot - powerOfTwo;
+    const NodeId loc = allocateInLayer(layer);
+    if (loc.l >= 0) {
+        assert_invariant(loc.l == int8_t(layer));
+        size_t dimension = 1u << powerOfTwo;
+        // find the location of in the texture from the morton code (quadtree position)
+        auto [x, y] = unmorton(loc.code);
+        // scale to our maximum allocation size
+        result.left   = int32_t(x) << powerOfTwo;
+        result.bottom = int32_t(y) << powerOfTwo;
+        result.width  = dimension;
+        result.height = dimension;
+    }
+    return result;
+}
+
+void AtlasAllocator::clear(size_t maxTextureSize) noexcept {
+    std::fill(mQuadTree.begin(), mQuadTree.end(), Node{});
+    mMaxTextureSizePot = (sizeof(maxTextureSize) * 8 - 1u) - utils::clz(maxTextureSize);
+}
+
+AtlasAllocator::NodeId AtlasAllocator::allocateInLayer(size_t maxHeight) noexcept {
+    using namespace QuadTreeUtils;
+
+    NodeId candidate{ -1, 0 };
+    if (UTILS_UNLIKELY(maxHeight > QuadTree::height())) {
+        return candidate;
+    }
+
+    const int8_t n = int8_t(maxHeight);
+
+    QuadTree::traverse(0, 0, n,
+            [this, n, &candidate](NodeId const& curr) -> QuadTree::TraversalResult {
+
+                // we should never reach a level higher than n
+                assert_invariant(curr.l <= n);
+
+                // get the node being processed
+                const size_t i = index(curr.l, curr.code);
+                Node const& node = mQuadTree[i];
+
+                // if the node is allocated, the whole tree below it is unavailable
+                if (node.isAllocated()) {
+                    // if a node is allocated it can't have children
+                    assert_invariant(!node.hasChildren());
+                    return QuadTree::TraversalResult::SKIP_CHILDREN;
+                }
+
+                // we're looking for the `deepest` unallocated node that has no children,
+                // up to the depth we're trying to allocate
+                if (curr.l > candidate.l && !node.hasChildren()) {
+                    candidate = curr;
+                    // Special case where we found a fitting node, just exit.
+                    if (curr.l == n) {
+                        return QuadTree::TraversalResult::EXIT;
+                    }
+                }
+
+                // We only want to find a fitting node that already has siblings, in order
+                // to accomplish a "best fit" allocation. So if we're a parent of a 'fitting'
+                // node and don't have children, we skip the children recursion.
+                if (curr.l == n - 1 && !node.hasChildren()) {
+                    return QuadTree::TraversalResult::SKIP_CHILDREN;
+                }
+
+                // continue going down the tree
+                return QuadTree::TraversalResult::RECURSE;
+            });
+
+    if (candidate.l >= 0) {
+        const size_t i = index(candidate.l, candidate.code);
+        Node& allocation = mQuadTree[i];
+
+        if (candidate.l == n) {
+            allocation.allocated = true;
+            if (UTILS_LIKELY(n > 0)) {
+                // the root doesn't have a parent
+                const size_t p = parent(candidate.l, candidate.code);
+                Node& parent = mQuadTree[p];
+                assert_invariant(!parent.isAllocated());
+                assert_invariant(parent.hasChildren());
+                assert_invariant(!parent.hasAllChildren());
+                parent.children++;
+            }
+        } else if (candidate.l < int8_t(QuadTree::height())) {
+            // we need to create the hierarchy down to the level we need
+            assert_invariant(!allocation.isAllocated());
+            assert_invariant(!allocation.hasChildren());
+
+            NodeId found{};
+            QuadTree::traverse(candidate.l, candidate.code,
+                    [this, n, &found](NodeId const& curr) -> QuadTree::TraversalResult {
+                        size_t i = index(curr.l, curr.code);
+                        Node& node = mQuadTree[i];
+                        if (curr.l == n) {
+                            found = curr;
+                            node.allocated = true;
+                            return QuadTree::TraversalResult::EXIT;
+                        }
+                        node.children++;
+                        return QuadTree::TraversalResult::RECURSE;
+                    });
+
+            candidate = found;
+        }
+    }
+    return candidate;
+}
+
+} // namespace filament

filament/src/AtlasAllocator.h

@@ -0,0 +1,95 @@
+/*
+ * Copyright (C) 2022 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef TNT_FILAMENT_ATLASALLOCATOR_H
+#define TNT_FILAMENT_ATLASALLOCATOR_H
+
+#include <utils/QuadTree.h>
+
+#include <filament/Viewport.h>
+
+class AtlasAllocator_AllocateFirstLevel_Test;
+class AtlasAllocator_AllocateSecondLevel_Test;
+class AtlasAllocator_AllocateMixed0_Test;
+class AtlasAllocator_AllocateMixed1_Test;
+class AtlasAllocator_AllocateMixed2_Test;
+
+namespace filament {
+
+/*
+ * A 2D allocator. Allocations must be square and have a power-of-two size.
+ * AtlasAllocator hard codes a depth of 4, that is only 4 allocation sizes are permitted.
+ * This doesn't actually allocate memory, just manages space within an abstract 2D image.
+ */
+class AtlasAllocator {
+
+    /*
+     * A quadtree is used to track the allocated regions. Each node of the quadtree stores a
+     * `Node` data structure below.
+     * The `Node` tracks if it is allocated as well as the number of children it has. It doesn't
+     * track which children though.
+     */
+    struct Node {
+        // whether this node is allocated. Implies no children.
+        constexpr bool isAllocated() const noexcept { return allocated; }
+        // whether this node has children. Implies it's not allocated.
+        constexpr bool hasChildren() const noexcept { return children != 0; }
+        // whether this node has all four children. Implies hasChildren().
+        constexpr bool hasAllChildren() const noexcept { return children == 4; }
+        bool allocated      : 1;    // true / false
+        uint8_t children    : 3;    // 0, 1, 2, 3, 4
+    };
+
+    static constexpr size_t QUAD_TREE_DEPTH = 4u;
+    using QuadTree = utils::QuadTreeArray<Node, QUAD_TREE_DEPTH>;
+    using NodeId = QuadTree::NodeId;
+
+public:
+    /*
+     * Create allocator and specify the maximum texture size. Must be a power of two.
+     * Allocations size allowed are the four power-of-two smaller or equal to this size.
+     */
+    explicit AtlasAllocator(size_t maxTextureSize = 1024) noexcept;
+
+    /*
+     * Allocates a square of size `textureSize`. Must be one of the power-of-two allowed
+     * (see above).
+     * Returns the location of the allocation within the maxTextureSize^2 square.
+     */
+    Viewport allocate(size_t textureSize) noexcept;
+
+    /*
+     * Frees all allocations and reset the maximum texture size.
+     */
+    void clear(size_t maxTextureSize = 1024) noexcept;
+
+private:
+    friend AtlasAllocator_AllocateFirstLevel_Test;
+    friend AtlasAllocator_AllocateSecondLevel_Test;
+    friend AtlasAllocator_AllocateMixed0_Test;
+    friend AtlasAllocator_AllocateMixed1_Test;
+    friend AtlasAllocator_AllocateMixed2_Test;
+
+    QuadTree::NodeId allocateInLayer(size_t n) noexcept;
+
+    // quad-tree array to store the allocated list
+    QuadTree mQuadTree{};
+    uint8_t mMaxTextureSizePot = 0;
+};
+
+} // namespace filament
+
+#endif // TNT_FILAMENT_ATLASALLOCATOR_H

filament/src/Engine.cpp

@@ -273,4 +273,10 @@ FeatureLevel Engine::getActiveFeatureLevel() const noexcept {
     return downcast(this)->getActiveFeatureLevel();
 }
 
+#if defined(__EMSCRIPTEN__)
+void Engine::resetBackendState() noexcept {
+    downcast(this)->resetBackendState();
+}
+#endif
+
 } // namespace filament

filament/src/MaterialInstance.cpp

@@ -292,4 +292,44 @@ MaterialInstance* MaterialInstance::duplicate(MaterialInstance const* other, con
 }
 
 
+float MaterialInstance::getMaskThreshold() const noexcept {
+    return downcast(this)->getMaskThreshold();
+}
+
+float MaterialInstance::getSpecularAntiAliasingVariance() const noexcept {
+    return downcast(this)->getSpecularAntiAliasingVariance();
+}
+
+float MaterialInstance::getSpecularAntiAliasingThreshold() const noexcept {
+    return downcast(this)->getSpecularAntiAliasingThreshold();
+}
+
+bool MaterialInstance::isDoubleSided() const noexcept {
+    return downcast(this)->isDoubleSided();
+}
+
+TransparencyMode MaterialInstance::getTransparencyMode() const noexcept {
+    return downcast(this)->getTransparencyMode();
+}
+
+CullingMode MaterialInstance::getCullingMode() const noexcept {
+    return downcast(this)->getCullingMode();
+}
+
+bool MaterialInstance::isColorWriteEnabled() const noexcept {
+    return downcast(this)->isColorWriteEnabled();
+}
+
+bool MaterialInstance::isDepthWriteEnabled() const noexcept {
+    return downcast(this)->isDepthWriteEnabled();
+}
+
+bool MaterialInstance::isStencilWriteEnabled() const noexcept {
+    return downcast(this)->isStencilWriteEnabled();
+}
+
+bool MaterialInstance::isDepthCullingEnabled() const noexcept {
+    return downcast(this)->isDepthCullingEnabled();
+}
+
 } // namespace filament

filament/src/PerShadowMapUniforms.cpp

@@ -0,0 +1,169 @@
+/*
+ * Copyright (C) 2022 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "PerShadowMapUniforms.h"
+
+#include "ShadowMapManager.h"
+
+#include "details/Camera.h"
+#include "details/Engine.h"
+
+#include <private/filament/EngineEnums.h>
+
+namespace filament {
+
+using namespace backend;
+using namespace math;
+
+PerShadowMapUniforms::PerShadowMapUniforms(FEngine& engine) noexcept {
+    DriverApi& driver = engine.getDriverApi();
+    mUniformBufferHandle = driver.createBufferObject(sizeof(PerViewUib),
+            BufferObjectBinding::UNIFORM, BufferUsage::DYNAMIC);
+}
+
+void PerShadowMapUniforms::terminate(DriverApi& driver) {
+    driver.destroyBufferObject(mUniformBufferHandle);
+}
+
+PerViewUib& PerShadowMapUniforms::edit(Transaction const& transaction) noexcept {
+    assert_invariant(transaction.uniforms);
+    return *transaction.uniforms;
+}
+
+void PerShadowMapUniforms::prepareCamera(Transaction const& transaction,
+        FEngine& engine, const CameraInfo& camera) noexcept {
+    mat4f const& viewFromWorld = camera.view;
+    mat4f const& worldFromView = camera.model;
+    mat4f const& clipFromView  = camera.projection;
+
+    const mat4f viewFromClip{ inverse((mat4)camera.projection) };
+    const mat4f clipFromWorld{ highPrecisionMultiply(clipFromView, viewFromWorld) };
+    const mat4f worldFromClip{ highPrecisionMultiply(worldFromView, viewFromClip) };
+
+    auto& s = edit(transaction);
+    s.viewFromWorldMatrix = viewFromWorld;    // view
+    s.worldFromViewMatrix = worldFromView;    // model
+    s.clipFromViewMatrix  = clipFromView;     // projection
+    s.viewFromClipMatrix  = viewFromClip;     // 1/projection
+    s.clipFromWorldMatrix = clipFromWorld;    // projection * view
+    s.worldFromClipMatrix = worldFromClip;    // 1/(projection * view)
+    s.clipTransform = camera.clipTransfrom;
+    s.cameraPosition = float3{ camera.getPosition() };
+    s.worldOffset = camera.getWorldOffset();
+    s.cameraFar = camera.zf;
+    s.oneOverFarMinusNear = 1.0f / (camera.zf - camera.zn);
+    s.nearOverFarMinusNear = camera.zn / (camera.zf - camera.zn);
+
+    // with a clip-space of [-w, w] ==> z' = -z
+    // with a clip-space of [0,  w] ==> z' = (w - z)/2
+    s.clipControl = engine.getDriverApi().getClipSpaceParams();
+}
+
+void PerShadowMapUniforms::prepareLodBias(Transaction const& transaction,
+        float bias) noexcept {
+    auto& s = edit(transaction);
+    s.lodBias = bias;
+}
+
+void PerShadowMapUniforms::prepareViewport(Transaction const& transaction,
+        const filament::Viewport& viewport,
+        uint32_t xoffset, uint32_t yoffset) noexcept {
+    const float w = float(viewport.width);
+    const float h = float(viewport.height);
+    auto& s = edit(transaction);
+    s.resolution = float4{ w, h, 1.0f / w, 1.0f / h };
+    s.origin = float2{ viewport.left, viewport.bottom };
+    s.offset = float2{ xoffset, yoffset };
+}
+
+void PerShadowMapUniforms::prepareTime(Transaction const& transaction,
+        FEngine& engine, math::float4 const& userTime) noexcept {
+    auto& s = edit(transaction);
+    const uint64_t oneSecondRemainder = engine.getEngineTime().count() % 1000000000;
+    const float fraction = float(double(oneSecondRemainder) / 1000000000.0);
+    s.time = fraction;
+    s.userTime = userTime;
+}
+
+void PerShadowMapUniforms::prepareDirectionalLight(Transaction const& transaction,
+        FEngine& engine,
+        float exposure,
+        float3 const& sceneSpaceDirection,
+        PerShadowMapUniforms::LightManagerInstance directionalLight) noexcept {
+    FLightManager& lcm = engine.getLightManager();
+    auto& s = edit(transaction);
+
+    const float3 l = -sceneSpaceDirection; // guaranteed normalized
+
+    if (directionalLight.isValid()) {
+        const float4 colorIntensity = {
+                lcm.getColor(directionalLight), lcm.getIntensity(directionalLight) * exposure };
+
+        s.lightDirection = l;
+        s.lightColorIntensity = colorIntensity;
+        s.lightChannels = lcm.getLightChannels(directionalLight);
+
+        const bool isSun = lcm.isSunLight(directionalLight);
+        // The last parameter must be < 0.0f for regular directional lights
+        float4 sun{ 0.0f, 0.0f, 0.0f, -1.0f };
+        if (UTILS_UNLIKELY(isSun && colorIntensity.w > 0.0f)) {
+            // currently we have only a single directional light, so it's probably likely that it's
+            // also the Sun. However, conceptually, most directional lights won't be sun lights.
+            float radius = lcm.getSunAngularRadius(directionalLight);
+            float haloSize = lcm.getSunHaloSize(directionalLight);
+            float haloFalloff = lcm.getSunHaloFalloff(directionalLight);
+            sun.x = std::cos(radius);
+            sun.y = std::sin(radius);
+            sun.z = 1.0f / (std::cos(radius * haloSize) - sun.x);
+            sun.w = haloFalloff;
+        }
+        s.sun = sun;
+    } else {
+        // Disable the sun if there's no directional light
+        s.sun = float4{ 0.0f, 0.0f, 0.0f, -1.0f };
+    }
+}
+
+void PerShadowMapUniforms::prepareShadowMapping(Transaction const& transaction,
+        bool highPrecision) noexcept {
+    auto& s = edit(transaction);
+    constexpr float low  = 5.54f; // ~ std::log(std::numeric_limits<math::half>::max()) * 0.5f;
+    constexpr float high = 42.0f; // ~ std::log(std::numeric_limits<float>::max()) * 0.5f;
+    s.vsmExponent = highPrecision ? high : low;
+}
+
+
+PerShadowMapUniforms::Transaction PerShadowMapUniforms::open(backend::DriverApi& driver) noexcept {
+    Transaction transaction;
+    // TODO: use out-of-line buffer if too large
+    transaction.uniforms = (PerViewUib *)driver.allocate(sizeof(PerViewUib));
+    assert_invariant(transaction.uniforms);
+    return transaction;
+}
+
+void PerShadowMapUniforms::commit(Transaction& transaction,
+        backend::DriverApi& driver) noexcept {
+    driver.updateBufferObject(mUniformBufferHandle, {
+        transaction.uniforms, sizeof(PerViewUib) }, 0);
+    transaction.uniforms = nullptr;
+}
+
+void PerShadowMapUniforms::bind(backend::DriverApi& driver) noexcept {
+    driver.bindUniformBuffer(+UniformBindingPoints::PER_VIEW, mUniformBufferHandle);
+}
+
+} // namespace filament
+

filament/src/PerShadowMapUniforms.h

@@ -0,0 +1,95 @@
+/*
+ * Copyright (C) 2022 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef TNT_FILAMENT_PERSHADOWMAPUNIFORMS_H
+#define TNT_FILAMENT_PERSHADOWMAPUNIFORMS_H
+
+#include <filament/Viewport.h>
+
+#include <private/filament/UibStructs.h>
+#include <private/backend/SamplerGroup.h>
+
+#include "TypedUniformBuffer.h"
+
+#include <backend/Handle.h>
+
+#include <utils/EntityInstance.h>
+
+#include <random>
+
+namespace filament {
+
+struct CameraInfo;
+
+class FEngine;
+class LightManager;
+
+/*
+ * PerShadowMapUniforms manages the UBO needed to generate our shadow maps. Internally it just
+ * holds onto a `PerViewUniform` UBO handle, but doesn't keep any shadow copy of it, instead it
+ * writes the data directly into the commandstream, for this reason partial update of the data
+ * is not possible.
+ */
+class PerShadowMapUniforms {
+
+    using LightManagerInstance = utils::EntityInstance<LightManager>;
+
+public:
+    class Transaction {
+        friend PerShadowMapUniforms;
+        PerViewUib* uniforms = nullptr;
+        Transaction() = default; // disallow creation by the caller
+    };
+
+    explicit PerShadowMapUniforms(FEngine& engine) noexcept;
+
+    void terminate(backend::DriverApi& driver);
+
+    static void prepareCamera(Transaction const& transaction,
+            FEngine& engine, const CameraInfo& camera) noexcept;
+
+    static void prepareLodBias(Transaction const& transaction,
+            float bias) noexcept;
+
+    static void prepareViewport(Transaction const& transaction,
+            const filament::Viewport& viewport, uint32_t xoffset, uint32_t yoffset) noexcept;
+
+    static void prepareTime(Transaction const& transaction,
+            FEngine& engine, math::float4 const& userTime) noexcept;
+
+    static void prepareShadowMapping(Transaction const& transaction,
+            bool highPrecision) noexcept;
+
+    static void prepareDirectionalLight(Transaction const& transaction,
+            FEngine& engine, float exposure,
+            math::float3 const& sceneSpaceDirection, LightManagerInstance instance) noexcept;
+
+    static Transaction open(backend::DriverApi& driver) noexcept;
+
+    // update local data into GPU UBO
+    void commit(Transaction& transaction, backend::DriverApi& driver) noexcept;
+
+    // bind this UBO
+    void bind(backend::DriverApi& driver) noexcept;
+
+private:
+    static PerViewUib& edit(Transaction const& transaction) noexcept;
+    backend::Handle<backend::HwBufferObject> mUniformBufferHandle;
+};
+
+} // namespace filament
+
+#endif //TNT_FILAMENT_PERSHADOWMAPUNIFORMS_H

filament/src/PerViewUniforms.cpp

@@ -34,15 +34,13 @@
 
 #include <math/mat4.h>
 
-
 namespace filament {
 
 using namespace backend;
 using namespace math;
 
 PerViewUniforms::PerViewUniforms(FEngine& engine) noexcept
-        : mEngine(engine),
-          mSamplers(PerViewSib::SAMPLER_COUNT) {
+        : mSamplers(PerViewSib::SAMPLER_COUNT) {
     DriverApi& driver = engine.getDriverApi();
 
     mSamplerGroupHandle = driver.createSamplerGroup(mSamplers.getSize());
@@ -50,10 +48,6 @@ PerViewUniforms::PerViewUniforms(FEngine& engine) noexcept
     mUniformBufferHandle = driver.createBufferObject(mUniforms.getSize(),
             BufferObjectBinding::UNIFORM, BufferUsage::DYNAMIC);
 
-    // with a clip-space of [-w, w] ==> z' = -z
-    // with a clip-space of [0,  w] ==> z' = (w - z)/2
-    mClipControl = driver.getClipSpaceParams();
-
     if (engine.getDFG().isValid()) {
         TextureSampler sampler(TextureSampler::MagFilter::LINEAR);
         mSamplers.setSampler(PerViewSib::IBL_DFG_LUT,
@@ -66,7 +60,7 @@ void PerViewUniforms::terminate(DriverApi& driver) {
     driver.destroySamplerGroup(mSamplerGroupHandle);
 }
 
-void PerViewUniforms::prepareCamera(const CameraInfo& camera) noexcept {
+void PerViewUniforms::prepareCamera(FEngine& engine, const CameraInfo& camera) noexcept {
     mat4f const& viewFromWorld = camera.view;
     mat4f const& worldFromView = camera.model;
     mat4f const& clipFromView  = camera.projection;
@@ -88,17 +82,15 @@ void PerViewUniforms::prepareCamera(const CameraInfo& camera) noexcept {
     s.cameraFar = camera.zf;
     s.oneOverFarMinusNear = 1.0f / (camera.zf - camera.zn);
     s.nearOverFarMinusNear = camera.zn / (camera.zf - camera.zn);
-    s.clipControl = mClipControl;
+
+    // with a clip-space of [-w, w] ==> z' = -z
+    // with a clip-space of [0,  w] ==> z' = (w - z)/2
+    s.clipControl = engine.getDriverApi().getClipSpaceParams();
 }
 
-void PerViewUniforms::prepareUpscaler(math::float2 scale,
-        DynamicResolutionOptions const& options) noexcept {
+void PerViewUniforms::prepareLodBias(float bias) noexcept {
     auto& s = mUniforms.edit();
-    if (options.quality >= QualityLevel::HIGH) {
-        s.lodBias = std::log2(std::min(scale.x, scale.y));
-    } else {
-        s.lodBias = 0.0f;
-    }
+    s.lodBias = bias;
 }
 
 void PerViewUniforms::prepareExposure(float ev100) noexcept {
@@ -118,17 +110,19 @@ void PerViewUniforms::prepareViewport(const filament::Viewport& viewport,
     s.offset = float2{ xoffset, yoffset };
 }
 
-void PerViewUniforms::prepareTime(math::float4 const& userTime) noexcept {
+void PerViewUniforms::prepareTime(FEngine& engine, math::float4 const& userTime) noexcept {
     auto& s = mUniforms.edit();
-    const uint64_t oneSecondRemainder = mEngine.getEngineTime().count() % 1000000000;
+    const uint64_t oneSecondRemainder = engine.getEngineTime().count() % 1000000000;
     const float fraction = float(double(oneSecondRemainder) / 1000000000.0);
     s.time = fraction;
     s.userTime = userTime;
 }
 
-void PerViewUniforms::prepareTemporalNoise(TemporalAntiAliasingOptions const& options) noexcept {
+void PerViewUniforms::prepareTemporalNoise(FEngine& engine,
+        TemporalAntiAliasingOptions const& options) noexcept {
+    std::uniform_real_distribution<float> uniformDistribution{ 0.0f, 1.0f };
     auto& s = mUniforms.edit();
-    const float temporalNoise = mUniformDistribution(mEngine.getRandomEngine());
+    const float temporalNoise = uniformDistribution(engine.getRandomEngine());
     s.temporalNoise = options.enabled ? temporalNoise : 0.0f;
 }
 
@@ -217,11 +211,11 @@ void PerViewUniforms::prepareStructure(Handle<HwTexture> structure) noexcept {
     mSamplers.setSampler(PerViewSib::STRUCTURE, { structure, {}});
 }
 
-void PerViewUniforms::prepareDirectionalLight(
+void PerViewUniforms::prepareDirectionalLight(FEngine& engine,
         float exposure,
         float3 const& sceneSpaceDirection,
         PerViewUniforms::LightManagerInstance directionalLight) noexcept {
-    FLightManager& lcm = mEngine.getLightManager();
+    FLightManager& lcm = engine.getLightManager();
     auto& s = mUniforms.edit();
 
     const float3 l = -sceneSpaceDirection; // guaranteed normalized
@@ -255,9 +249,8 @@ void PerViewUniforms::prepareDirectionalLight(
     }
 }
 
-void PerViewUniforms::prepareAmbientLight(FIndirectLight const& ibl,
+void PerViewUniforms::prepareAmbientLight(FEngine& engine, FIndirectLight const& ibl,
         float intensity, float exposure) noexcept {
-    auto& engine = mEngine;
     auto& s = mUniforms.edit();
 
     // Set up uniforms and sampler for the IBL, guaranteed to be non-null at this point.

filament/src/PerViewUniforms.h

@@ -48,6 +48,11 @@ class FIndirectLight;
 class Froxelizer;
 class LightManager;
 
+/*
+ * PerViewUniforms manages the UBO and samplers needed to render the color passes. Internally it
+ * holds onto handles for the PER_VIEW UBO and SamplerGroup. This class maintains a shadow copy
+ * of the UBO/sampler data, so it is possible to partially update it between commits.
+ */
 class PerViewUniforms {
 
     using LightManagerInstance = utils::EntityInstance<LightManager>;
@@ -63,8 +68,8 @@ public:
 
     void terminate(backend::DriverApi& driver);
 
-    void prepareCamera(const CameraInfo& camera) noexcept;
-    void prepareUpscaler(math::float2 scale, DynamicResolutionOptions const& options) noexcept;
+    void prepareCamera(FEngine& engine, const CameraInfo& camera) noexcept;
+    void prepareLodBias(float bias) noexcept;
 
     /*
      * @param viewport  viewport (should be same as RenderPassParams::viewport)
@@ -75,8 +80,8 @@ public:
      */
     void prepareViewport(const filament::Viewport& viewport, uint32_t xoffset, uint32_t yoffset) noexcept;
 
-    void prepareTime(math::float4 const& userTime) noexcept;
-    void prepareTemporalNoise(TemporalAntiAliasingOptions const& options) noexcept;
+    void prepareTime(FEngine& engine, math::float4 const& userTime) noexcept;
+    void prepareTemporalNoise(FEngine& engine, TemporalAntiAliasingOptions const& options) noexcept;
     void prepareExposure(float ev100) noexcept;
     void prepareFog(math::float3 const& cameraPosition, FogOptions const& options) noexcept;
     void prepareStructure(TextureHandle structure) noexcept;
@@ -95,10 +100,11 @@ public:
 
     void prepareShadowMapping(bool highPrecision) noexcept;
 
-    void prepareDirectionalLight(float exposure,
+    void prepareDirectionalLight(FEngine& engine, float exposure,
             math::float3 const& sceneSpaceDirection, LightManagerInstance instance) noexcept;
 
-    void prepareAmbientLight(FIndirectLight const& ibl, float intensity, float exposure) noexcept;
+    void prepareAmbientLight(FEngine& engine,
+            FIndirectLight const& ibl, float intensity, float exposure) noexcept;
 
     void prepareDynamicLights(Froxelizer& froxelizer) noexcept;
 
@@ -126,13 +132,10 @@ public:
     void unbindSamplers() noexcept;
 
 private:
-    FEngine& mEngine;
-    math::float2 mClipControl{};
     TypedUniformBuffer<PerViewUib> mUniforms;
     backend::SamplerGroup mSamplers;
     backend::Handle<backend::HwBufferObject> mUniformBufferHandle;
     backend::Handle<backend::HwSamplerGroup> mSamplerGroupHandle;
-    std::uniform_real_distribution<float> mUniformDistribution{ 0.0f, 1.0f };
     static void prepareShadowSampling(PerViewUib& uniforms,
             ShadowMappingUniforms const& shadowMappingUniforms) noexcept;
 };

filament/src/PostProcessManager.cpp

@@ -399,9 +399,9 @@ PostProcessManager::StructurePassOutput PostProcessManager::structure(FrameGraph
                 auto out = resources.getRenderPassInfo();
                 renderPass.setRenderFlags(structureRenderFlags);
                 renderPass.setVariant(structureVariant);
-                renderPass.appendCommands(RenderPass::CommandTypeFlags::SSAO);
-                renderPass.sortCommands();
-                renderPass.execute(resources.getPassName(), out.target, out.params);
+                renderPass.appendCommands(mEngine, RenderPass::CommandTypeFlags::SSAO);
+                renderPass.sortCommands(mEngine);
+                renderPass.execute(mEngine, resources.getPassName(), out.target, out.params);
             });
 
     auto depth = structurePass->depth;
@@ -544,9 +544,9 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::ssr(FrameGraph& fg,
                 // the SCREEN_SPACE ReflectionMode.
                 renderPass.setVariant(Variant{Variant::SPECIAL_SSR});
                 // generate all our drawing commands, except blended objects.
-                renderPass.appendCommands(RenderPass::CommandTypeFlags::SCREEN_SPACE_REFLECTIONS);
-                renderPass.sortCommands();
-                renderPass.execute(resources.getPassName(), out.target, out.params);
+                renderPass.appendCommands(mEngine, RenderPass::CommandTypeFlags::SCREEN_SPACE_REFLECTIONS);
+                renderPass.sortCommands(mEngine);
+                renderPass.execute(mEngine, resources.getPassName(), out.target, out.params);
             });
 
     return ssrPass->reflections;
@@ -2476,7 +2476,7 @@ void PostProcessManager::prepareTaa(FrameGraph& fg, filament::Viewport const& sv
 
     fg.addTrivialSideEffectPass("Jitter Camera",
             [=, &uniforms] (DriverApi& driver) {
-        uniforms.prepareCamera(*inoutCameraInfo);
+        uniforms.prepareCamera(mEngine, *inoutCameraInfo);
         uniforms.commit(driver);
     });
 }

filament/src/RenderPass.cpp

@@ -39,7 +39,7 @@ using namespace backend;
 
 RenderPass::RenderPass(FEngine& engine,
         RenderPass::Arena& arena) noexcept
-        : mEngine(engine), mCommandArena(arena),
+        : mCommandArena(arena),
           mCustomCommands(engine.getPerRenderPassAllocator()) {
 }
 
@@ -79,26 +79,13 @@ void RenderPass::setCamera(const CameraInfo& camera) noexcept {
     mCameraForwardVector = camera.getForwardVector();
 }
 
-void RenderPass::overridePolygonOffset(backend::PolygonOffset const* polygonOffset) noexcept {
-    if ((mPolygonOffsetOverride = (polygonOffset != nullptr))) {
-        mPolygonOffset = *polygonOffset;
-    }
-}
-
-
 void RenderPass::setScissorViewport(backend::Viewport viewport) noexcept {
     assert_invariant(viewport.width  <= std::numeric_limits<int32_t>::max());
     assert_invariant(viewport.height <= std::numeric_limits<int32_t>::max());
     mScissorViewport = viewport;
 }
 
-void RenderPass::overrideScissor(backend::Viewport const* scissor) noexcept {
-    if ((mScissorOverride = (scissor != nullptr))) {
-        mScissor = *scissor;
-    }
-}
-
-void RenderPass::appendCommands(CommandTypeFlags const commandTypeFlags) noexcept {
+void RenderPass::appendCommands(FEngine& engine, CommandTypeFlags const commandTypeFlags) noexcept {
     SYSTRACE_CONTEXT();
 
     assert_invariant(mRenderableSoa);
@@ -110,7 +97,6 @@ void RenderPass::appendCommands(CommandTypeFlags const commandTypeFlags) noexcep
         return;
     }
 
-    FEngine& engine = mEngine;
     JobSystem& js = engine.getJobSystem();
     const RenderFlags renderFlags = mFlags;
     const Variant variant = mVariant;
@@ -179,7 +165,7 @@ void RenderPass::appendCustomCommand(Pass pass, CustomCommand custom, uint32_t o
     curr->key = cmd;
 }
 
-void RenderPass::sortCommands() noexcept {
+void RenderPass::sortCommands(FEngine& engine) noexcept {
     SYSTRACE_NAME("sort and trim commands");
 
     std::sort(mCommandBegin, mCommandEnd);
@@ -192,12 +178,28 @@ void RenderPass::sortCommands() noexcept {
 
     resize(uint32_t(last - mCommandBegin));
 
-    if (mEngine.isAutomaticInstancingEnabled()) {
-        instanceify();
+    if (engine.isAutomaticInstancingEnabled()) {
+        instanceify(engine);
     }
 }
 
-void RenderPass::instanceify() noexcept {
+void RenderPass::execute(FEngine& engine, const char* name,
+        backend::Handle<backend::HwRenderTarget> renderTarget,
+        backend::RenderPassParams params) const noexcept {
+
+    DriverApi& driver = engine.getDriverApi();
+
+    // this is a good time to flush the CommandStream, because we're about to potentially
+    // output a lot of commands. This guarantees here that we have at least
+    // FILAMENT_MIN_COMMAND_BUFFERS_SIZE_IN_MB bytes (1MiB by default).
+    engine.flush();
+
+    driver.beginRenderPass(renderTarget, params);
+    getExecutor().execute(engine, name);
+    driver.endRenderPass();
+}
+
+void RenderPass::instanceify(FEngine& engine) noexcept {
     SYSTRACE_NAME("instanceify");
 
     // instanceify works by scanning the **sorted** command stream, looking for repeat draw
@@ -284,7 +286,7 @@ void RenderPass::instanceify() noexcept {
 #endif
 
         // we have instanced primitives
-        DriverApi& driver = mEngine.getDriverApi();
+        DriverApi& driver = engine.getDriverApi();
 
         // TODO: maybe use a pool? so we can reuse the buffer.
         // create a ubo to hold the instanced primitive data
@@ -356,8 +358,8 @@ void RenderPass::setupColorCommand(Command& cmdDraw, Variant variant,
 
     cmdDraw.primitive.rasterState.inverseFrontFaces = inverseFrontFaces;
     cmdDraw.primitive.rasterState.culling = mi->getCullingMode();
-    cmdDraw.primitive.rasterState.colorWrite = mi->getColorWrite();
-    cmdDraw.primitive.rasterState.depthWrite = mi->getDepthWrite();
+    cmdDraw.primitive.rasterState.colorWrite = mi->isColorWriteEnabled();
+    cmdDraw.primitive.rasterState.depthWrite = mi->isDepthWriteEnabled();
     cmdDraw.primitive.rasterState.depthFunc = mi->getDepthFunc();
     cmdDraw.primitive.mi = mi;
     cmdDraw.primitive.materialVariant = variant;
@@ -542,11 +544,6 @@ RenderPass::Command* RenderPass::generateCommandsImpl(uint32_t extraFlags,
          */
         for (size_t pi = 0, c = primitives.size(); pi < c; ++pi) {
             auto const& primitive = primitives[pi];
-            // handle the case where this primitive is empty / no-op
-            if (primitive.getPrimitiveType() == PrimitiveType::NONE) {
-                continue;
-            }
-
             auto const& morphTargets = morphing.targets[pi];
             FMaterialInstance const* const mi = primitive.getMaterialInstance();
             FMaterial const* const ma = mi->getMaterial();
@@ -609,6 +606,9 @@ RenderPass::Command* RenderPass::generateCommandsImpl(uint32_t extraFlags,
                     // cancel command if asked to filter translucent objects
                     key |= select(filterTranslucentObjects);
 
+                    // cancel command if both front and back faces are culled
+                    key |= select(mi->getCullingMode() == CullingMode::FRONT_AND_BACK);
+
                     *curr = cmdColor;
                     curr->key = key;
                     ++curr;
@@ -635,6 +635,10 @@ RenderPass::Command* RenderPass::generateCommandsImpl(uint32_t extraFlags,
                 }
 
                 *curr = cmdColor;
+
+                // cancel command if both front and back faces are culled
+                curr->key |= select(mi->getCullingMode() == CullingMode::FRONT_AND_BACK);
+
                 ++curr;
             }
 
@@ -659,11 +663,16 @@ RenderPass::Command* RenderPass::generateCommandsImpl(uint32_t extraFlags,
 
                 // FIXME: should writeDepthForShadowCasters take precedence over mi->getDepthWrite()?
                 cmdDepth.primitive.rasterState.depthWrite = (1 // only keep bit 0
-                        & (mi->getDepthWrite() | (mode == TransparencyMode::TWO_PASSES_ONE_SIDE))
+                        & (mi->isDepthWriteEnabled() | (mode == TransparencyMode::TWO_PASSES_ONE_SIDE))
                         & !(filterTranslucentObjects & translucent)
                         & !(depthFilterAlphaMaskedObjects & rs.alphaToCoverage))
                             | writeDepthForShadowCasters;
+
                 *curr = cmdDepth;
+
+                // cancel command if both front and back faces are culled
+                curr->key |= select(mi->getCullingMode() == CullingMode::FRONT_AND_BACK);
+
                 ++curr;
             }
         }
@@ -686,25 +695,25 @@ void RenderPass::updateSummedPrimitiveCounts(
 
 // ------------------------------------------------------------------------------------------------
 
-void RenderPass::Executor::execute(const char* name,
-        backend::Handle<backend::HwRenderTarget> renderTarget,
-        backend::RenderPassParams const& params) const noexcept {
-    FEngine& engine = mEngine;
-    DriverApi& driver = engine.getDriverApi();
+void RenderPass::Executor::overridePolygonOffset(backend::PolygonOffset const* polygonOffset) noexcept {
+    if ((mPolygonOffsetOverride = (polygonOffset != nullptr))) {
+        mPolygonOffset = *polygonOffset;
+    }
+}
 
-    // this is a good time to flush the CommandStream, because we're about to potentially
-    // output a lot of commands. This guarantees here that we have at least
-    // FILAMENT_MIN_COMMAND_BUFFERS_SIZE_IN_MB bytes (1MiB by default).
-    engine.flush();
+void RenderPass::Executor::overrideScissor(backend::Viewport const* scissor) noexcept {
+    if ((mScissorOverride = (scissor != nullptr))) {
+        mScissor = *scissor;
+    }
+}
 
-    driver.beginRenderPass(renderTarget, params);
-    recordDriverCommands(engine, driver, mBegin, mEnd, params.readOnlyDepthStencil);
-    driver.endRenderPass();
+void RenderPass::Executor::execute(FEngine& engine, const char* name) const noexcept {
+    execute(engine.getDriverApi(), mCommands.begin(), mCommands.end());
 }
 
 UTILS_NOINLINE // no need to be inlined
-void RenderPass::Executor::recordDriverCommands(FEngine& engine, backend::DriverApi& driver,
-        const Command* first, const Command* last, uint16_t readOnlyDepthStencil) const noexcept {
+void RenderPass::Executor::execute(backend::DriverApi& driver,
+        const Command* first, const Command* last) const noexcept {
     SYSTRACE_CALL();
 
     if (first != last) {
@@ -724,7 +733,7 @@ void RenderPass::Executor::recordDriverCommands(FEngine& engine, backend::Driver
         Handle<HwBufferObject> uboHandle = mUboHandle;
         FMaterialInstance const* UTILS_RESTRICT mi = nullptr;
         FMaterial const* UTILS_RESTRICT ma = nullptr;
-        auto customCommands = mCustomCommands.data();
+        auto const* UTILS_RESTRICT pCustomCommands = mCustomCommands.data();
 
         first--;
         while (++first != last) {
@@ -737,7 +746,7 @@ void RenderPass::Executor::recordDriverCommands(FEngine& engine, backend::Driver
             if (UTILS_UNLIKELY((first->key & CUSTOM_MASK) != uint64_t(CustomCommand::PASS))) {
                 uint32_t index = (first->key & CUSTOM_INDEX_MASK) >> CUSTOM_INDEX_SHIFT;
                 assert_invariant(index < mCustomCommands.size());
-                customCommands[index]();
+                pCustomCommands[index]();
                 continue;
             }
 
@@ -745,11 +754,6 @@ void RenderPass::Executor::recordDriverCommands(FEngine& engine, backend::Driver
             const PrimitiveInfo info = first->primitive;
             pipeline.rasterState = info.rasterState;
 
-#ifndef NDEBUG
-            const bool readOnlyDepthGuaranteed = readOnlyDepthStencil & RenderPassParams::READONLY_DEPTH;
-            assert_invariant(!readOnlyDepthGuaranteed || !pipeline.rasterState.depthWrite);
-#endif
-
             if (UTILS_UNLIKELY(mi != info.mi)) {
                 // this is always taken the first time
                 mi = info.mi;
@@ -831,15 +835,13 @@ void RenderPass::Executor::recordDriverCommands(FEngine& engine, backend::Driver
 // ------------------------------------------------------------------------------------------------
 
 RenderPass::Executor::Executor(RenderPass const* pass, Command const* b, Command const* e) noexcept
-        : mEngine(pass->mEngine), mBegin(b), mEnd(e),
-          mCustomCommands(pass->mCustomCommands),
+        : mCommands(b, e),
+          mCustomCommands(pass->mCustomCommands.data(), pass->mCustomCommands.size()),
           mUboHandle(pass->mUboHandle),
           mInstancedUboHandle(pass->mInstancedUboHandle),
-          mPolygonOffset(pass->mPolygonOffset),
-          mScissor(pass->mScissor),
           mScissorViewport(pass->mScissorViewport),
-          mPolygonOffsetOverride(pass->mPolygonOffsetOverride),
-          mScissorOverride(pass->mScissorOverride) {
+          mPolygonOffsetOverride(false),
+          mScissorOverride(false) {
     assert_invariant(b >= pass->begin());
     assert_invariant(e <= pass->end());
 }

filament/src/RenderPass.h

@@ -276,15 +276,9 @@ public:
     // allocated commands ARE NOT freed, they're owned by the Arena
     ~RenderPass() noexcept;
 
-    // if non-null, overrides the material's polygon offset
-    void overridePolygonOffset(backend::PolygonOffset const* polygonOffset) noexcept;
-
     // a box that both offsets the viewport and clips it
     void setScissorViewport(backend::Viewport viewport) noexcept;
 
-    // if non-null, overrides the material's scissor
-    void overrideScissor(backend::Viewport const* scissor) noexcept;
-
     // specifies the geometry to generate commands for
     void setGeometry(FScene::RenderableSoa const& soa, utils::Range<uint32_t> vr,
             backend::Handle<backend::HwBufferObject> uboHandle) noexcept;
@@ -310,58 +304,69 @@ public:
 
     // This is the main function of this class, this appends commands to the pass using
     // the current camera, geometry and flags set. This can be called multiple times if needed.
-    void appendCommands(CommandTypeFlags commandTypeFlags) noexcept;
+    void appendCommands(FEngine& engine, CommandTypeFlags commandTypeFlags) noexcept;
 
-    // sorts and instancify commands then trims sentinels
-    void sortCommands() noexcept;
+    // sorts and instanceify commands then trims sentinels
+    void sortCommands(FEngine& engine) noexcept;
 
     // Helper to execute all the commands generated by this RenderPass
-    void execute(const char* name,
+    void execute(FEngine& engine, const char* name,
             backend::Handle<backend::HwRenderTarget> renderTarget,
-            backend::RenderPassParams params) const noexcept {
-        getExecutor().execute(name, renderTarget, params);
-    }
+            backend::RenderPassParams params) const noexcept;
 
     /*
      * Executor holds the range of commands to execute for a given pass
      */
     class Executor {
         using CustomCommandFn = std::function<void()>;
-        using CustomCommandVector = std::vector<CustomCommandFn,
-                utils::STLAllocator<CustomCommandFn, LinearAllocatorArena>>;
-
         friend class RenderPass;
-        FEngine& mEngine;
-        Command const* mBegin;
-        Command const* mEnd;
-        const CustomCommandVector mCustomCommands;
-        const backend::Handle<backend::HwBufferObject> mUboHandle;
-        const backend::Handle<backend::HwBufferObject> mInstancedUboHandle;
-        const backend::PolygonOffset mPolygonOffset;
-        const backend::Viewport mScissor;
-        const backend::Viewport mScissorViewport;
-        const bool mPolygonOffsetOverride : 1;
-        const bool mScissorOverride : 1;
+
+        // these fields are constant after creation
+        utils::Slice<Command> mCommands;
+        utils::Slice<CustomCommandFn> mCustomCommands;
+        backend::Handle<backend::HwBufferObject> mUboHandle;
+        backend::Handle<backend::HwBufferObject> mInstancedUboHandle;
+        backend::Viewport mScissorViewport;
+
+        backend::Viewport mScissor{};            // value of scissor override
+        backend::PolygonOffset mPolygonOffset{}; // value of the override
+        bool mPolygonOffsetOverride : 1;         // whether to override the polygon offset setting
+        bool mScissorOverride : 1;               // whether to override the polygon offset setting
 
         Executor(RenderPass const* pass, Command const* b, Command const* e) noexcept;
 
-        void recordDriverCommands(FEngine& engine, backend::DriverApi& driver,
-                const Command* first, const Command* last, uint16_t readOnlyDepthStencil) const noexcept;
+        void execute(backend::DriverApi& driver,
+                const Command* first, const Command* last) const noexcept;
 
     public:
+        Executor() = default;
         Executor(Executor const& rhs);
+        Executor& operator=(Executor const& rhs) = default;
         ~Executor() noexcept;
-        void execute(const char* name,
-                backend::Handle<backend::HwRenderTarget> renderTarget,
-                backend::RenderPassParams const& params) const noexcept;
+
+        // if non-null, overrides the material's polygon offset
+        void overridePolygonOffset(backend::PolygonOffset const* polygonOffset) noexcept;
+
+        // if non-null, overrides the material's scissor
+        void overrideScissor(backend::Viewport const* scissor) noexcept;
+
+        void execute(FEngine& engine, const char* name) const noexcept;
     };
 
     // returns a new executor for this pass
+    Executor getExecutor() {
+        return { this, mCommandBegin, mCommandEnd };
+    }
+
     Executor getExecutor() const {
         return { this, mCommandBegin, mCommandEnd };
     }
 
     // returns a new executor for this pass with a custom range
+    Executor getExecutor(Command const* b, Command const* e) {
+        return { this, b, e };
+    }
+
     Executor getExecutor(Command const* b, Command const* e) const {
         return { this, b, e };
     }
@@ -376,7 +381,7 @@ private:
 
     Command* append(size_t count) noexcept;
     void resize(size_t count) noexcept;
-    void instanceify() noexcept;
+    void instanceify(FEngine& engine) noexcept;
 
     // on 64-bits systems, we process batches of 256 (64 bytes) cache-lines, or 512 (32 bytes) commands
     // on 32-bits systems, we process batches of 512 (32 bytes) cache-lines, or 512 (32 bytes) commands
@@ -406,7 +411,6 @@ private:
             FScene::RenderableSoa& renderableData, utils::Range<uint32_t> vr) noexcept;
 
     // a reference to the Engine, mostly to get to things like JobSystem
-    FEngine& mEngine;
 
     // Arena where all Commands are allocated. The Arena owns the commands.
     Arena& mCommandArena;
@@ -440,24 +444,14 @@ private:
     // Additional visibility mask
     FScene::VisibleMaskType mVisibilityMask = std::numeric_limits<FScene::VisibleMaskType>::max();
 
-    // whether to override the polygon offset setting
-    bool mPolygonOffsetOverride = false;
-
-    // whether to override the polygon offset setting
-    bool mScissorOverride = false;
-
-    // value of the override
-    backend::PolygonOffset mPolygonOffset{};
-
-    // value of scissor override
-    backend::Viewport mScissor{};
-
     backend::Viewport mScissorViewport{ 0, 0,
             std::numeric_limits<int32_t>::max(),
             std::numeric_limits<int32_t>::max() };
 
     // a vector for our custom commands
-    mutable Executor::CustomCommandVector mCustomCommands;
+    using CustomCommandVector = std::vector<Executor::CustomCommandFn,
+            utils::STLAllocator<Executor::CustomCommandFn, LinearAllocatorArena>>;
+    mutable CustomCommandVector mCustomCommands;
 };
 
 } // namespace filament

filament/src/RenderPrimitive.h

@@ -71,7 +71,7 @@ private:
     AttributeBitset mEnabledAttributes = {};
     uint16_t mBlendOrder = 0;
     bool mGlobalBlendOrderEnabled = false;
-    backend::PrimitiveType mPrimitiveType = backend::PrimitiveType::NONE;
+    backend::PrimitiveType mPrimitiveType = backend::PrimitiveType::TRIANGLES;
     UTILS_UNUSED uint8_t reserved[4] = {};
 };
 

filament/src/RendererUtils.cpp

@@ -173,7 +173,7 @@ FrameGraphId<FrameGraphTexture> RendererUtils::colorPass(
                 view.prepareSSAO(data.ssao ?
                         resources.getTexture(data.ssao) : engine.getOneTextureArray());
 
-                view.prepareShadowMap(view.getVsmShadowOptions().highPrecision);
+                view.prepareShadowMapping(view.getVsmShadowOptions().highPrecision);
 
                 // set shadow sampler
                 view.prepareShadow(data.shadows ?
@@ -200,6 +200,7 @@ FrameGraphId<FrameGraphTexture> RendererUtils::colorPass(
 
                 view.prepareViewport(static_cast<filament::Viewport&>(out.params.viewport),
                         config.xoffset, config.yoffset);
+
                 view.commitUniforms(driver);
 
                 out.params.clearColor = data.clearColor;
@@ -217,7 +218,14 @@ FrameGraphId<FrameGraphTexture> RendererUtils::colorPass(
                 if (colorGradingConfig.asSubpass || colorGradingConfig.customResolve) {
                     out.params.subpassMask = 1;
                 }
-                passExecutor.execute(resources.getPassName(), out.target, out.params);
+
+                // this is a good time to flush the CommandStream, because we're about to potentially
+                // output a lot of commands. This guarantees here that we have at least
+                // FILAMENT_MIN_COMMAND_BUFFERS_SIZE_IN_MB bytes (1MiB by default).
+                engine.flush();
+                driver.beginRenderPass(out.target, out.params);
+                passExecutor.execute(engine, resources.getPassName());
+                driver.endRenderPass();
 
                 // color pass is typically heavy, and we don't have much CPU work left after
                 // this point, so flushing now allows us to start the GPU earlier and reduce

filament/src/ShadowMap.cpp

@@ -41,8 +41,10 @@ using namespace backend;
 static constexpr bool USE_DEPTH_CLAMP = false;
 
 ShadowMap::ShadowMap(FEngine& engine) noexcept
-        : mShadowType(ShadowType::DIRECTIONAL),
-          mHasVisibleShadows(false) {
+        : mPerShadowMapUniforms(engine),
+          mShadowType(ShadowType::DIRECTIONAL),
+          mHasVisibleShadows(false),
+          mFace(0) {
     Entity entities[2];
     engine.getEntityManager().create(2, entities);
     mCamera = engine.createCamera(entities[0]);
@@ -60,16 +62,19 @@ void ShadowMap::terminate(FEngine& engine) {
         engine.destroyCameraComponent(e);
     }
     engine.getEntityManager().destroy(sizeof(entities) / sizeof(Entity), entities);
+    mPerShadowMapUniforms.terminate(engine.getDriverApi());
 }
 
 ShadowMap::~ShadowMap() = default;
 
-void ShadowMap::initialize(size_t lightIndex, ShadowType shadowType, uint16_t shadowIndex,
+void ShadowMap::initialize(size_t lightIndex, ShadowType shadowType,
+        uint16_t shadowIndex, uint8_t face,
         LightManager::ShadowOptions const* options) {
     mLightIndex = lightIndex;
     mShadowIndex = shadowIndex;
     mOptions = options;
     mShadowType = shadowType;
+    mFace = face;
 }
 
 mat4f ShadowMap::getDirectionalLightViewMatrix(float3 direction, float3 position) noexcept {
@@ -1238,4 +1243,67 @@ void ShadowMap::updateSceneInfoSpot(mat4f const& Mv, FScene const& scene,
     );
 }
 
+filament::Viewport ShadowMap::getViewport() const noexcept {
+    // We set a viewport with a 1-texel border for when we index outside the
+    // texture.
+    // DON'T CHANGE this unless ShadowMap::getTextureCoordsMapping() is updated too.
+    // see: ShadowMap::getTextureCoordsMapping()
+    //
+    // For floating-point depth textures, the 1-texel border could be set to
+    // FLOAT_MAX to avoid clamping in the shadow shader (see sampleDepth inside
+    // shadowing.fs). Unfortunately, the APIs don't seem let us clear depth
+    // attachments to anything greater than 1.0, so we'd need a way to do this other
+    // than clearing.
+    const uint32_t dim = mOptions->mapSize;
+    if (isPointShadow()) {
+        // for point-light we don't have a border
+        return { 0, 0, dim, dim };
+    } else {
+        return { 1, 1, dim - 2, dim - 2 };
+    }
+}
+
+// ------------------------------------------------------------------------------------------------
+
+void ShadowMap::prepareCamera(Transaction const& transaction,
+        FEngine& engine, const CameraInfo& cameraInfo) noexcept {
+    PerShadowMapUniforms::prepareCamera(transaction, engine, cameraInfo);
+    PerShadowMapUniforms::prepareLodBias(transaction, 0.0f);
+}
+
+void ShadowMap::prepareViewport(Transaction const& transaction,
+        const filament::Viewport& viewport) noexcept {
+    PerShadowMapUniforms::prepareViewport(transaction, viewport, 0, 0);
+}
+
+void ShadowMap::prepareTime(Transaction const& transaction,
+        FEngine& engine, math::float4 const& userTime) noexcept {
+    PerShadowMapUniforms::prepareTime(transaction, engine, userTime);
+}
+
+void ShadowMap::prepareDirectionalLight(Transaction const& transaction,
+        FEngine& engine, math::float3 const& sceneSpaceDirection,
+        LightManager::Instance instance) noexcept {
+    PerShadowMapUniforms::prepareDirectionalLight(transaction,
+            engine, 1.0f, sceneSpaceDirection, instance);
+}
+
+void ShadowMap::prepareShadowMapping(Transaction const& transaction,
+        bool highPrecision) noexcept {
+    PerShadowMapUniforms::prepareShadowMapping(transaction, highPrecision);
+}
+
+PerShadowMapUniforms::Transaction ShadowMap::open(DriverApi& driver) noexcept {
+    return PerShadowMapUniforms::open(driver);
+}
+
+void ShadowMap::commit(Transaction& transaction,
+        backend::DriverApi& driver) const noexcept {
+    mPerShadowMapUniforms.commit(transaction, driver);
+}
+
+void ShadowMap::bindPerViewUniformsAndSamplers(backend::DriverApi& driver) const noexcept {
+    mPerShadowMapUniforms.bind(driver);
+}
+
 } // namespace filament

filament/src/ShadowMap.h

@@ -19,12 +19,14 @@
 
 #include "components/LightManager.h"
 
+#include "PerShadowMapUniforms.h"
+
 #include "details/Camera.h"
 #include "details/Scene.h"
 
-#include "backend/DriverApiForward.h"
-#include "private/backend/SamplerGroup.h"
-#include "math/mat4.h"
+#include "private/backend/DriverApi.h"
+
+#include <filament/Viewport.h>
 
 #include <math/mat4.h>
 #include <math/vec4.h>
@@ -122,7 +124,7 @@ public:
     static math::mat4f getPointLightViewMatrix(backend::TextureCubemapFace face,
             math::float3 position) noexcept;
 
-    void initialize(size_t lightIndex, ShadowType shadowType, uint16_t shadowIndex,
+    void initialize(size_t lightIndex, ShadowType shadowType, uint16_t shadowIndex, uint8_t face,
             LightManager::ShadowOptions const* options);
 
     struct ShaderParameters {
@@ -172,6 +174,8 @@ public:
     static void updateSceneInfoSpot(const math::mat4f& Mv, FScene const& scene,
             SceneInfo& sceneInfo);
 
+    filament::Viewport getViewport() const noexcept;
+
     LightManager::ShadowOptions const* getShadowOptions() const noexcept { return mOptions; }
     size_t getLightIndex() const { return mLightIndex; }
     uint16_t getShadowIndex() const { return mShadowIndex; }
@@ -181,6 +185,25 @@ public:
     bool isSpotShadow() const noexcept { return mShadowType == ShadowType::SPOT; }
     bool isPointShadow() const noexcept { return mShadowType == ShadowType::POINT; }
     ShadowType getShadowType() const noexcept { return mShadowType; }
+    uint8_t getFace() const noexcept { return mFace; }
+
+    using Transaction = PerShadowMapUniforms::Transaction;
+
+    static void prepareCamera(Transaction const& transaction,
+            FEngine& engine, const CameraInfo& cameraInfo) noexcept;
+    static void prepareViewport(Transaction const& transaction,
+            const filament::Viewport& viewport) noexcept;
+    static void prepareTime(Transaction const& transaction,
+            FEngine& engine, math::float4 const& userTime) noexcept;
+    static void prepareDirectionalLight(Transaction const& transaction,
+            FEngine& engine, math::float3 const& sceneSpaceDirection,
+            LightManager::Instance instance) noexcept;
+    static void prepareShadowMapping(Transaction const& transaction,
+            bool highPrecision) noexcept;
+    static PerShadowMapUniforms::Transaction open(backend::DriverApi& driver) noexcept;
+    void commit(Transaction& transaction,
+            backend::DriverApi& driver) const noexcept;
+    void bindPerViewUniformsAndSamplers(backend::DriverApi& driver) const noexcept;
 
 private:
     struct Segment {
@@ -278,8 +301,10 @@ private:
             { 2, 6, 7, 3 },  // top
     };
 
-    FCamera* mCamera = nullptr;                 //  8
-    FCamera* mDebugCamera = nullptr;            //  8
+    mutable PerShadowMapUniforms mPerShadowMapUniforms;                     // 4
+
+    FCamera* mCamera = nullptr;                                             //  8
+    FCamera* mDebugCamera = nullptr;                                        //  8
 
     // The data below technically belongs to ShadowMapManager, but it simplifies allocations
     // to store it here. This data is always associated with this shadow map anyway.
@@ -287,8 +312,9 @@ private:
     uint32_t mLightIndex = 0;   // which light are we shadowing             // 4
     uint16_t mShadowIndex = 0;  // our index in the shadowMap vector        // 2
     uint8_t mLayer = 0;         // our layer in the shadowMap texture       // 1
-    ShadowType mShadowType : 2;
-    bool mHasVisibleShadows : 2;
+    ShadowType mShadowType  : 2;                                            // :2
+    bool mHasVisibleShadows : 2;                                            // :2
+    uint8_t mFace           : 3;                                            // :3
 };
 
 } // namespace filament

filament/src/ShadowMapManager.cpp

@@ -98,22 +98,34 @@ void ShadowMapManager::setDirectionalShadowMap(size_t lightIndex,
     assert_invariant(options->shadowCascades <= CONFIG_MAX_SHADOW_CASCADES);
     for (size_t c = 0; c < options->shadowCascades; c++) {
         auto* pShadowMap = getCascadeShadowMap(c);
-        pShadowMap->initialize(lightIndex, ShadowType::DIRECTIONAL, c, options);
+        pShadowMap->initialize(lightIndex, ShadowType::DIRECTIONAL, c, 0, options);
         mCascadeShadowMaps.push_back(pShadowMap);
     }
 }
 
 void ShadowMapManager::addShadowMap(size_t lightIndex, bool spotlight,
         LightManager::ShadowOptions const* options) noexcept {
-    const size_t c = mSpotShadowMaps.size();
-    assert_invariant(c < CONFIG_MAX_SHADOWMAP_PUNCTUAL);
-    auto* pShadowMap = getPointOrSpotShadowMap(c);
-    pShadowMap->initialize(lightIndex, spotlight ? ShadowType::SPOT : ShadowType::POINT, c, options);
-    mSpotShadowMaps.push_back(pShadowMap);
+    if (spotlight) {
+        const size_t c = mSpotShadowMaps.size();
+        assert_invariant(c < CONFIG_MAX_SHADOWMAP_PUNCTUAL);
+        auto* pShadowMap = getPointOrSpotShadowMap(c);
+        pShadowMap->initialize(lightIndex, ShadowType::SPOT, c, 0, options);
+        mSpotShadowMaps.push_back(pShadowMap);
+    } else {
+        // point-light
+        for (size_t face = 0; face < 6; face++) {
+            const size_t c = mSpotShadowMaps.size();
+            assert_invariant(c < CONFIG_MAX_SHADOWMAP_PUNCTUAL);
+            auto* pShadowMap = getPointOrSpotShadowMap(c);
+            pShadowMap->initialize(lightIndex, ShadowType::POINT, c, face, options);
+            mSpotShadowMaps.push_back(pShadowMap);
+        }
+    }
 }
 
-FrameGraphId<FrameGraphTexture> ShadowMapManager::render(FrameGraph& fg, FEngine& engine,
-        RenderPass const& pass, FView& view, CameraInfo const& mainCameraInfo) noexcept {
+FrameGraphId<FrameGraphTexture> ShadowMapManager::render(FEngine& engine, FrameGraph& fg,
+        RenderPass const& pass, FView& view, CameraInfo const& mainCameraInfo,
+        float4 const& userTime) noexcept {
 
     const float moment2 = std::numeric_limits<half>::max();
     const float moment1 = std::sqrt(moment2);
@@ -132,12 +144,10 @@ FrameGraphId<FrameGraphTexture> ShadowMapManager::render(FrameGraph& fg, FEngine
 
     struct PrepareShadowPassData {
         struct ShadowPass {
-            mutable RenderPass pass;
-            mutable CameraInfo cameraInfo;
+            mutable RenderPass::Executor executor;
             ShadowMap* shadowMap;
             utils::Range<uint32_t> range;
             FScene::VisibleMaskType visibilityMask;
-            uint8_t face = 0;
         };
         // the actual shadow map atlas (currently a 2D texture array)
         FrameGraphId<FrameGraphTexture> shadows;
@@ -166,7 +176,7 @@ FrameGraphId<FrameGraphTexture> ShadowMapManager::render(FrameGraph& fg, FEngine
                         // for the directional light, we already know if it has visible shadows.
                         if (pShadowMap->hasVisibleShadows()) {
                             passList.push_back({
-                                    pass, {}, pShadowMap, directionalShadowCastersRange,
+                                    {}, pShadowMap, directionalShadowCastersRange,
                                     VISIBLE_DIR_SHADOW_RENDERABLE });
                         }
                     }
@@ -178,16 +188,8 @@ FrameGraphId<FrameGraphTexture> ShadowMapManager::render(FrameGraph& fg, FEngine
                     for (auto* pShadowMap : mSpotShadowMaps) {
                         assert_invariant(!pShadowMap->isDirectionalShadow());
                         passList.push_back({
-                                pass, {}, pShadowMap, spotShadowCastersRange,
+                                {}, pShadowMap, spotShadowCastersRange,
                                 VISIBLE_DYN_SHADOW_RENDERABLE });
-                        if (pShadowMap->isPointShadow()) {
-                            // add the 5 extra faces
-                            for (uint8_t face = 1; face < 6; face++) {
-                                passList.push_back({
-                                        pass, {}, pShadowMap, spotShadowCastersRange,
-                                        VISIBLE_DYN_SHADOW_RENDERABLE, face });
-                            }
-                        }
                     }
                 }
 
@@ -197,10 +199,8 @@ FrameGraphId<FrameGraphTexture> ShadowMapManager::render(FrameGraph& fg, FEngine
                 // "read" from one of its resource (only writes), so the FrameGraph culls it.
                 builder.sideEffect();
             },
-            [this, &engine, &view, scene, &mainCameraInfo](
+            [this, &engine, &view, scene, mainCameraInfo, userTime, passTemplate = pass](
                     FrameGraphResources const& resources, auto const& data, DriverApi& driver) {
-                // set uniforms needed to render this ShadowMap
-                view.prepareShadowMap(view.getVsmShadowOptions().highPrecision);
 
                 // Note: we could almost parallel_for the loop below, the problem currently is
                 // that updatePrimitivesLod() updates temporary global state.
@@ -215,7 +215,6 @@ FrameGraphId<FrameGraphTexture> ShadowMapManager::render(FrameGraph& fg, FEngine
                     // for spot shadow map, we need to do the culling
                     switch (shadowMap.getShadowType()) {
                         case ShadowType::DIRECTIONAL:
-                            // nothing to do, it's been done already
                             break;
                         case ShadowType::SPOT:
                             prepareSpotShadowMap(shadowMap, engine, view, mainCameraInfo,
@@ -225,7 +224,7 @@ FrameGraphId<FrameGraphTexture> ShadowMapManager::render(FrameGraph& fg, FEngine
                         case ShadowType::POINT:
                             preparePointShadowMap(shadowMap, engine, view, mainCameraInfo,
                                     scene->getRenderableData(), entry.range,
-                                    scene->getLightData(), entry.face, mSceneInfo);
+                                    scene->getLightData(), mSceneInfo);
                             break;
                     }
 
@@ -243,33 +242,50 @@ FrameGraphId<FrameGraphTexture> ShadowMapManager::render(FrameGraph& fg, FEngine
                         // cameraInfo only valid after calling update
                         const CameraInfo cameraInfo{ shadowMap.getCamera() };
 
+                        auto transaction = ShadowMap::open(driver);
+                        ShadowMap::prepareCamera(transaction, engine, cameraInfo);
+                        ShadowMap::prepareViewport(transaction, shadowMap.getViewport());
+                        ShadowMap::prepareTime(transaction, engine, userTime);
+                        ShadowMap::prepareShadowMapping(transaction,
+                                view.getVsmShadowOptions().highPrecision);
+                        if (shadowMap.getShadowType() == ShadowType::DIRECTIONAL) {
+                            ShadowMap::prepareDirectionalLight(transaction, engine,
+                                    scene->getLightData().elementAt<FScene::DIRECTION>(0),
+                                    scene->getLightData().elementAt<FScene::LIGHT_INSTANCE>(0));
+                        }
+                        shadowMap.commit(transaction, driver);
+
                         // updatePrimitivesLod must be run before RenderPass::appendCommands.
                         view.updatePrimitivesLod(engine,
                                 cameraInfo, scene->getRenderableData(), entry.range);
 
+                        RenderPass pass(passTemplate);
+                        pass.setCamera(cameraInfo);
+                        pass.setVisibilityMask(entry.visibilityMask);
+                        pass.setGeometry(scene->getRenderableData(),
+                                entry.range, scene->getRenderableUBO());
+                        pass.appendCommands(engine, RenderPass::SHADOW);
+                        pass.sortCommands(engine);
+
+                        entry.executor = pass.getExecutor();
+
                         if (!view.hasVSM()) {
                             auto li = scene->getLightData().elementAt<FScene::LIGHT_INSTANCE>(
-                                    shadowMap.getLightIndex());
+                                    entry.shadowMap->getLightIndex());
+
                             auto const& params = engine.getLightManager().getShadowParams(li);
+
                             const PolygonOffset polygonOffset = { // handle reversed Z
                                     .slope    = -params.options.polygonOffsetSlope,
                                     .constant = -params.options.polygonOffsetConstant
                             };
-                            entry.pass.overridePolygonOffset(&polygonOffset);
+                            entry.executor.overridePolygonOffset(&polygonOffset);
                         }
 
                         constexpr const backend::Viewport disabledScissor{ 0, 0,
                                 (uint32_t)std::numeric_limits<int32_t>::max(),
                                 (uint32_t)std::numeric_limits<int32_t>::max() };
-
-                        entry.pass.setCamera(cameraInfo);
-                        entry.pass.setVisibilityMask(entry.visibilityMask);
-                        entry.pass.setGeometry(scene->getRenderableData(),
-                                entry.range, scene->getRenderableUBO());
-                        entry.pass.overrideScissor(&disabledScissor);
-                        entry.pass.appendCommands(RenderPass::SHADOW);
-                        entry.pass.sortCommands();
-                        entry.cameraInfo = cameraInfo;
+                        entry.executor.overrideScissor(&disabledScissor);
                     }
                 }
 
@@ -287,8 +303,7 @@ FrameGraphId<FrameGraphTexture> ShadowMapManager::render(FrameGraph& fg, FEngine
     struct ShadowPassData {
         FrameGraphId<FrameGraphTexture> tempBlurSrc{};  // temporary shadowmap when blurring
         FrameGraphId<FrameGraphTexture> output;
-        uint32_t blurRt{};
-        uint32_t shadowRt{};
+        uint32_t rt{};
     };
 
     auto const& passList = prepareShadowPass.getData().passList;
@@ -297,7 +312,7 @@ FrameGraphId<FrameGraphTexture> ShadowMapManager::render(FrameGraph& fg, FEngine
             continue;
         }
         
-        const uint8_t layer = entry.shadowMap->getLayer() + entry.face;
+        const uint8_t layer = entry.shadowMap->getLayer();
         const auto* options = entry.shadowMap->getShadowOptions();
 
         auto& shadowPass = fg.addPass<ShadowPassData>("Shadow Pass",
@@ -346,7 +361,7 @@ FrameGraphId<FrameGraphTexture> ShadowMapManager::render(FrameGraph& fg, FEngine
                             data.tempBlurSrc = builder.write(data.tempBlurSrc,
                                     FrameGraphTexture::Usage::COLOR_ATTACHMENT);
 
-                            data.blurRt = builder.declareRenderPass("Temp Shadow RT", {
+                            data.rt = builder.declareRenderPass("Temp Shadow RT", {
                                     .attachments = {
                                             .color = { data.tempBlurSrc },
                                             .depth = depth },
@@ -365,9 +380,13 @@ FrameGraphId<FrameGraphTexture> ShadowMapManager::render(FrameGraph& fg, FEngine
                     }
 
                     // finally, create the shadowmap render target -- one per layer.
-                    data.shadowRt = builder.declareRenderPass("Shadow RT", renderTargetDesc);
+                    auto rt = builder.declareRenderPass("Shadow RT", renderTargetDesc);
+
+                    // render either directly into the shadowmap, or to the temporary texture for
+                    // blurring.
+                    data.rt = blur ? data.rt : rt;
                 },
-                [=, &entry, &view](FrameGraphResources const& resources,
+                [=, &engine, &entry](FrameGraphResources const& resources,
                         auto const& data, DriverApi& driver) {
 
                     // Note: we capture entry by reference here. That's actually okay because
@@ -377,41 +396,15 @@ FrameGraphId<FrameGraphTexture> ShadowMapManager::render(FrameGraph& fg, FEngine
                     // It wouldn't work to capture by copy because entry.pass wouldn't be
                     // initialized, as this happens in an `execute` block.
 
-                    CameraInfo const& cameraInfo = entry.cameraInfo;
+                    entry.shadowMap->bindPerViewUniformsAndSamplers(driver);
 
-                    const auto& executor = entry.pass.getExecutor();
-                    const bool blur = view.hasVSM() && options->vsm.blurWidth > 0.0f;
+                    auto rt = resources.getRenderPassInfo(data.rt);
+                    rt.params.viewport = entry.shadowMap->getViewport();
 
-                    // set uniforms relative to the camera
-                    view.prepareCamera(cameraInfo);
-
-                    // We set a viewport with a 1-texel border for when we index outside the
-                    // texture.
-                    // DON'T CHANGE this unless ShadowMap::getTextureCoordsMapping() is updated too.
-                    // see: ShadowMap::getTextureCoordsMapping()
-                    //
-                    // For floating-point depth textures, the 1-texel border could be set to
-                    // FLOAT_MAX to avoid clamping in the shadow shader (see sampleDepth inside
-                    // shadowing.fs). Unfortunately, the APIs don't seem let us clear depth
-                    // attachments to anything greater than 1.0, so we'd need a way to do this other
-                    // than clearing.
-                    const uint32_t dim = options->mapSize;
-                    filament::Viewport viewport{ 1, 1, dim - 2, dim - 2 };
-                    if (entry.shadowMap->isPointShadow()) {
-                        // for point-light we don't have a border
-                        viewport = { 0, 0, dim, dim };
-                    }
-
-                    view.prepareViewport(viewport, 0, 0);
-
-                    view.commitUniforms(driver);
-
-                    // render either directly into the shadowmap, or to the temporary texture for
-                    // blurring.
-                    auto rt = resources.getRenderPassInfo(blur ? data.blurRt : data.shadowRt);
-                    rt.params.viewport = viewport;
-
-                    executor.execute("Shadow Pass", rt.target, rt.params);
+                    engine.flush();
+                    driver.beginRenderPass(rt.target, rt.params);
+                    entry.executor.execute(engine, "Shadow Pass");
+                    driver.endRenderPass();
                 });
 
 
@@ -701,9 +694,10 @@ void ShadowMapManager::prepareSpotShadowMap(ShadowMap& shadowMap,
 void ShadowMapManager::preparePointShadowMap(ShadowMap& shadowMap,
         FEngine& engine, FView& view, CameraInfo const& mainCameraInfo,
         FScene::RenderableSoa& renderableData, utils::Range<uint32_t> range,
-        FScene::LightSoa& lightData, uint8_t face,
+        FScene::LightSoa& lightData,
         ShadowMap::SceneInfo const& sceneInfo) noexcept {
 
+    const uint8_t face = shadowMap.getFace();
     const size_t lightIndex = shadowMap.getLightIndex();
     FLightManager::ShadowOptions const* const options = shadowMap.getShadowOptions();
 
@@ -794,9 +788,15 @@ ShadowMapManager::ShadowTechnique ShadowMapManager::updateSpotShadowMaps(FEngine
         shadowTechnique |= ShadowTechnique::SHADOW_MAP;
         for (auto const* pShadowMap : mSpotShadowMaps) {
             const size_t lightIndex = pShadowMap->getLightIndex();
-            shadowInfo[lightIndex].castsShadows = true;     // FIXME: is that set correctly?
-            shadowInfo[lightIndex].index = pShadowMap->getShadowIndex();
-            shadowInfo[lightIndex].layer = pShadowMap->getLayer();
+
+            // FIXME: currently we have one slot per shadowmap in the UBO, but we now have up to
+            //        6 shadowmap per light. So for now, we only write the data of the face 0,
+            //        and the shader will figure out where to find the other face (layer+face)
+            if (pShadowMap->getFace() == 0) {
+                shadowInfo[lightIndex].castsShadows = true;     // FIXME: is that set correctly?
+                shadowInfo[lightIndex].index = pShadowMap->getShadowIndex();
+                shadowInfo[lightIndex].layer = pShadowMap->getLayer();
+            }
         }
     }
 
@@ -837,10 +837,6 @@ void ShadowMapManager::calculateTextureRequirements(FEngine& engine, FView& view
         maxDimension = std::max(maxDimension, options->mapSize);
         elvsm = elvsm || options->vsm.elvsm;
         pShadowMap->setLayer(layer++);
-        if (pShadowMap->isPointShadow()) {
-            // FIXME: we need to make sure we don't exceed 256
-            layer += 5;
-        }
     }
 
     const uint8_t layersNeeded = layer;

filament/src/ShadowMapManager.h

@@ -90,8 +90,8 @@ public:
             FScene::RenderableSoa& renderableData, FScene::LightSoa& lightData) noexcept;
 
     // Renders all the shadow maps.
-    FrameGraphId<FrameGraphTexture> render(FrameGraph& fg, FEngine& engine,
-            RenderPass const& pass, FView& view, CameraInfo const& mainCameraInfo) noexcept;
+    FrameGraphId<FrameGraphTexture> render(FEngine& engine, FrameGraph& fg, RenderPass const& pass,
+            FView& view, CameraInfo const& mainCameraInfo, math::float4 const& userTime) noexcept;
 
     ShadowMap* getCascadeShadowMap(size_t cascade) noexcept {
         assert_invariant(cascade < CONFIG_MAX_SHADOW_CASCADES);
@@ -140,7 +140,7 @@ private:
     void preparePointShadowMap(ShadowMap& map,
             FEngine& engine, FView& view, CameraInfo const& mainCameraInfo,
             FScene::RenderableSoa& renderableData, utils::Range<uint32_t> range,
-            FScene::LightSoa& lightData, uint8_t face,
+            FScene::LightSoa& lightData,
             ShadowMap::SceneInfo const& sceneInfo) noexcept;
 
     static void updateSpotVisibilityMasks(
@@ -205,7 +205,7 @@ private:
     mutable TypedUniformBuffer<ShadowUib> mShadowUb;
     backend::Handle<backend::HwBufferObject> mShadowUbh;
 
-    ShadowMappingUniforms mShadowMappingUniforms;
+    ShadowMappingUniforms mShadowMappingUniforms = {};
 
     ShadowMap::SceneInfo mSceneInfo;
 
@@ -219,7 +219,7 @@ private:
 
     // inline storage for all our ShadowMap objects, we can't easily use a std::array<> directly.
     // because ShadowMap doesn't have a default ctor, and we avoid out-of-line allocations.
-    // Each ShadowMap is currently 32 bytes.
+    // Each ShadowMap is currently 40 bytes (total of 2.5KB for 64 shadow maps)
     using ShadowMapStorage = std::aligned_storage<sizeof(ShadowMap), alignof(ShadowMap)>::type;
     std::array<ShadowMapStorage, CONFIG_MAX_SHADOW_LAYERS> mShadowMapCache;
 };

filament/src/TypedUniformBuffer.h

@@ -46,7 +46,7 @@ public:
     // return if any uniform has been changed
     bool isDirty() const noexcept { return mSomethingDirty; }
 
-    // mark the whole buffer as clean (no modified uniforms)
+    // mark the whole buffer as "clean" (no modified uniforms)
     void clean() const noexcept { mSomethingDirty = false; }
 
     // helper functions

filament/src/components/RenderableManager.cpp

@@ -747,4 +747,8 @@ bool FRenderableManager::getLightChannel(Instance ci, unsigned int channel) cons
     return false;
 }
 
+size_t FRenderableManager::getPrimitiveCount(Instance instance, uint8_t level) const noexcept {
+    return getRenderPrimitives(instance, level).size();
+}
+
 } // namespace filament

filament/src/components/RenderableManager.h

@@ -160,7 +160,7 @@ public:
     }
 
     inline size_t getLevelCount(Instance) const noexcept { return 1; }
-    inline size_t getPrimitiveCount(Instance instance, uint8_t level) const noexcept;
+    size_t getPrimitiveCount(Instance instance, uint8_t level) const noexcept;
     void setMaterialInstanceAt(Instance instance, uint8_t level,
             size_t primitiveIndex, FMaterialInstance const* materialInstance);
     MaterialInstance* getMaterialInstanceAt(Instance instance, uint8_t level, size_t primitiveIndex) const noexcept;
@@ -411,10 +411,6 @@ utils::Slice<FRenderableManager::MorphTargets>& FRenderableManager::getMorphTarg
     return mManager[instance].morphTargets;
 }
 
-size_t FRenderableManager::getPrimitiveCount(Instance instance, uint8_t level) const noexcept {
-    return getRenderPrimitives(instance, level).size();
-}
-
 } // namespace filament
 
 #endif // TNT_FILAMENT_COMPONENTS_RENDERABLEMANAGER_H

filament/src/details/Engine.cpp

@@ -1072,4 +1072,10 @@ Engine::FeatureLevel FEngine::setActiveFeatureLevel(FeatureLevel featureLevel) {
     return (mActiveFeatureLevel = std::max(mActiveFeatureLevel, featureLevel));
 }
 
+#if defined(__EMSCRIPTEN__)
+void FEngine::resetBackendState() noexcept {
+    getDriverApi().resetState();
+}
+#endif
+
 } // namespace filament

filament/src/details/Engine.h

@@ -374,6 +374,10 @@ public:
         return FEngine::getActiveFeatureLevel() >= neededFeatureLevel;
     }
 
+#if defined(__EMSCRIPTEN__)
+    void resetBackendState() noexcept;
+#endif
+
 private:
     static Config validateConfig(const Config* pConfig) noexcept;
 

filament/src/details/MaterialInstance.cpp

@@ -35,17 +35,30 @@ namespace filament {
 
 using namespace backend;
 
-FMaterialInstance::FMaterialInstance() noexcept = default;
+FMaterialInstance::FMaterialInstance() noexcept
+        : mCulling(CullingMode::BACK),
+          mDepthFunc(RasterState::DepthFunc::LE),
+          mColorWrite(false),
+          mDepthWrite(false),
+          mHasScissor(false),
+          mIsDoubleSided(false),
+          mTransparencyMode(TransparencyMode::DEFAULT) {
+}
 
 FMaterialInstance::FMaterialInstance(FEngine& engine,
         FMaterialInstance const* other, const char* name)
         : mMaterial(other->mMaterial),
           mPolygonOffset(other->mPolygonOffset),
+          mStencilState(other->mStencilState),
+          mMaskThreshold(other->mMaskThreshold),
+          mSpecularAntiAliasingVariance(other->mSpecularAntiAliasingVariance),
+          mSpecularAntiAliasingThreshold(other->mSpecularAntiAliasingThreshold),
           mCulling(other->mCulling),
+          mDepthFunc(other->mDepthFunc),
           mColorWrite(other->mColorWrite),
           mDepthWrite(other->mDepthWrite),
-          mStencilState(other->mStencilState),
-          mDepthFunc(other->mDepthFunc),
+          mHasScissor(false),
+          mIsDoubleSided(other->mIsDoubleSided),
           mScissorRect(other->mScissorRect),
           mName(name ? CString(name) : other->mName) {
 
@@ -63,10 +76,23 @@ FMaterialInstance::FMaterialInstance(FEngine& engine,
         mSbHandle = driver.createSamplerGroup(mSamplers.getSize());
     }
 
-    mMaterialSortingKey = RenderPass::makeMaterialSortingKey(
-            material->getId(), material->generateMaterialInstanceId());
+    if (material->hasDoubleSidedCapability()) {
+        setDoubleSided(mIsDoubleSided);
+    }
+
+    if (material->getBlendingMode() == BlendingMode::MASKED) {
+        setMaskThreshold(mMaskThreshold);
+    }
+
+    if (material->hasSpecularAntiAliasing()) {
+        setSpecularAntiAliasingThreshold(mSpecularAntiAliasingThreshold);
+        setSpecularAntiAliasingVariance(mSpecularAntiAliasingVariance);
+    }
 
     setTransparencyMode(material->getTransparencyMode());
+
+    mMaterialSortingKey = RenderPass::makeMaterialSortingKey(
+            material->getId(), material->generateMaterialInstanceId());
 }
 
 FMaterialInstance* FMaterialInstance::duplicate(
@@ -183,14 +209,29 @@ void FMaterialInstance::setParameterImpl(std::string_view name,
 
 void FMaterialInstance::setMaskThreshold(float threshold) noexcept {
     setParameter("_maskThreshold", math::saturate(threshold));
+    mMaskThreshold = math::saturate(threshold);
+}
+
+float FMaterialInstance::getMaskThreshold() const noexcept {
+    return mMaskThreshold;
 }
 
 void FMaterialInstance::setSpecularAntiAliasingVariance(float variance) noexcept {
     setParameter("_specularAntiAliasingVariance", math::saturate(variance));
+    mSpecularAntiAliasingVariance = math::saturate(variance);
+}
+
+float FMaterialInstance::getSpecularAntiAliasingVariance() const noexcept {
+    return mSpecularAntiAliasingVariance;
 }
 
 void FMaterialInstance::setSpecularAntiAliasingThreshold(float threshold) noexcept {
     setParameter("_specularAntiAliasingThreshold", math::saturate(threshold * threshold));
+    mSpecularAntiAliasingThreshold = std::sqrt(math::saturate(threshold * threshold));
+}
+
+float FMaterialInstance::getSpecularAntiAliasingThreshold() const noexcept {
+    return mSpecularAntiAliasingThreshold;
 }
 
 void FMaterialInstance::setDoubleSided(bool doubleSided) noexcept {
@@ -202,6 +243,11 @@ void FMaterialInstance::setDoubleSided(bool doubleSided) noexcept {
     if (doubleSided) {
         setCullingMode(CullingMode::NONE);
     }
+    mIsDoubleSided = doubleSided;
+}
+
+bool FMaterialInstance::isDoubleSided() const noexcept {
+    return mIsDoubleSided;
 }
 
 void FMaterialInstance::setTransparencyMode(TransparencyMode mode) noexcept {
@@ -212,6 +258,10 @@ void FMaterialInstance::setDepthCulling(bool enable) noexcept {
     mDepthFunc = enable ? RasterState::DepthFunc::GE : RasterState::DepthFunc::A;
 }
 
+bool FMaterialInstance::isDepthCullingEnabled() const noexcept {
+    return mDepthFunc != RasterState::DepthFunc::A;
+}
+
 const char* FMaterialInstance::getName() const noexcept {
     // To decide whether to use the parent material name as a fallback, we check for the nullness of
     // the instance's CString rather than calling empty(). This allows instances to override the

filament/src/details/MaterialInstance.h

@@ -89,9 +89,11 @@ public:
 
     backend::CullingMode getCullingMode() const noexcept { return mCulling; }
 
-    bool getColorWrite() const noexcept { return mColorWrite; }
+    bool isColorWriteEnabled() const noexcept { return mColorWrite; }
 
-    bool getDepthWrite() const noexcept { return mDepthWrite; }
+    bool isDepthWriteEnabled() const noexcept { return mDepthWrite; }
+
+    bool isStencilWriteEnabled() const noexcept { return mStencilState.stencilWrite; }
 
     backend::StencilState getStencilState() const noexcept { return mStencilState; }
 
@@ -108,12 +110,20 @@ public:
 
     void setMaskThreshold(float threshold) noexcept;
 
+    float getMaskThreshold() const noexcept;
+
     void setSpecularAntiAliasingVariance(float variance) noexcept;
 
+    float getSpecularAntiAliasingVariance() const noexcept;
+
     void setSpecularAntiAliasingThreshold(float threshold) noexcept;
 
+    float getSpecularAntiAliasingThreshold() const noexcept;
+
     void setDoubleSided(bool doubleSided) noexcept;
 
+    bool isDoubleSided() const noexcept;
+
     void setTransparencyMode(TransparencyMode mode) noexcept;
 
     void setCullingMode(CullingMode culling) noexcept { mCulling = culling; }
@@ -126,6 +136,8 @@ public:
 
     void setDepthCulling(bool enable) noexcept;
 
+    bool isDepthCullingEnabled() const noexcept;
+
     void setStencilCompareFunction(StencilCompareFunc func, StencilFace face) noexcept {
         if (any(face & StencilFace::FRONT)) {
             mStencilState.front.stencilFunc = func;
@@ -222,19 +234,26 @@ private:
 
     // keep these grouped, they're accessed together in the render-loop
     FMaterial const* mMaterial = nullptr;
+
     backend::Handle<backend::HwBufferObject> mUbHandle;
     backend::Handle<backend::HwSamplerGroup> mSbHandle;
-
     UniformBuffer mUniforms;
     backend::SamplerGroup mSamplers;
-    backend::PolygonOffset mPolygonOffset;
-    backend::CullingMode mCulling;
-    bool mColorWrite;
-    bool mDepthWrite;
-    bool mHasScissor = false;
-    backend::StencilState mStencilState = {};
-    backend::RasterState::DepthFunc mDepthFunc;
-    TransparencyMode mTransparencyMode;
+
+    backend::PolygonOffset mPolygonOffset{};
+    backend::StencilState mStencilState{};
+
+    float mMaskThreshold = 0.0f;
+    float mSpecularAntiAliasingVariance = 0.0f;
+    float mSpecularAntiAliasingThreshold = 0.0f;
+
+    backend::CullingMode mCulling : 2;
+    backend::RasterState::DepthFunc mDepthFunc : 3;
+    bool mColorWrite : 1;
+    bool mDepthWrite : 1;
+    bool mHasScissor : 1;
+    bool mIsDoubleSided : 1;
+    TransparencyMode mTransparencyMode : 2;
 
     uint64_t mMaterialSortingKey = 0;
 

filament/src/details/Renderer.cpp

@@ -656,7 +656,7 @@ void FRenderer::renderJob(ArenaScope& arena, FView& view) {
 
         RenderPass shadowPass(pass);
         shadowPass.setVariant(shadowVariant);
-        auto shadows = view.renderShadowMaps(fg, engine, cameraInfo, shadowPass);
+        auto shadows = view.renderShadowMaps(engine, fg, cameraInfo, mShaderUserTime, shadowPass);
         blackboard["shadows"] = shadows;
     }
 
@@ -740,8 +740,8 @@ void FRenderer::renderJob(ArenaScope& arena, FView& view) {
 
     // view set-ups that need to happen before rendering
     fg.addTrivialSideEffectPass("Prepare View Uniforms",
-            [=, &uniforms = view.getPerViewUniforms()](DriverApi& driver) {
-                uniforms.prepareCamera(cameraInfo);
+            [=, &view, &engine](DriverApi& driver) {
+                view.prepareCamera(engine, cameraInfo);
 
                 // The code here is a little fragile. In theory, we need to call prepareViewport()
                 // for each render pass, because the viewport parameters depend on the resolution.
@@ -761,11 +761,14 @@ void FRenderer::renderJob(ArenaScope& arena, FView& view) {
                 // The reason why this bug is acceptable is that the viewport parameters are
                 // currently only used for generating noise, so it's not too bad.
 
-                uniforms.prepareViewport(svp,
+                view.prepareViewport(svp,
                         xvp.left   * aoOptions.resolution,
                         xvp.bottom * aoOptions.resolution);
 
-                uniforms.commit(driver);
+                view.commitUniforms(driver);
+
+                // set uniforms and samplers for the color passes
+                view.bindPerViewUniformsAndSamplers(driver);
             });
 
     // --------------------------------------------------------------------------------------------
@@ -850,8 +853,8 @@ void FRenderer::renderJob(ArenaScope& arena, FView& view) {
     // This one doesn't need to be a FrameGraph pass because it always happens by construction
     // (i.e. it won't be culled, unless everything is culled), so no need to complexify things.
     pass.setVariant(variant);
-    pass.appendCommands(RenderPass::COLOR);
-    pass.sortCommands();
+    pass.appendCommands(engine, RenderPass::COLOR);
+    pass.sortCommands(engine);
 
     FrameGraphTexture::Descriptor desc = {
             .width = config.width,

filament/src/details/View.cpp

@@ -383,7 +383,7 @@ void FView::prepareLighting(FEngine& engine, FEngine::DriverApi& driver, ArenaSc
         intensity = skybox ? skybox->getIntensity() : FIndirectLight::DEFAULT_INTENSITY;
     }
 
-    mPerViewUniforms.prepareAmbientLight(*ibl, intensity, exposure);
+    mPerViewUniforms.prepareAmbientLight(engine, *ibl, intensity, exposure);
 
     /*
      * Directional light (always at index 0)
@@ -391,7 +391,7 @@ void FView::prepareLighting(FEngine& engine, FEngine::DriverApi& driver, ArenaSc
 
     FLightManager::Instance directionalLight = lightData.elementAt<FScene::LIGHT_INSTANCE>(0);
     const float3 sceneSpaceDirection = lightData.elementAt<FScene::DIRECTION>(0); // guaranteed normalized
-    mPerViewUniforms.prepareDirectionalLight(exposure, sceneSpaceDirection, directionalLight);
+    mPerViewUniforms.prepareDirectionalLight(engine, exposure, sceneSpaceDirection, directionalLight);
     mHasDirectionalLight = directionalLight.isValid();
 }
 
@@ -602,13 +602,23 @@ void FView::prepare(FEngine& engine, DriverApi& driver, ArenaScope& arena,
      * Update driver state
      */
 
-    mPerViewUniforms.prepareTime(userTime);
+    mPerViewUniforms.prepareTime(engine, userTime);
     mPerViewUniforms.prepareFog(cameraInfo.getPosition(), mFogOptions);
-    mPerViewUniforms.prepareTemporalNoise(mTemporalAntiAliasingOptions);
+    mPerViewUniforms.prepareTemporalNoise(engine, mTemporalAntiAliasingOptions);
     mPerViewUniforms.prepareBlending(needsAlphaChannel);
+}
 
-    // set uniforms and samplers
-    bindPerViewUniformsAndSamplers(driver);
+void FView::bindPerViewUniformsAndSamplers(FEngine::DriverApi& driver) const noexcept {
+    mPerViewUniforms.bind(driver);
+
+    driver.bindUniformBuffer(+UniformBindingPoints::LIGHTS,
+            mLightUbh);
+
+    driver.bindUniformBuffer(+UniformBindingPoints::SHADOW,
+            mShadowMapManager.getShadowUniformsHandle());
+
+    driver.bindUniformBuffer(+UniformBindingPoints::FROXEL_RECORDS,
+            mFroxelizer.getRecordBuffer());
 }
 
 void FView::computeVisibilityMasks(
@@ -656,12 +666,14 @@ UTILS_NOINLINE
 
 void FView::prepareUpscaler(float2 scale) const noexcept {
     SYSTRACE_CALL();
-    mPerViewUniforms.prepareUpscaler(scale, mDynamicResolution);
+    const float bias = (mDynamicResolution.quality >= QualityLevel::HIGH) ?
+            std::log2(std::min(scale.x, scale.y)) : 0.0f;
+    mPerViewUniforms.prepareLodBias(bias);
 }
 
-void FView::prepareCamera(const CameraInfo& cameraInfo) const noexcept {
+void FView::prepareCamera(FEngine& engine, const CameraInfo& cameraInfo) const noexcept {
     SYSTRACE_CALL();
-    mPerViewUniforms.prepareCamera(cameraInfo);
+    mPerViewUniforms.prepareCamera(engine, cameraInfo);
 }
 
 void FView::prepareViewport(const filament::Viewport& viewport,
@@ -711,7 +723,7 @@ void FView::prepareShadow(Handle<HwTexture> texture) const noexcept {
     }
 }
 
-void FView::prepareShadowMap(bool highPrecision) const noexcept {
+void FView::prepareShadowMapping(bool highPrecision) const noexcept {
     mPerViewUniforms.prepareShadowMapping(highPrecision);
 }
 
@@ -894,9 +906,9 @@ void FView::updatePrimitivesLod(FEngine& engine, const CameraInfo&,
     }
 }
 
-FrameGraphId<FrameGraphTexture> FView::renderShadowMaps(FrameGraph& fg, FEngine& engine,
-        CameraInfo const& cameraInfo, RenderPass const& pass) noexcept {
-    return mShadowMapManager.render(fg, engine, pass, *this, cameraInfo);
+FrameGraphId<FrameGraphTexture> FView::renderShadowMaps(FEngine& engine, FrameGraph& fg,
+        CameraInfo const& cameraInfo, float4 const& userTime, RenderPass const& pass) noexcept {
+    return mShadowMapManager.render(engine, fg, pass, *this, cameraInfo, userTime);
 }
 
 void FView::commitFrameHistory(FEngine& engine) noexcept {

filament/src/details/View.h

@@ -92,6 +92,8 @@ public:
             filament::Viewport const& viewport, CameraInfo const& cameraInfo,
             math::float4 const& userTime, bool needsAlphaChannel) noexcept;
 
+    void bindPerViewUniformsAndSamplers(FEngine::DriverApi& driver) const noexcept;
+
     void setScene(FScene* scene) { mScene = scene; }
     FScene const* getScene() const noexcept { return mScene; }
     FScene* getScene() noexcept { return mScene; }
@@ -132,7 +134,7 @@ public:
     }
 
     void prepareUpscaler(math::float2 scale) const noexcept;
-    void prepareCamera(const CameraInfo& cameraInfo) const noexcept;
+    void prepareCamera(FEngine& engine, const CameraInfo& cameraInfo) const noexcept;
     void prepareViewport(const Viewport& viewport, uint32_t xoffset, uint32_t yoffset) const noexcept;
     void prepareShadowing(FEngine& engine, backend::DriverApi& driver,
             FScene::RenderableSoa& renderableData, FScene::LightSoa& lightData,
@@ -148,7 +150,7 @@ public:
             ScreenSpaceReflectionsOptions const& ssrOptions) const noexcept;
     void prepareStructure(backend::Handle<backend::HwTexture> structure) const noexcept;
     void prepareShadow(backend::Handle<backend::HwTexture> structure) const noexcept;
-    void prepareShadowMap(bool highPrecision) const noexcept;
+    void prepareShadowMapping(bool highPrecision) const noexcept;
 
     void cleanupRenderPasses() const noexcept;
     void froxelize(FEngine& engine, math::mat4f const& viewMatrix) const noexcept;
@@ -165,8 +167,9 @@ public:
     bool hasPCSS() const noexcept { return mShadowType == ShadowType::PCSS; }
     bool hasPicking() const noexcept { return mActivePickingQueriesList != nullptr; }
 
-    FrameGraphId<FrameGraphTexture> renderShadowMaps(FrameGraph& fg, FEngine& engine,
-            CameraInfo const& cameraInfo, RenderPass const& pass) noexcept;
+    FrameGraphId<FrameGraphTexture> renderShadowMaps(FEngine& engine, FrameGraph& fg,
+            CameraInfo const& cameraInfo, math::float4 const& userTime,
+            RenderPass const& pass) noexcept;
 
     void updatePrimitivesLod(
             FEngine& engine, const CameraInfo& camera,
@@ -451,13 +454,6 @@ private:
             Culler::result_type* visibleMask,
             size_t count);
 
-    void bindPerViewUniformsAndSamplers(FEngine::DriverApi& driver) const noexcept {
-        mPerViewUniforms.bind(driver);
-        driver.bindUniformBuffer(+UniformBindingPoints::LIGHTS, mLightUbh);
-        driver.bindUniformBuffer(+UniformBindingPoints::SHADOW, mShadowMapManager.getShadowUniformsHandle());
-        driver.bindUniformBuffer(+UniformBindingPoints::FROXEL_RECORDS, mFroxelizer.getRecordBuffer());
-    }
-
     // Clean-up the whole history, free all resources. This is typically called when the View is
     // being terminated.
     void drainFrameHistory(FEngine& engine) noexcept;

filament/test/CMakeLists.txt

@@ -42,6 +42,7 @@ list(APPEND RESGEN_SOURCE ${DUMMY_SRC})
 # away in Release builds
 if (TNT_DEV)
     add_executable(test_${TARGET}
+            filament_AtlasAllocator_test.cpp
             filament_test_exposure.cpp
             filament_rendering_test.cpp
             filament_framegraph_test.cpp

filament/test/filament_AtlasAllocator_test.cpp

@@ -0,0 +1,149 @@
+/*
+ * Copyright (C) 2022 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include <gtest/gtest.h>
+
+#include "AtlasAllocator.h"
+
+using namespace filament;
+
+TEST(AtlasAllocator, AllocateFirstLevel) {
+
+    AtlasAllocator allocator;
+
+    auto a = allocator.allocateInLayer(0);
+    EXPECT_TRUE(a.l == 0 && a.code == 0);
+
+    auto b = allocator.allocateInLayer(1);
+    EXPECT_TRUE(b.l < 0);
+
+    auto c = allocator.allocateInLayer(2);
+    EXPECT_TRUE(c.l < 0);
+
+    auto d = allocator.allocateInLayer(5);
+    EXPECT_TRUE(d.l < 0);
+}
+
+TEST(AtlasAllocator, AllocateSecondLevel) {
+
+    AtlasAllocator allocator;
+
+    auto d0 = allocator.allocateInLayer(1);
+    EXPECT_TRUE(d0.l == 1 && d0.code == 0);
+
+    auto d1 = allocator.allocateInLayer(1);
+    EXPECT_TRUE(d1.l == 1 && d1.code == 1);
+
+    auto d2 = allocator.allocateInLayer(1);
+    EXPECT_TRUE(d2.l == 1 && d2.code == 2);
+
+    auto d3 = allocator.allocateInLayer(1);
+    EXPECT_TRUE(d3.l == 1 && d3.code == 3);
+}
+
+TEST(AtlasAllocator, AllocateMixed0) {
+    AtlasAllocator allocator;
+
+    auto e0 = allocator.allocateInLayer(1);
+    EXPECT_TRUE(e0.l == 1 && e0.code == 0);
+
+    auto e1 = allocator.allocateInLayer(1);
+    EXPECT_TRUE(e1.l == 1 && e1.code == 1);
+
+    auto e2 = allocator.allocateInLayer(1);
+    EXPECT_TRUE(e2.l == 1 && e2.code == 2);
+
+    auto e3 = allocator.allocateInLayer(2);
+    EXPECT_TRUE(e3.l == 2 && e3.code == 12);
+
+    auto e4 = allocator.allocateInLayer(1);
+    EXPECT_TRUE(e4.l < 0);
+}
+
+TEST(AtlasAllocator, AllocateMixed1) {
+    AtlasAllocator allocator;
+
+    auto e0 = allocator.allocateInLayer(1);
+    EXPECT_TRUE(e0.l == 1 && e0.code == 0);
+
+    auto e1 = allocator.allocateInLayer(1);
+    EXPECT_TRUE(e1.l == 1 && e1.code == 1);
+
+    auto e2 = allocator.allocateInLayer(2);
+    EXPECT_TRUE(e2.l == 2 && e2.code == 8);
+
+    auto e3 = allocator.allocateInLayer(1);
+    EXPECT_TRUE(e3.l == 1 && e3.code == 3);
+}
+
+TEST(AtlasAllocator, AllocateMixed2) {
+    AtlasAllocator allocator;
+
+    auto e0 = allocator.allocateInLayer(1);
+    EXPECT_TRUE(e0.l == 1 && e0.code == 0);
+
+    auto e1 = allocator.allocateInLayer(1);
+    EXPECT_TRUE(e1.l == 1 && e1.code == 1);
+
+    auto c0 = allocator.allocateInLayer(2);
+    EXPECT_TRUE(c0.l == 2 && c0.code == 8);
+    auto c1 = allocator.allocateInLayer(2);
+    EXPECT_TRUE(c1.l == 2 && c1.code == 9);
+    auto c2 = allocator.allocateInLayer(2);
+    EXPECT_TRUE(c2.l == 2 && c2.code == 10);
+    auto c3 = allocator.allocateInLayer(2);
+    EXPECT_TRUE(c3.l == 2 && c3.code == 11);
+
+    auto c4 = allocator.allocateInLayer(2);
+    EXPECT_TRUE(c4.l == 2 && c4.code == 12);
+    auto c5 = allocator.allocateInLayer(2);
+    EXPECT_TRUE(c5.l == 2 && c5.code == 13);
+    auto c6 = allocator.allocateInLayer(2);
+    EXPECT_TRUE(c6.l == 2 && c6.code == 14);
+    auto c7 = allocator.allocateInLayer(2);
+    EXPECT_TRUE(c7.l == 2 && c7.code == 15);
+}
+
+TEST(AtlasAllocator, AllocateBySize) {
+    AtlasAllocator allocator(256);
+
+    Viewport vp(0,0,256,256);
+    auto vp0 = allocator.allocate(256);
+    EXPECT_EQ(vp0, vp);
+
+    auto vp1 = allocator.allocate(128);
+    EXPECT_TRUE(vp1.empty());
+}
+
+TEST(AtlasAllocator, AllocateBySizeOneOfEach) {
+    AtlasAllocator allocator(256);
+
+    Viewport r0(0,0,128,128);
+    Viewport r1(128,0,64,64);
+    Viewport r2(192,0,32,32);
+
+    auto vp0 = allocator.allocate(128);
+    auto vp1 = allocator.allocate(64);
+    auto vp2 = allocator.allocate(32);
+    auto vp3 = allocator.allocate(16);
+
+    EXPECT_EQ(vp0, r0);
+    EXPECT_EQ(vp1, r1);
+    EXPECT_EQ(vp2, r2);
+    EXPECT_TRUE(vp3.empty());
+
+}
+

ios/CocoaPods/Filament.podspec

@@ -1,12 +1,12 @@
 Pod::Spec.new do |spec|
   spec.name = "Filament"
-  spec.version = "1.28.1"
+  spec.version = "1.28.2"
   spec.license = { :type => "Apache 2.0", :file => "LICENSE" }
   spec.homepage = "https://google.github.io/filament"
   spec.authors = "Google LLC."
   spec.summary = "Filament is a real-time physically based rendering engine for Android, iOS, Windows, Linux, macOS, and WASM/WebGL."
   spec.platform = :ios, "11.0"
-  spec.source = { :http => "https://github.com/google/filament/releases/download/v1.28.1/filament-v1.28.1-ios.tgz" }
+  spec.source = { :http => "https://github.com/google/filament/releases/download/v1.28.2/filament-v1.28.2-ios.tgz" }
 
   # Fix linking error with Xcode 12; we do not yet support the simulator on Apple silicon.
   spec.pod_target_xcconfig = {

ios/samples/backend-test/backend-test.xcodeproj/xcshareddata/xcschemes/backend-test Metal.xcscheme

@@ -27,10 +27,8 @@
       buildConfiguration = "Metal Debug"
       selectedDebuggerIdentifier = "Xcode.DebuggerFoundation.Debugger.LLDB"
       selectedLauncherIdentifier = "Xcode.DebuggerFoundation.Launcher.LLDB"
-      onlyGenerateCoverageForSpecifiedTargets = "NO"
-      shouldUseLaunchSchemeArgsEnv = "YES">
-      <Testables>
-      </Testables>
+      shouldUseLaunchSchemeArgsEnv = "YES"
+      onlyGenerateCoverageForSpecifiedTargets = "NO">
       <MacroExpansion>
          <BuildableReference
             BuildableIdentifier = "primary"
@@ -40,6 +38,8 @@
             ReferencedContainer = "container:backend-test.xcodeproj">
          </BuildableReference>
       </MacroExpansion>
+      <Testables>
+      </Testables>
       <CommandLineArguments>
       </CommandLineArguments>
    </TestAction>

ios/samples/backend-test/backend-test.xcodeproj/xcshareddata/xcschemes/backend-test OpenGL.xcscheme

@@ -27,10 +27,8 @@
       buildConfiguration = "OpenGL Debug"
       selectedDebuggerIdentifier = "Xcode.DebuggerFoundation.Debugger.LLDB"
       selectedLauncherIdentifier = "Xcode.DebuggerFoundation.Launcher.LLDB"
-      onlyGenerateCoverageForSpecifiedTargets = "NO"
-      shouldUseLaunchSchemeArgsEnv = "YES">
-      <Testables>
-      </Testables>
+      shouldUseLaunchSchemeArgsEnv = "YES"
+      onlyGenerateCoverageForSpecifiedTargets = "NO">
       <MacroExpansion>
          <BuildableReference
             BuildableIdentifier = "primary"
@@ -40,6 +38,8 @@
             ReferencedContainer = "container:backend-test.xcodeproj">
          </BuildableReference>
       </MacroExpansion>
+      <Testables>
+      </Testables>
       <CommandLineArguments>
       </CommandLineArguments>
    </TestAction>

ios/samples/gltf-viewer/gltf-viewer.xcodeproj/project.pbxproj

@@ -123,7 +123,7 @@
 				};
 			};
 			buildConfigurationList = 90327DE42AE5DF12C4BD2113 /* Build configuration list for PBXProject "gltf-viewer" */;
-			compatibilityVersion = "Xcode 10.0";
+			compatibilityVersion = "Xcode 11.0";
 			developmentRegion = en;
 			hasScannedForEncodings = 0;
 			knownRegions = (
@@ -175,7 +175,7 @@
 			);
 			runOnlyForDeploymentPostprocessing = 0;
 			shellPath = /bin/sh;
-			shellScript = "#/usr/bin/env/bash\n\nset -e\n\n# Compile resources.\n\n# The gltf-viewer app requires two resources:\n# 1. The IBL image\n# 2. The skybox image\n\nHOST_TOOLS_PATH=\"${HOST_TOOLS_PATH:-../../../out/release/filament/bin}\"\n\ncmgen_path=`find ${HOST_TOOLS_PATH} -name cmgen -type f | head -n 1`\n\n# Ensure that the required tools are present in the out/ directory.\n# These can be built by running ./build.sh -p desktop -i release at Filament's root directory.\n\nif [[ ! -e \"${cmgen_path}\" ]]; then\n  echo \"No cmgen binary could be found in ${HOST_TOOLS_PATH}.\"\n  echo \"Ensure Filament has been built/installed before building this app.\"\n  exit 1\nfi\n\n# cmgen consumes an HDR environment map and generates two mipmapped KTX files (IBL and skybox)\n\"${cmgen_path}\" \\\n    --deploy=\"${PROJECT_DIR}/generated/default_env\" \\\n    --format=ktx --size=256 --extract-blur=0.1 \\\n    \"${PROJECT_DIR}/../../../third_party/environments/lightroom_14b.hdr\"\n";
+			shellScript = "#/usr/bin/env/bash\n\nset -e\n\n# Compile resources.\n\n# The gltf-viewer app requires two resources:\n# 1. The IBL image\n# 2. The skybox image\n\nHOST_TOOLS_PATH=\"${HOST_TOOLS_PATH:-../../../out/release/filament/bin}\"\n\ncmgen_path=`find ${HOST_TOOLS_PATH} -name cmgen -type f | head -n 1`\n\n# Ensure that the required tools are present in the out/ directory.\n# These can be built by running ./build.sh -p desktop -i release at Filament's root directory.\n\nif [[ ! -e \"${cmgen_path}\" ]]; then\n  echo \"No cmgen binary could be found in ${HOST_TOOLS_PATH}.\"\n  echo \"Ensure Filament has been built/installed before building this app.\"\n  exit 1\nfi\n\n# cmgen consumes an HDR environment map and generates two mipmapped KTX files (IBL and skybox)\n\"${cmgen_path}\" \\\n    --quiet \\\n    --deploy=\"${PROJECT_DIR}/generated/default_env\" \\\n    --format=ktx --size=256 --extract-blur=0.1 \\\n    \"${PROJECT_DIR}/../../../third_party/environments/lightroom_14b.hdr\"\n";
 		};
 /* End PBXShellScriptBuildPhase section */
 
@@ -429,6 +429,7 @@
 					"-lzstd",
 					"-lcamutils",
 					"-ldracodec",
+					"-lmeshoptimizer",
 					"-lviewer",
 					"-lcivetweb",
 					"-lktxreader",
@@ -485,6 +486,7 @@
 					"-lzstd",
 					"-lcamutils",
 					"-ldracodec",
+					"-lmeshoptimizer",
 					"-lviewer",
 					"-lcivetweb",
 					"-lktxreader",
@@ -603,6 +605,7 @@
 					"-lzstd",
 					"-lcamutils",
 					"-ldracodec",
+					"-lmeshoptimizer",
 					"-lviewer",
 					"-lcivetweb",
 					"-lktxreader",
@@ -659,6 +662,7 @@
 					"-lzstd",
 					"-lcamutils",
 					"-ldracodec",
+					"-lmeshoptimizer",
 					"-lviewer",
 					"-lcivetweb",
 					"-lktxreader",

ios/samples/gltf-viewer/gltf-viewer.xcodeproj/xcshareddata/xcschemes/gltf-viewer Metal.xcscheme

@@ -1,7 +1,7 @@
 <?xml version="1.0" encoding="UTF-8"?>
 <Scheme
    LastUpgradeVersion = "1200"
-   version = "1.3">
+   version = "1.7">
    <BuildAction
       parallelizeBuildables = "YES"
       buildImplicitDependencies = "YES"
@@ -27,10 +27,8 @@
       buildConfiguration = "Metal Debug"
       selectedDebuggerIdentifier = "Xcode.DebuggerFoundation.Debugger.LLDB"
       selectedLauncherIdentifier = "Xcode.DebuggerFoundation.Launcher.LLDB"
-      onlyGenerateCoverageForSpecifiedTargets = "NO"
-      shouldUseLaunchSchemeArgsEnv = "YES">
-      <Testables>
-      </Testables>
+      shouldUseLaunchSchemeArgsEnv = "YES"
+      onlyGenerateCoverageForSpecifiedTargets = "NO">
       <MacroExpansion>
          <BuildableReference
             BuildableIdentifier = "primary"
@@ -40,6 +38,8 @@
             ReferencedContainer = "container:gltf-viewer.xcodeproj">
          </BuildableReference>
       </MacroExpansion>
+      <Testables>
+      </Testables>
       <CommandLineArguments>
       </CommandLineArguments>
    </TestAction>

ios/samples/gltf-viewer/gltf-viewer.xcodeproj/xcshareddata/xcschemes/gltf-viewer OpenGL.xcscheme

@@ -1,7 +1,7 @@
 <?xml version="1.0" encoding="UTF-8"?>
 <Scheme
    LastUpgradeVersion = "1200"
-   version = "1.3">
+   version = "1.7">
    <BuildAction
       parallelizeBuildables = "YES"
       buildImplicitDependencies = "YES"
@@ -27,10 +27,8 @@
       buildConfiguration = "OpenGL Debug"
       selectedDebuggerIdentifier = "Xcode.DebuggerFoundation.Debugger.LLDB"
       selectedLauncherIdentifier = "Xcode.DebuggerFoundation.Launcher.LLDB"
-      onlyGenerateCoverageForSpecifiedTargets = "NO"
-      shouldUseLaunchSchemeArgsEnv = "YES">
-      <Testables>
-      </Testables>
+      shouldUseLaunchSchemeArgsEnv = "YES"
+      onlyGenerateCoverageForSpecifiedTargets = "NO">
       <MacroExpansion>
          <BuildableReference
             BuildableIdentifier = "primary"
@@ -40,6 +38,8 @@
             ReferencedContainer = "container:gltf-viewer.xcodeproj">
          </BuildableReference>
       </MacroExpansion>
+      <Testables>
+      </Testables>
       <CommandLineArguments>
       </CommandLineArguments>
    </TestAction>

ios/samples/gltf-viewer/project.yml

@@ -20,7 +20,8 @@ targets:
         settings:
             base:
                 OTHER_LDFLAGS: ["-lgltfio_core", "-luberarchive", "-limage", "-lgeometry", "-lstb", "-lbasis_transcoder",
-                                "-lzstd", "-lcamutils", "-ldracodec", "-lviewer", "-lcivetweb", "-lktxreader", "-luberzlib"]
+                                "-lzstd", "-lcamutils", "-ldracodec", "-lmeshoptimizer", "-lviewer", "-lcivetweb", "-lktxreader",
+                                "-luberzlib"]
         preBuildScripts:
             - path: build-resources.sh
               name: Build Resources

ios/samples/hello-ar/hello-ar.xcodeproj/project.pbxproj

@@ -165,7 +165,7 @@
 				};
 			};
 			buildConfigurationList = 281E763643B2F1F3B903864B /* Build configuration list for PBXProject "hello-ar" */;
-			compatibilityVersion = "Xcode 10.0";
+			compatibilityVersion = "Xcode 11.0";
 			developmentRegion = en;
 			hasScannedForEncodings = 0;
 			knownRegions = (

ios/samples/hello-ar/hello-ar.xcodeproj/xcshareddata/xcschemes/hello-ar Metal.xcscheme

@@ -1,7 +1,7 @@
 <?xml version="1.0" encoding="UTF-8"?>
 <Scheme
    LastUpgradeVersion = "1200"
-   version = "1.3">
+   version = "1.7">
    <BuildAction
       parallelizeBuildables = "YES"
       buildImplicitDependencies = "YES"
@@ -27,10 +27,8 @@
       buildConfiguration = "Metal Debug"
       selectedDebuggerIdentifier = "Xcode.DebuggerFoundation.Debugger.LLDB"
       selectedLauncherIdentifier = "Xcode.DebuggerFoundation.Launcher.LLDB"
-      onlyGenerateCoverageForSpecifiedTargets = "NO"
-      shouldUseLaunchSchemeArgsEnv = "YES">
-      <Testables>
-      </Testables>
+      shouldUseLaunchSchemeArgsEnv = "YES"
+      onlyGenerateCoverageForSpecifiedTargets = "NO">
       <MacroExpansion>
          <BuildableReference
             BuildableIdentifier = "primary"
@@ -40,6 +38,8 @@
             ReferencedContainer = "container:hello-ar.xcodeproj">
          </BuildableReference>
       </MacroExpansion>
+      <Testables>
+      </Testables>
       <CommandLineArguments>
       </CommandLineArguments>
    </TestAction>

ios/samples/hello-ar/hello-ar.xcodeproj/xcshareddata/xcschemes/hello-ar OpenGL.xcscheme

@@ -1,7 +1,7 @@
 <?xml version="1.0" encoding="UTF-8"?>
 <Scheme
    LastUpgradeVersion = "1200"
-   version = "1.3">
+   version = "1.7">
    <BuildAction
       parallelizeBuildables = "YES"
       buildImplicitDependencies = "YES"
@@ -27,10 +27,8 @@
       buildConfiguration = "OpenGL Debug"
       selectedDebuggerIdentifier = "Xcode.DebuggerFoundation.Debugger.LLDB"
       selectedLauncherIdentifier = "Xcode.DebuggerFoundation.Launcher.LLDB"
-      onlyGenerateCoverageForSpecifiedTargets = "NO"
-      shouldUseLaunchSchemeArgsEnv = "YES">
-      <Testables>
-      </Testables>
+      shouldUseLaunchSchemeArgsEnv = "YES"
+      onlyGenerateCoverageForSpecifiedTargets = "NO">
       <MacroExpansion>
          <BuildableReference
             BuildableIdentifier = "primary"
@@ -40,6 +38,8 @@
             ReferencedContainer = "container:hello-ar.xcodeproj">
          </BuildableReference>
       </MacroExpansion>
+      <Testables>
+      </Testables>
       <CommandLineArguments>
       </CommandLineArguments>
    </TestAction>

ios/samples/hello-gltf/hello-gltf.xcodeproj/project.pbxproj

@@ -138,7 +138,7 @@
 				};
 			};
 			buildConfigurationList = C7514DDCFDA3C5F803E239F9 /* Build configuration list for PBXProject "hello-gltf" */;
-			compatibilityVersion = "Xcode 10.0";
+			compatibilityVersion = "Xcode 11.0";
 			developmentRegion = en;
 			hasScannedForEncodings = 0;
 			knownRegions = (
@@ -328,6 +328,8 @@
 					"-limage",
 					"-lgeometry",
 					"-ldracodec",
+					"-lmeshoptimizer",
+					"-lbasis_transcoder",
 					"-lktxreader",
 					"-lstb",
 					"-lzstd",
@@ -380,6 +382,8 @@
 					"-limage",
 					"-lgeometry",
 					"-ldracodec",
+					"-lmeshoptimizer",
+					"-lbasis_transcoder",
 					"-lktxreader",
 					"-lstb",
 					"-lzstd",
@@ -604,6 +608,8 @@
 					"-limage",
 					"-lgeometry",
 					"-ldracodec",
+					"-lmeshoptimizer",
+					"-lbasis_transcoder",
 					"-lktxreader",
 					"-lstb",
 					"-lzstd",
@@ -656,6 +662,8 @@
 					"-limage",
 					"-lgeometry",
 					"-ldracodec",
+					"-lmeshoptimizer",
+					"-lbasis_transcoder",
 					"-lktxreader",
 					"-lstb",
 					"-lzstd",

ios/samples/hello-gltf/hello-gltf.xcodeproj/xcshareddata/xcschemes/hello-gltf Metal.xcscheme

@@ -1,7 +1,7 @@
 <?xml version="1.0" encoding="UTF-8"?>
 <Scheme
    LastUpgradeVersion = "1200"
-   version = "1.3">
+   version = "1.7">
    <BuildAction
       parallelizeBuildables = "YES"
       buildImplicitDependencies = "YES"
@@ -27,10 +27,8 @@
       buildConfiguration = "Metal Debug"
       selectedDebuggerIdentifier = "Xcode.DebuggerFoundation.Debugger.LLDB"
       selectedLauncherIdentifier = "Xcode.DebuggerFoundation.Launcher.LLDB"
-      onlyGenerateCoverageForSpecifiedTargets = "NO"
-      shouldUseLaunchSchemeArgsEnv = "YES">
-      <Testables>
-      </Testables>
+      shouldUseLaunchSchemeArgsEnv = "YES"
+      onlyGenerateCoverageForSpecifiedTargets = "NO">
       <MacroExpansion>
          <BuildableReference
             BuildableIdentifier = "primary"
@@ -40,6 +38,8 @@
             ReferencedContainer = "container:hello-gltf.xcodeproj">
          </BuildableReference>
       </MacroExpansion>
+      <Testables>
+      </Testables>
       <CommandLineArguments>
       </CommandLineArguments>
    </TestAction>

ios/samples/hello-gltf/hello-gltf.xcodeproj/xcshareddata/xcschemes/hello-gltf OpenGL.xcscheme

@@ -1,7 +1,7 @@
 <?xml version="1.0" encoding="UTF-8"?>
 <Scheme
    LastUpgradeVersion = "1200"
-   version = "1.3">
+   version = "1.7">
    <BuildAction
       parallelizeBuildables = "YES"
       buildImplicitDependencies = "YES"
@@ -27,10 +27,8 @@
       buildConfiguration = "OpenGL Debug"
       selectedDebuggerIdentifier = "Xcode.DebuggerFoundation.Debugger.LLDB"
       selectedLauncherIdentifier = "Xcode.DebuggerFoundation.Launcher.LLDB"
-      onlyGenerateCoverageForSpecifiedTargets = "NO"
-      shouldUseLaunchSchemeArgsEnv = "YES">
-      <Testables>
-      </Testables>
+      shouldUseLaunchSchemeArgsEnv = "YES"
+      onlyGenerateCoverageForSpecifiedTargets = "NO">
       <MacroExpansion>
          <BuildableReference
             BuildableIdentifier = "primary"
@@ -40,6 +38,8 @@
             ReferencedContainer = "container:hello-gltf.xcodeproj">
          </BuildableReference>
       </MacroExpansion>
+      <Testables>
+      </Testables>
       <CommandLineArguments>
       </CommandLineArguments>
    </TestAction>

ios/samples/hello-gltf/hello-gltf/FilamentView/App.cpp

@@ -23,6 +23,7 @@
 #include <ktxreader/Ktx1Reader.h>
 #include <gltfio/AssetLoader.h>
 #include <gltfio/ResourceLoader.h>
+#include <gltfio/TextureProvider.h>
 #include <gltfio/materials/uberarchive.h>
 
 #include <fstream>
@@ -139,10 +140,21 @@ void App::setupMesh() {
     }
     app.asset = app.assetLoader->createAsset(buffer.data(), static_cast<uint32_t>(size));
 
-    filament::gltfio::ResourceLoader({
+    auto resourceLoader = new filament::gltfio::ResourceLoader({
         .engine = engine,
         .normalizeSkinningWeights = true
-    }).loadResources(app.asset);
+    });
+    auto stbDecoder = filament::gltfio::createStbProvider(engine);
+    auto ktxDecoder = filament::gltfio::createKtx2Provider(engine);
+
+    resourceLoader->addTextureProvider("image/png", stbDecoder);
+    resourceLoader->addTextureProvider("image/jpeg", stbDecoder);
+    resourceLoader->addTextureProvider("image/ktx2", ktxDecoder);
+    resourceLoader->loadResources(app.asset);
+
+    delete resourceLoader;
+    delete stbDecoder;
+    delete ktxDecoder;
 
     scene->addEntities(app.asset->getEntities(), app.asset->getEntityCount());
 }

ios/samples/hello-gltf/project.yml

@@ -12,7 +12,8 @@ targets:
             - FilamentApp
         settings:
             base:
-                OTHER_LDFLAGS: ["-lgltfio_core", "-luberarchive", "-limage", "-lgeometry", "-ldracodec", "-lktxreader", "-lstb", "-lzstd", "-luberzlib"]
+                OTHER_LDFLAGS: ["-lgltfio_core", "-luberarchive", "-limage", "-lgeometry", "-ldracodec", "-lmeshoptimizer",
+                                "-lbasis_transcoder", "-lktxreader", "-lstb", "-lzstd", "-luberzlib"]
         preBuildScripts:
             - path: build-resources.sh
               name: Build Resources

ios/samples/hello-pbr/hello-pbr.xcodeproj/project.pbxproj

@@ -120,7 +120,7 @@
 				};
 			};
 			buildConfigurationList = 0479354FB8D9A411347D91A6 /* Build configuration list for PBXProject "hello-pbr" */;
-			compatibilityVersion = "Xcode 10.0";
+			compatibilityVersion = "Xcode 11.0";
 			developmentRegion = en;
 			hasScannedForEncodings = 0;
 			knownRegions = (

ios/samples/hello-pbr/hello-pbr.xcodeproj/xcshareddata/xcschemes/hello-pbr Metal.xcscheme

@@ -1,7 +1,7 @@
 <?xml version="1.0" encoding="UTF-8"?>
 <Scheme
    LastUpgradeVersion = "1200"
-   version = "1.3">
+   version = "1.7">
    <BuildAction
       parallelizeBuildables = "YES"
       buildImplicitDependencies = "YES"
@@ -27,10 +27,8 @@
       buildConfiguration = "Metal Debug"
       selectedDebuggerIdentifier = "Xcode.DebuggerFoundation.Debugger.LLDB"
       selectedLauncherIdentifier = "Xcode.DebuggerFoundation.Launcher.LLDB"
-      onlyGenerateCoverageForSpecifiedTargets = "NO"
-      shouldUseLaunchSchemeArgsEnv = "YES">
-      <Testables>
-      </Testables>
+      shouldUseLaunchSchemeArgsEnv = "YES"
+      onlyGenerateCoverageForSpecifiedTargets = "NO">
       <MacroExpansion>
          <BuildableReference
             BuildableIdentifier = "primary"
@@ -40,6 +38,8 @@
             ReferencedContainer = "container:hello-pbr.xcodeproj">
          </BuildableReference>
       </MacroExpansion>
+      <Testables>
+      </Testables>
       <CommandLineArguments>
       </CommandLineArguments>
    </TestAction>

ios/samples/hello-pbr/hello-pbr.xcodeproj/xcshareddata/xcschemes/hello-pbr OpenGL.xcscheme

@@ -1,7 +1,7 @@
 <?xml version="1.0" encoding="UTF-8"?>
 <Scheme
    LastUpgradeVersion = "1200"
-   version = "1.3">
+   version = "1.7">
    <BuildAction
       parallelizeBuildables = "YES"
       buildImplicitDependencies = "YES"
@@ -27,10 +27,8 @@
       buildConfiguration = "OpenGL Debug"
       selectedDebuggerIdentifier = "Xcode.DebuggerFoundation.Debugger.LLDB"
       selectedLauncherIdentifier = "Xcode.DebuggerFoundation.Launcher.LLDB"
-      onlyGenerateCoverageForSpecifiedTargets = "NO"
-      shouldUseLaunchSchemeArgsEnv = "YES">
-      <Testables>
-      </Testables>
+      shouldUseLaunchSchemeArgsEnv = "YES"
+      onlyGenerateCoverageForSpecifiedTargets = "NO">
       <MacroExpansion>
          <BuildableReference
             BuildableIdentifier = "primary"
@@ -40,6 +38,8 @@
             ReferencedContainer = "container:hello-pbr.xcodeproj">
          </BuildableReference>
       </MacroExpansion>
+      <Testables>
+      </Testables>
       <CommandLineArguments>
       </CommandLineArguments>
    </TestAction>

ios/samples/hello-triangle/hello-triangle.xcodeproj/xcshareddata/xcschemes/hello-triangle Metal.xcscheme

@@ -27,10 +27,8 @@
       buildConfiguration = "Metal Debug"
       selectedDebuggerIdentifier = "Xcode.DebuggerFoundation.Debugger.LLDB"
       selectedLauncherIdentifier = "Xcode.DebuggerFoundation.Launcher.LLDB"
-      onlyGenerateCoverageForSpecifiedTargets = "NO"
-      shouldUseLaunchSchemeArgsEnv = "YES">
-      <Testables>
-      </Testables>
+      shouldUseLaunchSchemeArgsEnv = "YES"
+      onlyGenerateCoverageForSpecifiedTargets = "NO">
       <MacroExpansion>
          <BuildableReference
             BuildableIdentifier = "primary"
@@ -40,6 +38,8 @@
             ReferencedContainer = "container:hello-triangle.xcodeproj">
          </BuildableReference>
       </MacroExpansion>
+      <Testables>
+      </Testables>
       <CommandLineArguments>
       </CommandLineArguments>
    </TestAction>

ios/samples/hello-triangle/hello-triangle.xcodeproj/xcshareddata/xcschemes/hello-triangle OpenGL.xcscheme

@@ -27,10 +27,8 @@
       buildConfiguration = "OpenGL Debug"
       selectedDebuggerIdentifier = "Xcode.DebuggerFoundation.Debugger.LLDB"
       selectedLauncherIdentifier = "Xcode.DebuggerFoundation.Launcher.LLDB"
-      onlyGenerateCoverageForSpecifiedTargets = "NO"
-      shouldUseLaunchSchemeArgsEnv = "YES">
-      <Testables>
-      </Testables>
+      shouldUseLaunchSchemeArgsEnv = "YES"
+      onlyGenerateCoverageForSpecifiedTargets = "NO">
       <MacroExpansion>
          <BuildableReference
             BuildableIdentifier = "primary"
@@ -40,6 +38,8 @@
             ReferencedContainer = "container:hello-triangle.xcodeproj">
          </BuildableReference>
       </MacroExpansion>
+      <Testables>
+      </Testables>
       <CommandLineArguments>
       </CommandLineArguments>
    </TestAction>

ios/samples/transparent-rendering/transparent-rendering.xcodeproj/xcshareddata/xcschemes/transparent-rendering Metal.xcscheme

@@ -27,10 +27,8 @@
       buildConfiguration = "Metal Debug"
       selectedDebuggerIdentifier = "Xcode.DebuggerFoundation.Debugger.LLDB"
       selectedLauncherIdentifier = "Xcode.DebuggerFoundation.Launcher.LLDB"
-      onlyGenerateCoverageForSpecifiedTargets = "NO"
-      shouldUseLaunchSchemeArgsEnv = "YES">
-      <Testables>
-      </Testables>
+      shouldUseLaunchSchemeArgsEnv = "YES"
+      onlyGenerateCoverageForSpecifiedTargets = "NO">
       <MacroExpansion>
          <BuildableReference
             BuildableIdentifier = "primary"
@@ -40,6 +38,8 @@
             ReferencedContainer = "container:transparent-rendering.xcodeproj">
          </BuildableReference>
       </MacroExpansion>
+      <Testables>
+      </Testables>
       <CommandLineArguments>
       </CommandLineArguments>
    </TestAction>

ios/samples/transparent-rendering/transparent-rendering.xcodeproj/xcshareddata/xcschemes/transparent-rendering OpenGL.xcscheme

@@ -27,10 +27,8 @@
       buildConfiguration = "OpenGL Debug"
       selectedDebuggerIdentifier = "Xcode.DebuggerFoundation.Debugger.LLDB"
       selectedLauncherIdentifier = "Xcode.DebuggerFoundation.Launcher.LLDB"
-      onlyGenerateCoverageForSpecifiedTargets = "NO"
-      shouldUseLaunchSchemeArgsEnv = "YES">
-      <Testables>
-      </Testables>
+      shouldUseLaunchSchemeArgsEnv = "YES"
+      onlyGenerateCoverageForSpecifiedTargets = "NO">
       <MacroExpansion>
          <BuildableReference
             BuildableIdentifier = "primary"
@@ -40,6 +38,8 @@
             ReferencedContainer = "container:transparent-rendering.xcodeproj">
          </BuildableReference>
       </MacroExpansion>
+      <Testables>
+      </Testables>
       <CommandLineArguments>
       </CommandLineArguments>
    </TestAction>

libs/bluevk/README.md

@@ -0,0 +1,17 @@
+## Updating Vulkan headers
+
+To update the Vulkan headers, perform the following steps.
+
+First, find the latest version of the Vulkan headers here:
+https://github.com/KhronosGroup/Vulkan-Headers/tags
+
+Replace `v1.3.232` with the latest version of the headers in the following commands.
+```
+cd libs/bluevk
+curl -OL https://github.com/KhronosGroup/Vulkan-Headers/archive/refs/tags/v1.3.232.zip
+unzip v1.3.232.zip
+rsync -r Vulkan-Headers-1.3.232/include/vulkan/ include/vulkan --delete
+rsync -r Vulkan-Headers-1.3.232/include/vk_video/ include/vk_video --delete
+rm include/vulkan/*.hpp
+rm -r Vulkan-Headers-1.3.232 v1.3.232.zip
+```

libs/bluevk/include/bluevk/BlueVK.h

@@ -37,6 +37,27 @@
     #ifndef VK_NO_PROTOTYPES
     #define VK_NO_PROTOTYPES
     #endif
+
+    #if defined(__ANDROID__)
+    #define VK_USE_PLATFORM_ANDROID_KHR 1
+    #elif defined(IOS)
+    #define VK_USE_PLATFORM_IOS_MVK 1
+    #elif defined(__linux__)
+    #if defined(FILAMENT_SUPPORTS_XCB)
+    #define VK_USE_PLATFORM_XCB_KHR 1
+    #endif
+    #if defined(FILAMENT_SUPPORTS_XLIB)
+    #define VK_USE_PLATFORM_XLIB_KHR 1
+    #endif
+    #if defined(FILAMENT_SUPPORTS_WAYLAND)
+    #define VK_USE_PLATFORM_WAYLAND_KHR 1
+    #endif
+    #elif defined(__APPLE__)
+    #define VK_USE_PLATFORM_MACOS_MVK 1
+    #elif defined(WIN32)
+    #define VK_USE_PLATFORM_WIN32_KHR 1
+    #endif
+
     #include <vulkan/vulkan.h>
 #endif
 

libs/bluevk/include/vk_video/vulkan_video_codec_h264std.h

@@ -21,16 +21,13 @@ extern "C" {
 
 #define vulkan_video_codec_h264std 1
 #include <stdint.h>
-// Vulkan 0.9 provisional Vulkan video H.264 encode and decode std specification version number
-#define VK_STD_VULKAN_VIDEO_CODEC_H264_API_VERSION_0_9_5 VK_MAKE_VIDEO_STD_VERSION(0, 9, 5) // Patch version should always be set to 0
-
 #define STD_VIDEO_H264_CPB_CNT_LIST_SIZE  32
 #define STD_VIDEO_H264_SCALING_LIST_4X4_NUM_LISTS 6
 #define STD_VIDEO_H264_SCALING_LIST_4X4_NUM_ELEMENTS 16
-#define STD_VIDEO_H264_SCALING_LIST_8X8_NUM_LISTS 2
+#define STD_VIDEO_H264_SCALING_LIST_8X8_NUM_LISTS 6
 #define STD_VIDEO_H264_SCALING_LIST_8X8_NUM_ELEMENTS 64
-#define VK_STD_VULKAN_VIDEO_CODEC_H264_SPEC_VERSION VK_STD_VULKAN_VIDEO_CODEC_H264_API_VERSION_0_9_5
-#define VK_STD_VULKAN_VIDEO_CODEC_H264_EXTENSION_NAME "VK_STD_vulkan_video_codec_h264"
+#define STD_VIDEO_H264_MAX_NUM_LIST_REF   32
+#define STD_VIDEO_H264_MAX_CHROMA_PLANES  2
 
 typedef enum StdVideoH264ChromaFormatIdc {
     STD_VIDEO_H264_CHROMA_FORMAT_IDC_MONOCHROME = 0,
@@ -50,29 +47,29 @@ typedef enum StdVideoH264ProfileIdc {
     STD_VIDEO_H264_PROFILE_IDC_MAX_ENUM = 0x7FFFFFFF
 } StdVideoH264ProfileIdc;
 
-typedef enum StdVideoH264Level {
-    STD_VIDEO_H264_LEVEL_1_0 = 0,
-    STD_VIDEO_H264_LEVEL_1_1 = 1,
-    STD_VIDEO_H264_LEVEL_1_2 = 2,
-    STD_VIDEO_H264_LEVEL_1_3 = 3,
-    STD_VIDEO_H264_LEVEL_2_0 = 4,
-    STD_VIDEO_H264_LEVEL_2_1 = 5,
-    STD_VIDEO_H264_LEVEL_2_2 = 6,
-    STD_VIDEO_H264_LEVEL_3_0 = 7,
-    STD_VIDEO_H264_LEVEL_3_1 = 8,
-    STD_VIDEO_H264_LEVEL_3_2 = 9,
-    STD_VIDEO_H264_LEVEL_4_0 = 10,
-    STD_VIDEO_H264_LEVEL_4_1 = 11,
-    STD_VIDEO_H264_LEVEL_4_2 = 12,
-    STD_VIDEO_H264_LEVEL_5_0 = 13,
-    STD_VIDEO_H264_LEVEL_5_1 = 14,
-    STD_VIDEO_H264_LEVEL_5_2 = 15,
-    STD_VIDEO_H264_LEVEL_6_0 = 16,
-    STD_VIDEO_H264_LEVEL_6_1 = 17,
-    STD_VIDEO_H264_LEVEL_6_2 = 18,
-    STD_VIDEO_H264_LEVEL_INVALID = 0x7FFFFFFF,
-    STD_VIDEO_H264_LEVEL_MAX_ENUM = 0x7FFFFFFF
-} StdVideoH264Level;
+typedef enum StdVideoH264LevelIdc {
+    STD_VIDEO_H264_LEVEL_IDC_1_0 = 0,
+    STD_VIDEO_H264_LEVEL_IDC_1_1 = 1,
+    STD_VIDEO_H264_LEVEL_IDC_1_2 = 2,
+    STD_VIDEO_H264_LEVEL_IDC_1_3 = 3,
+    STD_VIDEO_H264_LEVEL_IDC_2_0 = 4,
+    STD_VIDEO_H264_LEVEL_IDC_2_1 = 5,
+    STD_VIDEO_H264_LEVEL_IDC_2_2 = 6,
+    STD_VIDEO_H264_LEVEL_IDC_3_0 = 7,
+    STD_VIDEO_H264_LEVEL_IDC_3_1 = 8,
+    STD_VIDEO_H264_LEVEL_IDC_3_2 = 9,
+    STD_VIDEO_H264_LEVEL_IDC_4_0 = 10,
+    STD_VIDEO_H264_LEVEL_IDC_4_1 = 11,
+    STD_VIDEO_H264_LEVEL_IDC_4_2 = 12,
+    STD_VIDEO_H264_LEVEL_IDC_5_0 = 13,
+    STD_VIDEO_H264_LEVEL_IDC_5_1 = 14,
+    STD_VIDEO_H264_LEVEL_IDC_5_2 = 15,
+    STD_VIDEO_H264_LEVEL_IDC_6_0 = 16,
+    STD_VIDEO_H264_LEVEL_IDC_6_1 = 17,
+    STD_VIDEO_H264_LEVEL_IDC_6_2 = 18,
+    STD_VIDEO_H264_LEVEL_IDC_INVALID = 0x7FFFFFFF,
+    STD_VIDEO_H264_LEVEL_IDC_MAX_ENUM = 0x7FFFFFFF
+} StdVideoH264LevelIdc;
 
 typedef enum StdVideoH264PocType {
     STD_VIDEO_H264_POC_TYPE_0 = 0,
@@ -197,6 +194,7 @@ typedef struct StdVideoH264HrdParameters {
     uint8_t     cpb_cnt_minus1;
     uint8_t     bit_rate_scale;
     uint8_t     cpb_size_scale;
+    uint8_t     reserved1;
     uint32_t    bit_rate_value_minus1[STD_VIDEO_H264_CPB_CNT_LIST_SIZE];
     uint32_t    cpb_size_value_minus1[STD_VIDEO_H264_CPB_CNT_LIST_SIZE];
     uint8_t     cbr_flag[STD_VIDEO_H264_CPB_CNT_LIST_SIZE];
@@ -207,19 +205,22 @@ typedef struct StdVideoH264HrdParameters {
 } StdVideoH264HrdParameters;
 
 typedef struct StdVideoH264SequenceParameterSetVui {
-    StdVideoH264AspectRatioIdc    aspect_ratio_idc;
-    uint16_t                      sar_width;
-    uint16_t                      sar_height;
-    uint8_t                       video_format;
-    uint8_t                       color_primaries;
-    uint8_t                       transfer_characteristics;
-    uint8_t                       matrix_coefficients;
-    uint32_t                      num_units_in_tick;
-    uint32_t                      time_scale;
-    StdVideoH264HrdParameters*    pHrdParameters;
-    uint8_t                       max_num_reorder_frames;
-    uint8_t                       max_dec_frame_buffering;
-    StdVideoH264SpsVuiFlags       flags;
+    StdVideoH264SpsVuiFlags             flags;
+    StdVideoH264AspectRatioIdc          aspect_ratio_idc;
+    uint16_t                            sar_width;
+    uint16_t                            sar_height;
+    uint8_t                             video_format;
+    uint8_t                             colour_primaries;
+    uint8_t                             transfer_characteristics;
+    uint8_t                             matrix_coefficients;
+    uint32_t                            num_units_in_tick;
+    uint32_t                            time_scale;
+    uint8_t                             max_num_reorder_frames;
+    uint8_t                             max_dec_frame_buffering;
+    uint8_t                             chroma_sample_loc_type_top_field;
+    uint8_t                             chroma_sample_loc_type_bottom_field;
+    uint32_t                            reserved1;
+    const StdVideoH264HrdParameters*    pHrdParameters;
 } StdVideoH264SequenceParameterSetVui;
 
 typedef struct StdVideoH264SpsFlags {
@@ -242,36 +243,38 @@ typedef struct StdVideoH264SpsFlags {
 } StdVideoH264SpsFlags;
 
 typedef struct StdVideoH264ScalingLists {
-    uint8_t    scaling_list_present_mask;
-    uint8_t    use_default_scaling_matrix_mask;
-    uint8_t    ScalingList4x4[STD_VIDEO_H264_SCALING_LIST_4X4_NUM_LISTS][STD_VIDEO_H264_SCALING_LIST_4X4_NUM_ELEMENTS];
-    uint8_t    ScalingList8x8[STD_VIDEO_H264_SCALING_LIST_8X8_NUM_LISTS][STD_VIDEO_H264_SCALING_LIST_8X8_NUM_ELEMENTS];
+    uint16_t    scaling_list_present_mask;
+    uint16_t    use_default_scaling_matrix_mask;
+    uint8_t     ScalingList4x4[STD_VIDEO_H264_SCALING_LIST_4X4_NUM_LISTS][STD_VIDEO_H264_SCALING_LIST_4X4_NUM_ELEMENTS];
+    uint8_t     ScalingList8x8[STD_VIDEO_H264_SCALING_LIST_8X8_NUM_LISTS][STD_VIDEO_H264_SCALING_LIST_8X8_NUM_ELEMENTS];
 } StdVideoH264ScalingLists;
 
 typedef struct StdVideoH264SequenceParameterSet {
-    StdVideoH264ProfileIdc                  profile_idc;
-    StdVideoH264Level                       level_idc;
-    uint8_t                                 seq_parameter_set_id;
-    StdVideoH264ChromaFormatIdc             chroma_format_idc;
-    uint8_t                                 bit_depth_luma_minus8;
-    uint8_t                                 bit_depth_chroma_minus8;
-    uint8_t                                 log2_max_frame_num_minus4;
-    StdVideoH264PocType                     pic_order_cnt_type;
-    uint8_t                                 log2_max_pic_order_cnt_lsb_minus4;
-    int32_t                                 offset_for_non_ref_pic;
-    int32_t                                 offset_for_top_to_bottom_field;
-    uint8_t                                 num_ref_frames_in_pic_order_cnt_cycle;
-    uint8_t                                 max_num_ref_frames;
-    uint32_t                                pic_width_in_mbs_minus1;
-    uint32_t                                pic_height_in_map_units_minus1;
-    uint32_t                                frame_crop_left_offset;
-    uint32_t                                frame_crop_right_offset;
-    uint32_t                                frame_crop_top_offset;
-    uint32_t                                frame_crop_bottom_offset;
-    StdVideoH264SpsFlags                    flags;
-    int32_t*                                pOffsetForRefFrame;
-    StdVideoH264ScalingLists*               pScalingLists;
-    StdVideoH264SequenceParameterSetVui*    pSequenceParameterSetVui;
+    StdVideoH264SpsFlags                          flags;
+    StdVideoH264ProfileIdc                        profile_idc;
+    StdVideoH264LevelIdc                          level_idc;
+    StdVideoH264ChromaFormatIdc                   chroma_format_idc;
+    uint8_t                                       seq_parameter_set_id;
+    uint8_t                                       bit_depth_luma_minus8;
+    uint8_t                                       bit_depth_chroma_minus8;
+    uint8_t                                       log2_max_frame_num_minus4;
+    StdVideoH264PocType                           pic_order_cnt_type;
+    int32_t                                       offset_for_non_ref_pic;
+    int32_t                                       offset_for_top_to_bottom_field;
+    uint8_t                                       log2_max_pic_order_cnt_lsb_minus4;
+    uint8_t                                       num_ref_frames_in_pic_order_cnt_cycle;
+    uint8_t                                       max_num_ref_frames;
+    uint8_t                                       reserved1;
+    uint32_t                                      pic_width_in_mbs_minus1;
+    uint32_t                                      pic_height_in_map_units_minus1;
+    uint32_t                                      frame_crop_left_offset;
+    uint32_t                                      frame_crop_right_offset;
+    uint32_t                                      frame_crop_top_offset;
+    uint32_t                                      frame_crop_bottom_offset;
+    uint32_t                                      reserved2;
+    const int32_t*                                pOffsetForRefFrame;
+    const StdVideoH264ScalingLists*               pScalingLists;
+    const StdVideoH264SequenceParameterSetVui*    pSequenceParameterSetVui;
 } StdVideoH264SequenceParameterSet;
 
 typedef struct StdVideoH264PpsFlags {
@@ -279,25 +282,24 @@ typedef struct StdVideoH264PpsFlags {
     uint32_t    redundant_pic_cnt_present_flag : 1;
     uint32_t    constrained_intra_pred_flag : 1;
     uint32_t    deblocking_filter_control_present_flag : 1;
-    uint32_t    weighted_bipred_idc_flag : 1;
     uint32_t    weighted_pred_flag : 1;
-    uint32_t    pic_order_present_flag : 1;
+    uint32_t    bottom_field_pic_order_in_frame_present_flag : 1;
     uint32_t    entropy_coding_mode_flag : 1;
     uint32_t    pic_scaling_matrix_present_flag : 1;
 } StdVideoH264PpsFlags;
 
 typedef struct StdVideoH264PictureParameterSet {
-    uint8_t                          seq_parameter_set_id;
-    uint8_t                          pic_parameter_set_id;
-    uint8_t                          num_ref_idx_l0_default_active_minus1;
-    uint8_t                          num_ref_idx_l1_default_active_minus1;
-    StdVideoH264WeightedBipredIdc    weighted_bipred_idc;
-    int8_t                           pic_init_qp_minus26;
-    int8_t                           pic_init_qs_minus26;
-    int8_t                           chroma_qp_index_offset;
-    int8_t                           second_chroma_qp_index_offset;
-    StdVideoH264PpsFlags             flags;
-    StdVideoH264ScalingLists*        pScalingLists;
+    StdVideoH264PpsFlags               flags;
+    uint8_t                            seq_parameter_set_id;
+    uint8_t                            pic_parameter_set_id;
+    uint8_t                            num_ref_idx_l0_default_active_minus1;
+    uint8_t                            num_ref_idx_l1_default_active_minus1;
+    StdVideoH264WeightedBipredIdc      weighted_bipred_idc;
+    int8_t                             pic_init_qp_minus26;
+    int8_t                             pic_init_qs_minus26;
+    int8_t                             chroma_qp_index_offset;
+    int8_t                             second_chroma_qp_index_offset;
+    const StdVideoH264ScalingLists*    pScalingLists;
 } StdVideoH264PictureParameterSet;
 
 

libs/bluevk/include/vk_video/vulkan_video_codec_h264std_decode.h

@@ -20,8 +20,12 @@ extern "C" {
 
 
 #define vulkan_video_codec_h264std_decode 1
+// Vulkan 0.9 provisional Vulkan video H.264 decode std specification version number
+#define VK_STD_VULKAN_VIDEO_CODEC_H264_DECODE_API_VERSION_0_9_8 VK_MAKE_VIDEO_STD_VERSION(0, 9, 8) // Patch version should always be set to 0
+
 #define STD_VIDEO_DECODE_H264_FIELD_ORDER_COUNT_LIST_SIZE 2
-#define STD_VIDEO_DECODE_H264_MVC_REF_LIST_SIZE 15
+#define VK_STD_VULKAN_VIDEO_CODEC_H264_DECODE_SPEC_VERSION VK_STD_VULKAN_VIDEO_CODEC_H264_DECODE_API_VERSION_0_9_8
+#define VK_STD_VULKAN_VIDEO_CODEC_H264_DECODE_EXTENSION_NAME "VK_STD_vulkan_video_codec_h264_decode"
 
 typedef enum StdVideoDecodeH264FieldOrderCount {
     STD_VIDEO_DECODE_H264_FIELD_ORDER_COUNT_TOP = 0,
@@ -39,57 +43,30 @@ typedef struct StdVideoDecodeH264PictureInfoFlags {
 } StdVideoDecodeH264PictureInfoFlags;
 
 typedef struct StdVideoDecodeH264PictureInfo {
+    StdVideoDecodeH264PictureInfoFlags    flags;
     uint8_t                               seq_parameter_set_id;
     uint8_t                               pic_parameter_set_id;
-    uint16_t                              reserved;
+    uint8_t                               reserved1;
+    uint8_t                               reserved2;
     uint16_t                              frame_num;
     uint16_t                              idr_pic_id;
     int32_t                               PicOrderCnt[STD_VIDEO_DECODE_H264_FIELD_ORDER_COUNT_LIST_SIZE];
-    StdVideoDecodeH264PictureInfoFlags    flags;
 } StdVideoDecodeH264PictureInfo;
 
 typedef struct StdVideoDecodeH264ReferenceInfoFlags {
     uint32_t    top_field_flag : 1;
     uint32_t    bottom_field_flag : 1;
-    uint32_t    is_long_term : 1;
+    uint32_t    used_for_long_term_reference : 1;
     uint32_t    is_non_existing : 1;
 } StdVideoDecodeH264ReferenceInfoFlags;
 
 typedef struct StdVideoDecodeH264ReferenceInfo {
+    StdVideoDecodeH264ReferenceInfoFlags    flags;
     uint16_t                                FrameNum;
     uint16_t                                reserved;
-    int32_t                                 PicOrderCnt[2];
-    StdVideoDecodeH264ReferenceInfoFlags    flags;
+    int32_t                                 PicOrderCnt[STD_VIDEO_DECODE_H264_FIELD_ORDER_COUNT_LIST_SIZE];
 } StdVideoDecodeH264ReferenceInfo;
 
-typedef struct StdVideoDecodeH264MvcElementFlags {
-    uint32_t    non_idr : 1;
-    uint32_t    anchor_pic : 1;
-    uint32_t    inter_view : 1;
-} StdVideoDecodeH264MvcElementFlags;
-
-typedef struct StdVideoDecodeH264MvcElement {
-    StdVideoDecodeH264MvcElementFlags    flags;
-    uint16_t                             viewOrderIndex;
-    uint16_t                             viewId;
-    uint16_t                             temporalId;
-    uint16_t                             priorityId;
-    uint16_t                             numOfAnchorRefsInL0;
-    uint16_t                             viewIdOfAnchorRefsInL0[STD_VIDEO_DECODE_H264_MVC_REF_LIST_SIZE];
-    uint16_t                             numOfAnchorRefsInL1;
-    uint16_t                             viewIdOfAnchorRefsInL1[STD_VIDEO_DECODE_H264_MVC_REF_LIST_SIZE];
-    uint16_t                             numOfNonAnchorRefsInL0;
-    uint16_t                             viewIdOfNonAnchorRefsInL0[STD_VIDEO_DECODE_H264_MVC_REF_LIST_SIZE];
-    uint16_t                             numOfNonAnchorRefsInL1;
-    uint16_t                             viewIdOfNonAnchorRefsInL1[STD_VIDEO_DECODE_H264_MVC_REF_LIST_SIZE];
-} StdVideoDecodeH264MvcElement;
-
-typedef struct StdVideoDecodeH264Mvc {
-    uint32_t                         viewId0;
-    uint32_t                         mvcElementCount;
-    StdVideoDecodeH264MvcElement*    pMvcElements;
-} StdVideoDecodeH264Mvc;
-
 
 #ifdef __cplusplus
 }

libs/bluevk/include/vk_video/vulkan_video_codec_h264std_encode.h

@@ -20,6 +20,32 @@ extern "C" {
 
 
 #define vulkan_video_codec_h264std_encode 1
+// Vulkan 0.9 provisional Vulkan video H.264 encode std specification version number
+#define VK_STD_VULKAN_VIDEO_CODEC_H264_ENCODE_API_VERSION_0_9_8 VK_MAKE_VIDEO_STD_VERSION(0, 9, 8) // Patch version should always be set to 0
+
+#define VK_STD_VULKAN_VIDEO_CODEC_H264_ENCODE_SPEC_VERSION VK_STD_VULKAN_VIDEO_CODEC_H264_ENCODE_API_VERSION_0_9_8
+#define VK_STD_VULKAN_VIDEO_CODEC_H264_ENCODE_EXTENSION_NAME "VK_STD_vulkan_video_codec_h264_encode"
+typedef struct StdVideoEncodeH264WeightTableFlags {
+    uint32_t    luma_weight_l0_flag;
+    uint32_t    chroma_weight_l0_flag;
+    uint32_t    luma_weight_l1_flag;
+    uint32_t    chroma_weight_l1_flag;
+} StdVideoEncodeH264WeightTableFlags;
+
+typedef struct StdVideoEncodeH264WeightTable {
+    StdVideoEncodeH264WeightTableFlags    flags;
+    uint8_t                               luma_log2_weight_denom;
+    uint8_t                               chroma_log2_weight_denom;
+    int8_t                                luma_weight_l0[STD_VIDEO_H264_MAX_NUM_LIST_REF];
+    int8_t                                luma_offset_l0[STD_VIDEO_H264_MAX_NUM_LIST_REF];
+    int8_t                                chroma_weight_l0[STD_VIDEO_H264_MAX_NUM_LIST_REF][STD_VIDEO_H264_MAX_CHROMA_PLANES];
+    int8_t                                chroma_offset_l0[STD_VIDEO_H264_MAX_NUM_LIST_REF][STD_VIDEO_H264_MAX_CHROMA_PLANES];
+    int8_t                                luma_weight_l1[STD_VIDEO_H264_MAX_NUM_LIST_REF];
+    int8_t                                luma_offset_l1[STD_VIDEO_H264_MAX_NUM_LIST_REF];
+    int8_t                                chroma_weight_l1[STD_VIDEO_H264_MAX_NUM_LIST_REF][STD_VIDEO_H264_MAX_CHROMA_PLANES];
+    int8_t                                chroma_offset_l1[STD_VIDEO_H264_MAX_NUM_LIST_REF][STD_VIDEO_H264_MAX_CHROMA_PLANES];
+} StdVideoEncodeH264WeightTable;
+
 typedef struct StdVideoEncodeH264SliceHeaderFlags {
     uint32_t    direct_spatial_mv_pred_flag : 1;
     uint32_t    num_ref_idx_active_override_flag : 1;
@@ -31,11 +57,11 @@ typedef struct StdVideoEncodeH264SliceHeaderFlags {
 typedef struct StdVideoEncodeH264PictureInfoFlags {
     uint32_t    idr_flag : 1;
     uint32_t    is_reference_flag : 1;
-    uint32_t    long_term_reference_flag : 1;
+    uint32_t    used_for_long_term_reference : 1;
 } StdVideoEncodeH264PictureInfoFlags;
 
 typedef struct StdVideoEncodeH264ReferenceInfoFlags {
-    uint32_t    is_long_term : 1;
+    uint32_t    used_for_long_term_reference : 1;
 } StdVideoEncodeH264ReferenceInfoFlags;
 
 typedef struct StdVideoEncodeH264RefMgmtFlags {
@@ -58,17 +84,19 @@ typedef struct StdVideoEncodeH264RefPicMarkingEntry {
 } StdVideoEncodeH264RefPicMarkingEntry;
 
 typedef struct StdVideoEncodeH264RefMemMgmtCtrlOperations {
-    StdVideoEncodeH264RefMgmtFlags           flags;
-    uint8_t                                  refList0ModOpCount;
-    StdVideoEncodeH264RefListModEntry*       pRefList0ModOperations;
-    uint8_t                                  refList1ModOpCount;
-    StdVideoEncodeH264RefListModEntry*       pRefList1ModOperations;
-    uint8_t                                  refPicMarkingOpCount;
-    StdVideoEncodeH264RefPicMarkingEntry*    pRefPicMarkingOperations;
+    StdVideoEncodeH264RefMgmtFlags                 flags;
+    uint8_t                                        refList0ModOpCount;
+    const StdVideoEncodeH264RefListModEntry*       pRefList0ModOperations;
+    uint8_t                                        refList1ModOpCount;
+    const StdVideoEncodeH264RefListModEntry*       pRefList1ModOperations;
+    uint8_t                                        refPicMarkingOpCount;
+    const StdVideoEncodeH264RefPicMarkingEntry*    pRefPicMarkingOperations;
 } StdVideoEncodeH264RefMemMgmtCtrlOperations;
 
 typedef struct StdVideoEncodeH264PictureInfo {
     StdVideoEncodeH264PictureInfoFlags    flags;
+    uint8_t                               seq_parameter_set_id;
+    uint8_t                               pic_parameter_set_id;
     StdVideoH264PictureType               pictureType;
     uint32_t                              frame_num;
     int32_t                               PicOrderCnt;
@@ -86,8 +114,6 @@ typedef struct StdVideoEncodeH264SliceHeader {
     StdVideoEncodeH264SliceHeaderFlags        flags;
     uint32_t                                  first_mb_in_slice;
     StdVideoH264SliceType                     slice_type;
-    uint8_t                                   seq_parameter_set_id;
-    uint8_t                                   pic_parameter_set_id;
     uint16_t                                  idr_pic_id;
     uint8_t                                   num_ref_idx_l0_active_minus1;
     uint8_t                                   num_ref_idx_l1_active_minus1;
@@ -95,6 +121,7 @@ typedef struct StdVideoEncodeH264SliceHeader {
     StdVideoH264DisableDeblockingFilterIdc    disable_deblocking_filter_idc;
     int8_t                                    slice_alpha_c0_offset_div2;
     int8_t                                    slice_beta_offset_div2;
+    const StdVideoEncodeH264WeightTable*      pWeightTable;
 } StdVideoEncodeH264SliceHeader;
 
 

libs/bluevk/include/vk_video/vulkan_video_codec_h265std.h

@@ -20,10 +20,7 @@ extern "C" {
 
 
 #define vulkan_video_codec_h265std 1
-// Vulkan 0.5 version number WIP
-#define VK_STD_VULKAN_VIDEO_CODEC_H265_API_VERSION_0_9_5 VK_MAKE_VIDEO_STD_VERSION(0, 9, 5) // Patch version should always be set to 0
-
-#define STD_VIDEO_H265_SUBLAYERS_MINUS1_LIST_SIZE 7
+#define STD_VIDEO_H265_SUBLAYERS_LIST_SIZE 7
 #define STD_VIDEO_H265_CPB_CNT_LIST_SIZE  32
 #define STD_VIDEO_H265_SCALING_LIST_4X4_NUM_LISTS 6
 #define STD_VIDEO_H265_SCALING_LIST_4X4_NUM_ELEMENTS 16
@@ -35,11 +32,16 @@ extern "C" {
 #define STD_VIDEO_H265_SCALING_LIST_32X32_NUM_ELEMENTS 64
 #define STD_VIDEO_H265_PREDICTOR_PALETTE_COMPONENTS_LIST_SIZE 3
 #define STD_VIDEO_H265_PREDICTOR_PALETTE_COMP_ENTRIES_LIST_SIZE 128
+#define STD_VIDEO_H265_MAX_DPB_SIZE       16
+#define STD_VIDEO_H265_MAX_LONG_TERM_REF_PICS_SPS 32
+#define STD_VIDEO_H265_CHROMA_QP_OFFSET_LIST_SIZE 6
 #define STD_VIDEO_H265_CHROMA_QP_OFFSET_TILE_COLS_LIST_SIZE 19
 #define STD_VIDEO_H265_CHROMA_QP_OFFSET_TILE_ROWS_LIST_SIZE 21
-#define STD_VIDEO_H265_CHROMA_QP_OFFSET_LIST_SIZE 6
-#define VK_STD_VULKAN_VIDEO_CODEC_H265_SPEC_VERSION VK_STD_VULKAN_VIDEO_CODEC_H265_API_VERSION_0_9_5
-#define VK_STD_VULKAN_VIDEO_CODEC_H265_EXTENSION_NAME "VK_STD_vulkan_video_codec_h265"
+#define STD_VIDEO_H265_MAX_NUM_LIST_REF   15
+#define STD_VIDEO_H265_MAX_CHROMA_PLANES  2
+#define STD_VIDEO_H265_MAX_SHORT_TERM_REF_PIC_SETS 64
+#define STD_VIDEO_H265_MAX_LONG_TERM_PICS 16
+#define STD_VIDEO_H265_MAX_DELTA_POC      48
 
 typedef enum StdVideoH265ChromaFormatIdc {
     STD_VIDEO_H265_CHROMA_FORMAT_IDC_MONOCHROME = 0,
@@ -60,23 +62,23 @@ typedef enum StdVideoH265ProfileIdc {
     STD_VIDEO_H265_PROFILE_IDC_MAX_ENUM = 0x7FFFFFFF
 } StdVideoH265ProfileIdc;
 
-typedef enum StdVideoH265Level {
-    STD_VIDEO_H265_LEVEL_1_0 = 0,
-    STD_VIDEO_H265_LEVEL_2_0 = 1,
-    STD_VIDEO_H265_LEVEL_2_1 = 2,
-    STD_VIDEO_H265_LEVEL_3_0 = 3,
-    STD_VIDEO_H265_LEVEL_3_1 = 4,
-    STD_VIDEO_H265_LEVEL_4_0 = 5,
-    STD_VIDEO_H265_LEVEL_4_1 = 6,
-    STD_VIDEO_H265_LEVEL_5_0 = 7,
-    STD_VIDEO_H265_LEVEL_5_1 = 8,
-    STD_VIDEO_H265_LEVEL_5_2 = 9,
-    STD_VIDEO_H265_LEVEL_6_0 = 10,
-    STD_VIDEO_H265_LEVEL_6_1 = 11,
-    STD_VIDEO_H265_LEVEL_6_2 = 12,
-    STD_VIDEO_H265_LEVEL_INVALID = 0x7FFFFFFF,
-    STD_VIDEO_H265_LEVEL_MAX_ENUM = 0x7FFFFFFF
-} StdVideoH265Level;
+typedef enum StdVideoH265LevelIdc {
+    STD_VIDEO_H265_LEVEL_IDC_1_0 = 0,
+    STD_VIDEO_H265_LEVEL_IDC_2_0 = 1,
+    STD_VIDEO_H265_LEVEL_IDC_2_1 = 2,
+    STD_VIDEO_H265_LEVEL_IDC_3_0 = 3,
+    STD_VIDEO_H265_LEVEL_IDC_3_1 = 4,
+    STD_VIDEO_H265_LEVEL_IDC_4_0 = 5,
+    STD_VIDEO_H265_LEVEL_IDC_4_1 = 6,
+    STD_VIDEO_H265_LEVEL_IDC_5_0 = 7,
+    STD_VIDEO_H265_LEVEL_IDC_5_1 = 8,
+    STD_VIDEO_H265_LEVEL_IDC_5_2 = 9,
+    STD_VIDEO_H265_LEVEL_IDC_6_0 = 10,
+    STD_VIDEO_H265_LEVEL_IDC_6_1 = 11,
+    STD_VIDEO_H265_LEVEL_IDC_6_2 = 12,
+    STD_VIDEO_H265_LEVEL_IDC_INVALID = 0x7FFFFFFF,
+    STD_VIDEO_H265_LEVEL_IDC_MAX_ENUM = 0x7FFFFFFF
+} StdVideoH265LevelIdc;
 
 typedef enum StdVideoH265SliceType {
     STD_VIDEO_H265_SLICE_TYPE_B = 0,
@@ -94,10 +96,33 @@ typedef enum StdVideoH265PictureType {
     STD_VIDEO_H265_PICTURE_TYPE_INVALID = 0x7FFFFFFF,
     STD_VIDEO_H265_PICTURE_TYPE_MAX_ENUM = 0x7FFFFFFF
 } StdVideoH265PictureType;
+
+typedef enum StdVideoH265AspectRatioIdc {
+    STD_VIDEO_H265_ASPECT_RATIO_IDC_UNSPECIFIED = 0,
+    STD_VIDEO_H265_ASPECT_RATIO_IDC_SQUARE = 1,
+    STD_VIDEO_H265_ASPECT_RATIO_IDC_12_11 = 2,
+    STD_VIDEO_H265_ASPECT_RATIO_IDC_10_11 = 3,
+    STD_VIDEO_H265_ASPECT_RATIO_IDC_16_11 = 4,
+    STD_VIDEO_H265_ASPECT_RATIO_IDC_40_33 = 5,
+    STD_VIDEO_H265_ASPECT_RATIO_IDC_24_11 = 6,
+    STD_VIDEO_H265_ASPECT_RATIO_IDC_20_11 = 7,
+    STD_VIDEO_H265_ASPECT_RATIO_IDC_32_11 = 8,
+    STD_VIDEO_H265_ASPECT_RATIO_IDC_80_33 = 9,
+    STD_VIDEO_H265_ASPECT_RATIO_IDC_18_11 = 10,
+    STD_VIDEO_H265_ASPECT_RATIO_IDC_15_11 = 11,
+    STD_VIDEO_H265_ASPECT_RATIO_IDC_64_33 = 12,
+    STD_VIDEO_H265_ASPECT_RATIO_IDC_160_99 = 13,
+    STD_VIDEO_H265_ASPECT_RATIO_IDC_4_3 = 14,
+    STD_VIDEO_H265_ASPECT_RATIO_IDC_3_2 = 15,
+    STD_VIDEO_H265_ASPECT_RATIO_IDC_2_1 = 16,
+    STD_VIDEO_H265_ASPECT_RATIO_IDC_EXTENDED_SAR = 255,
+    STD_VIDEO_H265_ASPECT_RATIO_IDC_INVALID = 0x7FFFFFFF,
+    STD_VIDEO_H265_ASPECT_RATIO_IDC_MAX_ENUM = 0x7FFFFFFF
+} StdVideoH265AspectRatioIdc;
 typedef struct StdVideoH265DecPicBufMgr {
-    uint32_t    max_latency_increase_plus1[STD_VIDEO_H265_SUBLAYERS_MINUS1_LIST_SIZE];
-    uint8_t     max_dec_pic_buffering_minus1[STD_VIDEO_H265_SUBLAYERS_MINUS1_LIST_SIZE];
-    uint8_t     max_num_reorder_pics[STD_VIDEO_H265_SUBLAYERS_MINUS1_LIST_SIZE];
+    uint32_t    max_latency_increase_plus1[STD_VIDEO_H265_SUBLAYERS_LIST_SIZE];
+    uint8_t     max_dec_pic_buffering_minus1[STD_VIDEO_H265_SUBLAYERS_LIST_SIZE];
+    uint8_t     max_num_reorder_pics[STD_VIDEO_H265_SUBLAYERS_LIST_SIZE];
 } StdVideoH265DecPicBufMgr;
 
 typedef struct StdVideoH265SubLayerHrdParameters {
@@ -119,20 +144,21 @@ typedef struct StdVideoH265HrdFlags {
 } StdVideoH265HrdFlags;
 
 typedef struct StdVideoH265HrdParameters {
-    uint8_t                               tick_divisor_minus2;
-    uint8_t                               du_cpb_removal_delay_increment_length_minus1;
-    uint8_t                               dpb_output_delay_du_length_minus1;
-    uint8_t                               bit_rate_scale;
-    uint8_t                               cpb_size_scale;
-    uint8_t                               cpb_size_du_scale;
-    uint8_t                               initial_cpb_removal_delay_length_minus1;
-    uint8_t                               au_cpb_removal_delay_length_minus1;
-    uint8_t                               dpb_output_delay_length_minus1;
-    uint8_t                               cpb_cnt_minus1[STD_VIDEO_H265_SUBLAYERS_MINUS1_LIST_SIZE];
-    uint16_t                              elemental_duration_in_tc_minus1[STD_VIDEO_H265_SUBLAYERS_MINUS1_LIST_SIZE];
-    StdVideoH265SubLayerHrdParameters*    pSubLayerHrdParametersNal[STD_VIDEO_H265_SUBLAYERS_MINUS1_LIST_SIZE];
-    StdVideoH265SubLayerHrdParameters*    pSubLayerHrdParametersVcl[STD_VIDEO_H265_SUBLAYERS_MINUS1_LIST_SIZE];
-    StdVideoH265HrdFlags                  flags;
+    StdVideoH265HrdFlags                        flags;
+    uint8_t                                     tick_divisor_minus2;
+    uint8_t                                     du_cpb_removal_delay_increment_length_minus1;
+    uint8_t                                     dpb_output_delay_du_length_minus1;
+    uint8_t                                     bit_rate_scale;
+    uint8_t                                     cpb_size_scale;
+    uint8_t                                     cpb_size_du_scale;
+    uint8_t                                     initial_cpb_removal_delay_length_minus1;
+    uint8_t                                     au_cpb_removal_delay_length_minus1;
+    uint8_t                                     dpb_output_delay_length_minus1;
+    uint8_t                                     cpb_cnt_minus1[STD_VIDEO_H265_SUBLAYERS_LIST_SIZE];
+    uint16_t                                    elemental_duration_in_tc_minus1[STD_VIDEO_H265_SUBLAYERS_LIST_SIZE];
+    uint16_t                                    reserved[3];
+    const StdVideoH265SubLayerHrdParameters*    pSubLayerHrdParametersNal;
+    const StdVideoH265SubLayerHrdParameters*    pSubLayerHrdParametersVcl;
 } StdVideoH265HrdParameters;
 
 typedef struct StdVideoH265VpsFlags {
@@ -142,15 +168,33 @@ typedef struct StdVideoH265VpsFlags {
     uint32_t    vps_poc_proportional_to_timing_flag : 1;
 } StdVideoH265VpsFlags;
 
+typedef struct StdVideoH265ProfileTierLevelFlags {
+    uint32_t    general_tier_flag : 1;
+    uint32_t    general_progressive_source_flag : 1;
+    uint32_t    general_interlaced_source_flag : 1;
+    uint32_t    general_non_packed_constraint_flag : 1;
+    uint32_t    general_frame_only_constraint_flag : 1;
+} StdVideoH265ProfileTierLevelFlags;
+
+typedef struct StdVideoH265ProfileTierLevel {
+    StdVideoH265ProfileTierLevelFlags    flags;
+    StdVideoH265ProfileIdc               general_profile_idc;
+    StdVideoH265LevelIdc                 general_level_idc;
+} StdVideoH265ProfileTierLevel;
+
 typedef struct StdVideoH265VideoParameterSet {
-    uint8_t                       vps_video_parameter_set_id;
-    uint8_t                       vps_max_sub_layers_minus1;
-    uint32_t                      vps_num_units_in_tick;
-    uint32_t                      vps_time_scale;
-    uint32_t                      vps_num_ticks_poc_diff_one_minus1;
-    StdVideoH265DecPicBufMgr*     pDecPicBufMgr;
-    StdVideoH265HrdParameters*    pHrdParameters;
-    StdVideoH265VpsFlags          flags;
+    StdVideoH265VpsFlags                   flags;
+    uint8_t                                vps_video_parameter_set_id;
+    uint8_t                                vps_max_sub_layers_minus1;
+    uint8_t                                reserved1;
+    uint8_t                                reserved2;
+    uint32_t                               vps_num_units_in_tick;
+    uint32_t                               vps_time_scale;
+    uint32_t                               vps_num_ticks_poc_diff_one_minus1;
+    uint32_t                               reserved3;
+    const StdVideoH265DecPicBufMgr*        pDecPicBufMgr;
+    const StdVideoH265HrdParameters*       pHrdParameters;
+    const StdVideoH265ProfileTierLevel*    pProfileTierLevel;
 } StdVideoH265VideoParameterSet;
 
 typedef struct StdVideoH265ScalingLists {
@@ -184,29 +228,32 @@ typedef struct StdVideoH265SpsVuiFlags {
 } StdVideoH265SpsVuiFlags;
 
 typedef struct StdVideoH265SequenceParameterSetVui {
-    uint8_t                       aspect_ratio_idc;
-    uint16_t                      sar_width;
-    uint16_t                      sar_height;
-    uint8_t                       video_format;
-    uint8_t                       colour_primaries;
-    uint8_t                       transfer_characteristics;
-    uint8_t                       matrix_coeffs;
-    uint8_t                       chroma_sample_loc_type_top_field;
-    uint8_t                       chroma_sample_loc_type_bottom_field;
-    uint16_t                      def_disp_win_left_offset;
-    uint16_t                      def_disp_win_right_offset;
-    uint16_t                      def_disp_win_top_offset;
-    uint16_t                      def_disp_win_bottom_offset;
-    uint32_t                      vui_num_units_in_tick;
-    uint32_t                      vui_time_scale;
-    uint32_t                      vui_num_ticks_poc_diff_one_minus1;
-    StdVideoH265HrdParameters*    pHrdParameters;
-    uint16_t                      min_spatial_segmentation_idc;
-    uint8_t                       max_bytes_per_pic_denom;
-    uint8_t                       max_bits_per_min_cu_denom;
-    uint8_t                       log2_max_mv_length_horizontal;
-    uint8_t                       log2_max_mv_length_vertical;
-    StdVideoH265SpsVuiFlags       flags;
+    StdVideoH265SpsVuiFlags             flags;
+    StdVideoH265AspectRatioIdc          aspect_ratio_idc;
+    uint16_t                            sar_width;
+    uint16_t                            sar_height;
+    uint8_t                             video_format;
+    uint8_t                             colour_primaries;
+    uint8_t                             transfer_characteristics;
+    uint8_t                             matrix_coeffs;
+    uint8_t                             chroma_sample_loc_type_top_field;
+    uint8_t                             chroma_sample_loc_type_bottom_field;
+    uint8_t                             reserved1;
+    uint8_t                             reserved2;
+    uint16_t                            def_disp_win_left_offset;
+    uint16_t                            def_disp_win_right_offset;
+    uint16_t                            def_disp_win_top_offset;
+    uint16_t                            def_disp_win_bottom_offset;
+    uint32_t                            vui_num_units_in_tick;
+    uint32_t                            vui_time_scale;
+    uint32_t                            vui_num_ticks_poc_diff_one_minus1;
+    uint16_t                            min_spatial_segmentation_idc;
+    uint16_t                            reserved3;
+    uint8_t                             max_bytes_per_pic_denom;
+    uint8_t                             max_bits_per_min_cu_denom;
+    uint8_t                             log2_max_mv_length_horizontal;
+    uint8_t                             log2_max_mv_length_vertical;
+    const StdVideoH265HrdParameters*    pHrdParameters;
 } StdVideoH265SequenceParameterSetVui;
 
 typedef struct StdVideoH265PredictorPaletteEntries {
@@ -216,6 +263,8 @@ typedef struct StdVideoH265PredictorPaletteEntries {
 typedef struct StdVideoH265SpsFlags {
     uint32_t    sps_temporal_id_nesting_flag : 1;
     uint32_t    separate_colour_plane_flag : 1;
+    uint32_t    conformance_window_flag : 1;
+    uint32_t    sps_sub_layer_ordering_info_present_flag : 1;
     uint32_t    scaling_list_enabled_flag : 1;
     uint32_t    sps_scaling_list_data_present_flag : 1;
     uint32_t    amp_enabled_flag : 1;
@@ -237,50 +286,80 @@ typedef struct StdVideoH265SpsFlags {
     uint32_t    high_precision_offsets_enabled_flag : 1;
     uint32_t    persistent_rice_adaptation_enabled_flag : 1;
     uint32_t    cabac_bypass_alignment_enabled_flag : 1;
+    uint32_t    sps_scc_extension_flag : 1;
     uint32_t    sps_curr_pic_ref_enabled_flag : 1;
     uint32_t    palette_mode_enabled_flag : 1;
-    uint32_t    sps_palette_predictor_initializer_present_flag : 1;
+    uint32_t    sps_palette_predictor_initializers_present_flag : 1;
     uint32_t    intra_boundary_filtering_disabled_flag : 1;
 } StdVideoH265SpsFlags;
 
+typedef struct StdVideoH265ShortTermRefPicSetFlags {
+    uint32_t    inter_ref_pic_set_prediction_flag : 1;
+    uint32_t    delta_rps_sign : 1;
+} StdVideoH265ShortTermRefPicSetFlags;
+
+typedef struct StdVideoH265ShortTermRefPicSet {
+    StdVideoH265ShortTermRefPicSetFlags    flags;
+    uint32_t                               delta_idx_minus1;
+    uint16_t                               use_delta_flag;
+    uint16_t                               abs_delta_rps_minus1;
+    uint16_t                               used_by_curr_pic_flag;
+    uint16_t                               used_by_curr_pic_s0_flag;
+    uint16_t                               used_by_curr_pic_s1_flag;
+    uint16_t                               reserved1;
+    uint8_t                                reserved2;
+    uint8_t                                reserved3;
+    uint8_t                                num_negative_pics;
+    uint8_t                                num_positive_pics;
+    uint16_t                               delta_poc_s0_minus1[STD_VIDEO_H265_MAX_DPB_SIZE];
+    uint16_t                               delta_poc_s1_minus1[STD_VIDEO_H265_MAX_DPB_SIZE];
+} StdVideoH265ShortTermRefPicSet;
+
+typedef struct StdVideoH265LongTermRefPicsSps {
+    uint32_t    used_by_curr_pic_lt_sps_flag;
+    uint32_t    lt_ref_pic_poc_lsb_sps[STD_VIDEO_H265_MAX_LONG_TERM_REF_PICS_SPS];
+} StdVideoH265LongTermRefPicsSps;
+
 typedef struct StdVideoH265SequenceParameterSet {
-    StdVideoH265ProfileIdc                  profile_idc;
-    StdVideoH265Level                       level_idc;
-    uint32_t                                pic_width_in_luma_samples;
-    uint32_t                                pic_height_in_luma_samples;
-    uint8_t                                 sps_video_parameter_set_id;
-    uint8_t                                 sps_max_sub_layers_minus1;
-    uint8_t                                 sps_seq_parameter_set_id;
-    uint8_t                                 chroma_format_idc;
-    uint8_t                                 bit_depth_luma_minus8;
-    uint8_t                                 bit_depth_chroma_minus8;
-    uint8_t                                 log2_max_pic_order_cnt_lsb_minus4;
-    uint8_t                                 sps_max_dec_pic_buffering_minus1;
-    uint8_t                                 log2_min_luma_coding_block_size_minus3;
-    uint8_t                                 log2_diff_max_min_luma_coding_block_size;
-    uint8_t                                 log2_min_luma_transform_block_size_minus2;
-    uint8_t                                 log2_diff_max_min_luma_transform_block_size;
-    uint8_t                                 max_transform_hierarchy_depth_inter;
-    uint8_t                                 max_transform_hierarchy_depth_intra;
-    uint8_t                                 num_short_term_ref_pic_sets;
-    uint8_t                                 num_long_term_ref_pics_sps;
-    uint8_t                                 pcm_sample_bit_depth_luma_minus1;
-    uint8_t                                 pcm_sample_bit_depth_chroma_minus1;
-    uint8_t                                 log2_min_pcm_luma_coding_block_size_minus3;
-    uint8_t                                 log2_diff_max_min_pcm_luma_coding_block_size;
-    uint32_t                                conf_win_left_offset;
-    uint32_t                                conf_win_right_offset;
-    uint32_t                                conf_win_top_offset;
-    uint32_t                                conf_win_bottom_offset;
-    StdVideoH265DecPicBufMgr*               pDecPicBufMgr;
-    StdVideoH265SpsFlags                    flags;
-    StdVideoH265ScalingLists*               pScalingLists;
-    StdVideoH265SequenceParameterSetVui*    pSequenceParameterSetVui;
-    uint8_t                                 palette_max_size;
-    uint8_t                                 delta_palette_max_predictor_size;
-    uint8_t                                 motion_vector_resolution_control_idc;
-    uint8_t                                 sps_num_palette_predictor_initializer_minus1;
-    StdVideoH265PredictorPaletteEntries*    pPredictorPaletteEntries;
+    StdVideoH265SpsFlags                          flags;
+    StdVideoH265ChromaFormatIdc                   chroma_format_idc;
+    uint32_t                                      pic_width_in_luma_samples;
+    uint32_t                                      pic_height_in_luma_samples;
+    uint8_t                                       sps_video_parameter_set_id;
+    uint8_t                                       sps_max_sub_layers_minus1;
+    uint8_t                                       sps_seq_parameter_set_id;
+    uint8_t                                       bit_depth_luma_minus8;
+    uint8_t                                       bit_depth_chroma_minus8;
+    uint8_t                                       log2_max_pic_order_cnt_lsb_minus4;
+    uint8_t                                       log2_min_luma_coding_block_size_minus3;
+    uint8_t                                       log2_diff_max_min_luma_coding_block_size;
+    uint8_t                                       log2_min_luma_transform_block_size_minus2;
+    uint8_t                                       log2_diff_max_min_luma_transform_block_size;
+    uint8_t                                       max_transform_hierarchy_depth_inter;
+    uint8_t                                       max_transform_hierarchy_depth_intra;
+    uint8_t                                       num_short_term_ref_pic_sets;
+    uint8_t                                       num_long_term_ref_pics_sps;
+    uint8_t                                       pcm_sample_bit_depth_luma_minus1;
+    uint8_t                                       pcm_sample_bit_depth_chroma_minus1;
+    uint8_t                                       log2_min_pcm_luma_coding_block_size_minus3;
+    uint8_t                                       log2_diff_max_min_pcm_luma_coding_block_size;
+    uint8_t                                       reserved1;
+    uint8_t                                       reserved2;
+    uint8_t                                       palette_max_size;
+    uint8_t                                       delta_palette_max_predictor_size;
+    uint8_t                                       motion_vector_resolution_control_idc;
+    uint8_t                                       sps_num_palette_predictor_initializers_minus1;
+    uint32_t                                      conf_win_left_offset;
+    uint32_t                                      conf_win_right_offset;
+    uint32_t                                      conf_win_top_offset;
+    uint32_t                                      conf_win_bottom_offset;
+    const StdVideoH265ProfileTierLevel*           pProfileTierLevel;
+    const StdVideoH265DecPicBufMgr*               pDecPicBufMgr;
+    const StdVideoH265ScalingLists*               pScalingLists;
+    const StdVideoH265ShortTermRefPicSet*         pShortTermRefPicSet;
+    const StdVideoH265LongTermRefPicsSps*         pLongTermRefPicsSps;
+    const StdVideoH265SequenceParameterSetVui*    pSequenceParameterSetVui;
+    const StdVideoH265PredictorPaletteEntries*    pPredictorPaletteEntries;
 } StdVideoH265SequenceParameterSet;
 
 typedef struct StdVideoH265PpsFlags {
@@ -312,44 +391,48 @@ typedef struct StdVideoH265PpsFlags {
     uint32_t    pps_curr_pic_ref_enabled_flag : 1;
     uint32_t    residual_adaptive_colour_transform_enabled_flag : 1;
     uint32_t    pps_slice_act_qp_offsets_present_flag : 1;
-    uint32_t    pps_palette_predictor_initializer_present_flag : 1;
+    uint32_t    pps_palette_predictor_initializers_present_flag : 1;
     uint32_t    monochrome_palette_flag : 1;
     uint32_t    pps_range_extension_flag : 1;
 } StdVideoH265PpsFlags;
 
 typedef struct StdVideoH265PictureParameterSet {
-    uint8_t                                 pps_pic_parameter_set_id;
-    uint8_t                                 pps_seq_parameter_set_id;
-    uint8_t                                 num_extra_slice_header_bits;
-    uint8_t                                 num_ref_idx_l0_default_active_minus1;
-    uint8_t                                 num_ref_idx_l1_default_active_minus1;
-    int8_t                                  init_qp_minus26;
-    uint8_t                                 diff_cu_qp_delta_depth;
-    int8_t                                  pps_cb_qp_offset;
-    int8_t                                  pps_cr_qp_offset;
-    uint8_t                                 num_tile_columns_minus1;
-    uint8_t                                 num_tile_rows_minus1;
-    uint16_t                                column_width_minus1[STD_VIDEO_H265_CHROMA_QP_OFFSET_TILE_COLS_LIST_SIZE];
-    uint16_t                                row_height_minus1[STD_VIDEO_H265_CHROMA_QP_OFFSET_TILE_ROWS_LIST_SIZE];
-    int8_t                                  pps_beta_offset_div2;
-    int8_t                                  pps_tc_offset_div2;
-    uint8_t                                 log2_parallel_merge_level_minus2;
-    StdVideoH265PpsFlags                    flags;
-    StdVideoH265ScalingLists*               pScalingLists;
-    uint8_t                                 log2_max_transform_skip_block_size_minus2;
-    uint8_t                                 diff_cu_chroma_qp_offset_depth;
-    uint8_t                                 chroma_qp_offset_list_len_minus1;
-    int8_t                                  cb_qp_offset_list[STD_VIDEO_H265_CHROMA_QP_OFFSET_LIST_SIZE];
-    int8_t                                  cr_qp_offset_list[STD_VIDEO_H265_CHROMA_QP_OFFSET_LIST_SIZE];
-    uint8_t                                 log2_sao_offset_scale_luma;
-    uint8_t                                 log2_sao_offset_scale_chroma;
-    int8_t                                  pps_act_y_qp_offset_plus5;
-    int8_t                                  pps_act_cb_qp_offset_plus5;
-    int8_t                                  pps_act_cr_qp_offset_plus5;
-    uint8_t                                 pps_num_palette_predictor_initializer;
-    uint8_t                                 luma_bit_depth_entry_minus8;
-    uint8_t                                 chroma_bit_depth_entry_minus8;
-    StdVideoH265PredictorPaletteEntries*    pPredictorPaletteEntries;
+    StdVideoH265PpsFlags                          flags;
+    uint8_t                                       pps_pic_parameter_set_id;
+    uint8_t                                       pps_seq_parameter_set_id;
+    uint8_t                                       sps_video_parameter_set_id;
+    uint8_t                                       num_extra_slice_header_bits;
+    uint8_t                                       num_ref_idx_l0_default_active_minus1;
+    uint8_t                                       num_ref_idx_l1_default_active_minus1;
+    int8_t                                        init_qp_minus26;
+    uint8_t                                       diff_cu_qp_delta_depth;
+    int8_t                                        pps_cb_qp_offset;
+    int8_t                                        pps_cr_qp_offset;
+    int8_t                                        pps_beta_offset_div2;
+    int8_t                                        pps_tc_offset_div2;
+    uint8_t                                       log2_parallel_merge_level_minus2;
+    uint8_t                                       log2_max_transform_skip_block_size_minus2;
+    uint8_t                                       diff_cu_chroma_qp_offset_depth;
+    uint8_t                                       chroma_qp_offset_list_len_minus1;
+    int8_t                                        cb_qp_offset_list[STD_VIDEO_H265_CHROMA_QP_OFFSET_LIST_SIZE];
+    int8_t                                        cr_qp_offset_list[STD_VIDEO_H265_CHROMA_QP_OFFSET_LIST_SIZE];
+    uint8_t                                       log2_sao_offset_scale_luma;
+    uint8_t                                       log2_sao_offset_scale_chroma;
+    int8_t                                        pps_act_y_qp_offset_plus5;
+    int8_t                                        pps_act_cb_qp_offset_plus5;
+    int8_t                                        pps_act_cr_qp_offset_plus3;
+    uint8_t                                       pps_num_palette_predictor_initializers;
+    uint8_t                                       luma_bit_depth_entry_minus8;
+    uint8_t                                       chroma_bit_depth_entry_minus8;
+    uint8_t                                       num_tile_columns_minus1;
+    uint8_t                                       num_tile_rows_minus1;
+    uint8_t                                       reserved1;
+    uint8_t                                       reserved2;
+    uint16_t                                      column_width_minus1[STD_VIDEO_H265_CHROMA_QP_OFFSET_TILE_COLS_LIST_SIZE];
+    uint16_t                                      row_height_minus1[STD_VIDEO_H265_CHROMA_QP_OFFSET_TILE_ROWS_LIST_SIZE];
+    uint32_t                                      reserved3;
+    const StdVideoH265ScalingLists*               pScalingLists;
+    const StdVideoH265PredictorPaletteEntries*    pPredictorPaletteEntries;
 } StdVideoH265PictureParameterSet;
 
 

libs/bluevk/include/vk_video/vulkan_video_codec_h265std_decode.h

@@ -20,7 +20,12 @@ extern "C" {
 
 
 #define vulkan_video_codec_h265std_decode 1
+// Vulkan 0.9 provisional Vulkan video H.265 decode std specification version number
+#define VK_STD_VULKAN_VIDEO_CODEC_H265_DECODE_API_VERSION_0_9_9 VK_MAKE_VIDEO_STD_VERSION(0, 9, 9) // Patch version should always be set to 0
+
 #define STD_VIDEO_DECODE_H265_REF_PIC_SET_LIST_SIZE 8
+#define VK_STD_VULKAN_VIDEO_CODEC_H265_DECODE_SPEC_VERSION VK_STD_VULKAN_VIDEO_CODEC_H265_DECODE_API_VERSION_0_9_9
+#define VK_STD_VULKAN_VIDEO_CODEC_H265_DECODE_EXTENSION_NAME "VK_STD_vulkan_video_codec_h265_decode"
 typedef struct StdVideoDecodeH265PictureInfoFlags {
     uint32_t    IrapPicFlag : 1;
     uint32_t    IdrPicFlag  : 1;
@@ -29,27 +34,27 @@ typedef struct StdVideoDecodeH265PictureInfoFlags {
 } StdVideoDecodeH265PictureInfoFlags;
 
 typedef struct StdVideoDecodeH265PictureInfo {
-    uint8_t                               vps_video_parameter_set_id;
-    uint8_t                               sps_seq_parameter_set_id;
+    StdVideoDecodeH265PictureInfoFlags    flags;
+    uint8_t                               sps_video_parameter_set_id;
+    uint8_t                               pps_seq_parameter_set_id;
     uint8_t                               pps_pic_parameter_set_id;
-    uint8_t                               num_short_term_ref_pic_sets;
+    uint8_t                               NumDeltaPocsOfRefRpsIdx;
     int32_t                               PicOrderCntVal;
     uint16_t                              NumBitsForSTRefPicSetInSlice;
-    uint8_t                               NumDeltaPocsOfRefRpsIdx;
+    uint16_t                              reserved;
     uint8_t                               RefPicSetStCurrBefore[STD_VIDEO_DECODE_H265_REF_PIC_SET_LIST_SIZE];
     uint8_t                               RefPicSetStCurrAfter[STD_VIDEO_DECODE_H265_REF_PIC_SET_LIST_SIZE];
     uint8_t                               RefPicSetLtCurr[STD_VIDEO_DECODE_H265_REF_PIC_SET_LIST_SIZE];
-    StdVideoDecodeH265PictureInfoFlags    flags;
 } StdVideoDecodeH265PictureInfo;
 
 typedef struct StdVideoDecodeH265ReferenceInfoFlags {
-    uint32_t    is_long_term : 1;
-    uint32_t    is_non_existing : 1;
+    uint32_t    used_for_long_term_reference : 1;
+    uint32_t    unused_for_reference : 1;
 } StdVideoDecodeH265ReferenceInfoFlags;
 
 typedef struct StdVideoDecodeH265ReferenceInfo {
-    int32_t                                 PicOrderCntVal;
     StdVideoDecodeH265ReferenceInfoFlags    flags;
+    int32_t                                 PicOrderCntVal;
 } StdVideoDecodeH265ReferenceInfo;
 
 

libs/bluevk/include/vk_video/vulkan_video_codec_h265std_encode.h

@@ -20,13 +20,37 @@ extern "C" {
 
 
 #define vulkan_video_codec_h265std_encode 1
-#define STD_VIDEO_ENCODE_H265_LUMA_LIST_SIZE 15
-#define STD_VIDEO_ENCODE_H265_CHROMA_LIST_SIZE 15
-#define STD_VIDEO_ENCODE_H265_CHROMA_LISTS_NUM 2
+// Vulkan 0.9 provisional Vulkan video H.265 encode std specification version number
+#define VK_STD_VULKAN_VIDEO_CODEC_H265_ENCODE_API_VERSION_0_9_9 VK_MAKE_VIDEO_STD_VERSION(0, 9, 9) // Patch version should always be set to 0
+
+#define VK_STD_VULKAN_VIDEO_CODEC_H265_ENCODE_SPEC_VERSION VK_STD_VULKAN_VIDEO_CODEC_H265_ENCODE_API_VERSION_0_9_9
+#define VK_STD_VULKAN_VIDEO_CODEC_H265_ENCODE_EXTENSION_NAME "VK_STD_vulkan_video_codec_h265_encode"
+typedef struct StdVideoEncodeH265WeightTableFlags {
+    uint16_t    luma_weight_l0_flag;
+    uint16_t    chroma_weight_l0_flag;
+    uint16_t    luma_weight_l1_flag;
+    uint16_t    chroma_weight_l1_flag;
+} StdVideoEncodeH265WeightTableFlags;
+
+typedef struct StdVideoEncodeH265WeightTable {
+    StdVideoEncodeH265WeightTableFlags    flags;
+    uint8_t                               luma_log2_weight_denom;
+    int8_t                                delta_chroma_log2_weight_denom;
+    int8_t                                delta_luma_weight_l0[STD_VIDEO_H265_MAX_NUM_LIST_REF];
+    int8_t                                luma_offset_l0[STD_VIDEO_H265_MAX_NUM_LIST_REF];
+    int8_t                                delta_chroma_weight_l0[STD_VIDEO_H265_MAX_NUM_LIST_REF][STD_VIDEO_H265_MAX_CHROMA_PLANES];
+    int8_t                                delta_chroma_offset_l0[STD_VIDEO_H265_MAX_NUM_LIST_REF][STD_VIDEO_H265_MAX_CHROMA_PLANES];
+    int8_t                                delta_luma_weight_l1[STD_VIDEO_H265_MAX_NUM_LIST_REF];
+    int8_t                                luma_offset_l1[STD_VIDEO_H265_MAX_NUM_LIST_REF];
+    int8_t                                delta_chroma_weight_l1[STD_VIDEO_H265_MAX_NUM_LIST_REF][STD_VIDEO_H265_MAX_CHROMA_PLANES];
+    int8_t                                delta_chroma_offset_l1[STD_VIDEO_H265_MAX_NUM_LIST_REF][STD_VIDEO_H265_MAX_CHROMA_PLANES];
+} StdVideoEncodeH265WeightTable;
+
 typedef struct StdVideoEncodeH265SliceSegmentHeaderFlags {
     uint32_t    first_slice_segment_in_pic_flag : 1;
     uint32_t    no_output_of_prior_pics_flag : 1;
     uint32_t    dependent_slice_segment_flag : 1;
+    uint32_t    pic_output_flag : 1;
     uint32_t    short_term_ref_pic_set_sps_flag : 1;
     uint32_t    slice_temporal_mvp_enable_flag : 1;
     uint32_t    slice_sao_luma_flag : 1;
@@ -34,47 +58,42 @@ typedef struct StdVideoEncodeH265SliceSegmentHeaderFlags {
     uint32_t    num_ref_idx_active_override_flag : 1;
     uint32_t    mvd_l1_zero_flag : 1;
     uint32_t    cabac_init_flag : 1;
-    uint32_t    slice_deblocking_filter_disable_flag : 1;
+    uint32_t    cu_chroma_qp_offset_enabled_flag : 1;
+    uint32_t    deblocking_filter_override_flag : 1;
+    uint32_t    slice_deblocking_filter_disabled_flag : 1;
     uint32_t    collocated_from_l0_flag : 1;
     uint32_t    slice_loop_filter_across_slices_enabled_flag : 1;
-    uint32_t    bLastSliceInPic : 1;
-    uint32_t    reservedBits : 18;
-    uint16_t    luma_weight_l0_flag;
-    uint16_t    chroma_weight_l0_flag;
-    uint16_t    luma_weight_l1_flag;
-    uint16_t    chroma_weight_l1_flag;
 } StdVideoEncodeH265SliceSegmentHeaderFlags;
 
+typedef struct StdVideoEncodeH265SliceSegmentLongTermRefPics {
+    uint8_t     num_long_term_sps;
+    uint8_t     num_long_term_pics;
+    uint8_t     lt_idx_sps[STD_VIDEO_H265_MAX_LONG_TERM_REF_PICS_SPS];
+    uint8_t     poc_lsb_lt[STD_VIDEO_H265_MAX_LONG_TERM_PICS];
+    uint16_t    used_by_curr_pic_lt_flag;
+    uint8_t     delta_poc_msb_present_flag[STD_VIDEO_H265_MAX_DELTA_POC];
+    uint8_t     delta_poc_msb_cycle_lt[STD_VIDEO_H265_MAX_DELTA_POC];
+} StdVideoEncodeH265SliceSegmentLongTermRefPics;
+
 typedef struct StdVideoEncodeH265SliceSegmentHeader {
-    StdVideoH265SliceType                        slice_type;
-    uint8_t                                      slice_pic_parameter_set_id;
-    uint8_t                                      num_short_term_ref_pic_sets;
-    uint32_t                                     slice_segment_address;
-    uint8_t                                      short_term_ref_pic_set_idx;
-    uint8_t                                      num_long_term_sps;
-    uint8_t                                      num_long_term_pics;
-    uint8_t                                      collocated_ref_idx;
-    uint8_t                                      num_ref_idx_l0_active_minus1;
-    uint8_t                                      num_ref_idx_l1_active_minus1;
-    uint8_t                                      luma_log2_weight_denom;
-    int8_t                                       delta_chroma_log2_weight_denom;
-    int8_t                                       delta_luma_weight_l0[STD_VIDEO_ENCODE_H265_LUMA_LIST_SIZE];
-    int8_t                                       luma_offset_l0[STD_VIDEO_ENCODE_H265_LUMA_LIST_SIZE];
-    int8_t                                       delta_chroma_weight_l0[STD_VIDEO_ENCODE_H265_CHROMA_LIST_SIZE][STD_VIDEO_ENCODE_H265_CHROMA_LISTS_NUM];
-    int8_t                                       delta_chroma_offset_l0[STD_VIDEO_ENCODE_H265_CHROMA_LIST_SIZE][STD_VIDEO_ENCODE_H265_CHROMA_LISTS_NUM];
-    int8_t                                       delta_luma_weight_l1[STD_VIDEO_ENCODE_H265_LUMA_LIST_SIZE];
-    int8_t                                       luma_offset_l1[STD_VIDEO_ENCODE_H265_LUMA_LIST_SIZE];
-    int8_t                                       delta_chroma_weight_l1[STD_VIDEO_ENCODE_H265_CHROMA_LIST_SIZE][STD_VIDEO_ENCODE_H265_CHROMA_LISTS_NUM];
-    int8_t                                       delta_chroma_offset_l1[STD_VIDEO_ENCODE_H265_CHROMA_LIST_SIZE][STD_VIDEO_ENCODE_H265_CHROMA_LISTS_NUM];
-    uint8_t                                      MaxNumMergeCand;
-    int8_t                                       slice_cb_qp_offset;
-    int8_t                                       slice_cr_qp_offset;
-    int8_t                                       slice_beta_offset_div2;
-    int8_t                                       slice_tc_offset_div2;
-    int8_t                                       slice_act_y_qp_offset;
-    int8_t                                       slice_act_cb_qp_offset;
-    int8_t                                       slice_act_cr_qp_offset;
-    StdVideoEncodeH265SliceSegmentHeaderFlags    flags;
+    StdVideoEncodeH265SliceSegmentHeaderFlags               flags;
+    StdVideoH265SliceType                                   slice_type;
+    uint32_t                                                slice_segment_address;
+    uint8_t                                                 short_term_ref_pic_set_idx;
+    uint8_t                                                 collocated_ref_idx;
+    uint8_t                                                 num_ref_idx_l0_active_minus1;
+    uint8_t                                                 num_ref_idx_l1_active_minus1;
+    uint8_t                                                 MaxNumMergeCand;
+    int8_t                                                  slice_cb_qp_offset;
+    int8_t                                                  slice_cr_qp_offset;
+    int8_t                                                  slice_beta_offset_div2;
+    int8_t                                                  slice_tc_offset_div2;
+    int8_t                                                  slice_act_y_qp_offset;
+    int8_t                                                  slice_act_cb_qp_offset;
+    int8_t                                                  slice_act_cr_qp_offset;
+    const StdVideoH265ShortTermRefPicSet*                   pShortTermRefPicSet;
+    const StdVideoEncodeH265SliceSegmentLongTermRefPics*    pLongTermRefPics;
+    const StdVideoEncodeH265WeightTable*                    pWeightTable;
 } StdVideoEncodeH265SliceSegmentHeader;
 
 typedef struct StdVideoEncodeH265ReferenceModificationFlags {
@@ -85,35 +104,38 @@ typedef struct StdVideoEncodeH265ReferenceModificationFlags {
 typedef struct StdVideoEncodeH265ReferenceModifications {
     StdVideoEncodeH265ReferenceModificationFlags    flags;
     uint8_t                                         referenceList0ModificationsCount;
-    uint8_t*                                        pReferenceList0Modifications;
+    const uint8_t*                                  pReferenceList0Modifications;
     uint8_t                                         referenceList1ModificationsCount;
-    uint8_t*                                        pReferenceList1Modifications;
+    const uint8_t*                                  pReferenceList1Modifications;
 } StdVideoEncodeH265ReferenceModifications;
 
 typedef struct StdVideoEncodeH265PictureInfoFlags {
     uint32_t    is_reference_flag : 1;
     uint32_t    IrapPicFlag : 1;
     uint32_t    long_term_flag : 1;
+    uint32_t    discardable_flag : 1;
+    uint32_t    cross_layer_bla_flag : 1;
 } StdVideoEncodeH265PictureInfoFlags;
 
 typedef struct StdVideoEncodeH265PictureInfo {
+    StdVideoEncodeH265PictureInfoFlags    flags;
     StdVideoH265PictureType               PictureType;
     uint8_t                               sps_video_parameter_set_id;
     uint8_t                               pps_seq_parameter_set_id;
+    uint8_t                               pps_pic_parameter_set_id;
     int32_t                               PicOrderCntVal;
     uint8_t                               TemporalId;
-    StdVideoEncodeH265PictureInfoFlags    flags;
 } StdVideoEncodeH265PictureInfo;
 
 typedef struct StdVideoEncodeH265ReferenceInfoFlags {
-    uint32_t    is_long_term : 1;
-    uint32_t    isUsedFlag : 1;
+    uint32_t    used_for_long_term_reference : 1;
+    uint32_t    unused_for_reference : 1;
 } StdVideoEncodeH265ReferenceInfoFlags;
 
 typedef struct StdVideoEncodeH265ReferenceInfo {
+    StdVideoEncodeH265ReferenceInfoFlags    flags;
     int32_t                                 PicOrderCntVal;
     uint8_t                                 TemporalId;
-    StdVideoEncodeH265ReferenceInfoFlags    flags;
 } StdVideoEncodeH265ReferenceInfo;
 
 

libs/bluevk/include/vulkan/vk_platform.h

@@ -2,45 +2,15 @@
 // File: vk_platform.h
 //
 /*
-** Copyright (c) 2014-2017 The Khronos Group Inc.
+** Copyright 2014-2022 The Khronos Group Inc.
 **
-** Licensed under the Apache License, Version 2.0 (the "License");
-** you may not use this file except in compliance with the License.
-** You may obtain a copy of the License at
-**
-**     http://www.apache.org/licenses/LICENSE-2.0
-**
-** Unless required by applicable law or agreed to in writing, software
-** distributed under the License is distributed on an "AS IS" BASIS,
-** WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-** See the License for the specific language governing permissions and
-** limitations under the License.
+** SPDX-License-Identifier: Apache-2.0
 */
 
 
 #ifndef VK_PLATFORM_H_
 #define VK_PLATFORM_H_
 
-#if defined(__ANDROID__)
-#define VK_USE_PLATFORM_ANDROID_KHR 1
-#elif defined(IOS)
-#define VK_USE_PLATFORM_IOS_MVK 1
-#elif defined(__linux__)
-#if defined(FILAMENT_SUPPORTS_XCB)
-#define VK_USE_PLATFORM_XCB_KHR 1
-#endif
-#if defined(FILAMENT_SUPPORTS_XLIB)
-#define VK_USE_PLATFORM_XLIB_KHR 1
-#endif
-#if defined(FILAMENT_SUPPORTS_WAYLAND)
-#define VK_USE_PLATFORM_WAYLAND_KHR 1
-#endif
-#elif defined(__APPLE__)
-#define VK_USE_PLATFORM_MACOS_MVK 1
-#elif defined(WIN32)
-#define VK_USE_PLATFORM_WIN32_KHR 1
-#endif
-
 #ifdef __cplusplus
 extern "C"
 {
@@ -72,7 +42,7 @@ extern "C"
     #define VKAPI_CALL __stdcall
     #define VKAPI_PTR  VKAPI_CALL
 #elif defined(__ANDROID__) && defined(__ARM_ARCH) && __ARM_ARCH < 7
-    #error "Vulkan isn't supported for the 'armeabi' NDK ABI"
+    #error "Vulkan is not supported for the 'armeabi' NDK ABI"
 #elif defined(__ANDROID__) && defined(__ARM_ARCH) && __ARM_ARCH >= 7 && defined(__ARM_32BIT_STATE)
     // On Android 32-bit ARM targets, Vulkan functions use the "hardfloat"
     // calling convention, i.e. float parameters are passed in registers. This
@@ -88,7 +58,9 @@ extern "C"
     #define VKAPI_PTR
 #endif
 
-#include <stddef.h>
+#if !defined(VK_NO_STDDEF_H)
+    #include <stddef.h>
+#endif // !defined(VK_NO_STDDEF_H)
 
 #if !defined(VK_NO_STDINT_H)
     #if defined(_MSC_VER) && (_MSC_VER < 1600)

libs/bluevk/include/vulkan/vulkan.h

@@ -38,7 +38,6 @@
 
 
 #ifdef VK_USE_PLATFORM_WAYLAND_KHR
-#include <wayland-client.h>
 #include "vulkan_wayland.h"
 #endif