Update glslang to e3bca2a

README.md

@@ -31,7 +31,7 @@ repositories {
 }
 
 dependencies {
-    implementation 'com.google.android.filament:filament-android:1.23.0'
+    implementation 'com.google.android.filament:filament-android:1.21.3'
 }
 ```
 
@@ -51,7 +51,7 @@ Here are all the libraries available in the group `com.google.android.filament`:
 iOS projects can use CocoaPods to install the latest release:
 
 ```
-pod 'Filament', '~> 1.23.0'
+pod 'Filament', '~> 1.21.3'
 ```
 
 ### Snapshots

RELEASE_GUIDE.md

@@ -50,13 +50,13 @@ Do not push to origin yet.
 git cherry-pick rc/$RELEASE
 ```
 
-Update the headers. The "main branch" header becomes a header for $NEXT_RELEASE, and a new "main
-branch" header is added.
+Update the headers. The "Next release" header becomes a header for $NEXT_RELEASE, and a new "Next
+release" header is added.
 
 For example, this:
 
 ```
-## main branch
+## Next release (main branch)
 - foo
 - bar
 
@@ -68,7 +68,7 @@ For example, this:
 becomes:
 
 ```
-## main branch
+## Next release (main branch)
 
 ## v1.9.4
 - foo

RELEASE_NOTES.md

@@ -3,37 +3,12 @@
 This file contains one line summaries of commits that are worthy of mentioning in release notes.
 A new header is inserted each time a *tag* is created.
 
-## main branch
-
-## v1.23.0
-
-- engine: Changed UBOs layout [⚠️ **Material breakage**].
-- engine: Normals on morphed models have been fixed (core Filament change).
-- Java: View has several minor changes due to generated code, such as field ordering.
-- gltfio: Fix crash when reloading glTF assets.
-- gltfio: introduce cross-fade animation API [**NEW API**].
-
-## v1.22.2
-
-- Java: Minor API change: rename `ssctStartTraceDistance` to `ssctShadowDistance`. [⚠️ **API Change**]
-- Java: Minor API change: rename `blendingMode` to `blendMode`. [⚠️ **API Change**]
-- engine: Fix some memory leaks.
-
-## v1.22.1
-
-- Metal: Shaders now use `half` floating-point arithmetic when possible for improved performance. [⚠️ **Recompile Materials**]
-- engine: add support for presentation time in `Renderer`
-- engine: added guard bands support for screen-space effects
-- gltfio: Add multi-scene support.
-- gltfio: Various glTF-related cleanup and enhancements.
-- gltfio: Add support for KHR_texture_basisu.
+## v1.22.1 (currently main branch)
 
 ## v1.22.0
 
 - engine: Changed UBOs layout [⚠️ **Material breakage**].
 - engine: Improve effects relying on mipmapping
-- engine: Fix assert seen with VSM shadows.
-- WebGL: Fix `isTextureFormatSupported` for ETC2 formats.
 
 ## v1.21.3
 

android/build.gradle

@@ -81,10 +81,10 @@ buildscript {
         'minSdk': 19,
         'targetSdk': 31,
         'compileSdk': 31,
-        'kotlin': '1.6.21',
-        'kotlin_coroutines': '1.6.1',
+        'kotlin': '1.6.10',
+        'kotlin_coroutines': '1.6.0',
         'buildTools': '32.0.0',
-        'ndk': '24.0.8215888'
+        'ndk': '23.1.7779620'
     ]
 
     ext.deps = [
@@ -101,12 +101,11 @@ buildscript {
 
     dependencies {
         // NOTE: See TODO in gradle.properties once we move to Gradle 7.4
-        classpath 'com.android.tools.build:gradle:7.2.0'
+        classpath 'com.android.tools.build:gradle:7.1.1'
         classpath "org.jetbrains.kotlin:kotlin-gradle-plugin:${versions.kotlin}"
     }
 
     ext.cmakeArgs = [
-        "--no-warn-unused-cli",
         "-DANDROID_PIE=ON",
         "-DANDROID_PLATFORM=21",
         "-DANDROID_STL=c++_static",
@@ -118,6 +117,8 @@ buildscript {
 
     ext.cppFlags = [
         "-std=c++17",
+        "-Wno-unused-command-line-argument",
+        "-Wl,--hash-style=both", // Required to support API levels below 23
         "-fno-stack-protector",
         "-fno-exceptions",
         "-fno-unwind-tables",
@@ -130,13 +131,8 @@ buildscript {
         "-fomit-frame-pointer",
         "-ffunction-sections",
         "-fdata-sections",
-        "-no-canonical-prefixes",
-        "-Wformat",
-        "-Werror=format-security",
-        "-Wno-unused-command-line-argument",
         "-Wl,--gc-sections",
         "-Wl,-Bsymbolic-functions",
-        "-Wl,--hash-style=both", // Required to support API levels below 23
     ]
 
     ext.abis = abis

android/buildSrc/build.gradle

@@ -1,3 +1,4 @@
+
 plugins {
     id 'groovy-gradle-plugin'
 }

android/filamat-android/build.gradle

@@ -1,6 +1,4 @@
 android {
-    namespace 'com.google.android.filament.filamat'
-
     flavorDimensions "functionality"
     productFlavors {
         full {

android/filament-android/build.gradle

@@ -1,7 +1,3 @@
-android {
-    namespace 'com.google.android.filament'
-}
-
 dependencies {
     implementation deps.androidx.annotations
 }

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

@@ -156,9 +156,12 @@ Java_com_google_android_filament_Renderer_nResetUserTime(JNIEnv*, jclass, jlong
 }
 
 extern "C" JNIEXPORT void JNICALL
-Java_com_google_android_filament_Renderer_nSetDisplayInfo(JNIEnv*, jclass, jlong nativeRenderer, jfloat refreshRate) {
+Java_com_google_android_filament_Renderer_nSetDisplayInfo(JNIEnv*, jclass, jlong nativeRenderer,
+        jfloat refreshRate, jlong presentationDeadlineNanos, jlong vsyncOffsetNanos) {
     Renderer *renderer = (Renderer *) nativeRenderer;
-    renderer->setDisplayInfo({ .refreshRate = refreshRate });
+    renderer->setDisplayInfo({ .refreshRate = refreshRate,
+                               .presentationDeadlineNanos = (uint64_t)presentationDeadlineNanos,
+                               .vsyncOffsetNanos = (uint64_t)vsyncOffsetNanos });
 }
 
 extern "C" JNIEXPORT void JNICALL
@@ -180,10 +183,3 @@ Java_com_google_android_filament_Renderer_nSetClearOptions(JNIEnv *, jclass ,
                                 .clear = (bool) clear,
                                 .discard = (bool) discard});
 }
-
-extern "C" JNIEXPORT void JNICALL
-Java_com_google_android_filament_Renderer_nSetPresentationTime(JNIEnv *, jclass ,
-    jlong nativeRenderer, jlong monotonicClockNanos) {
-    Renderer *renderer = (Renderer *) nativeRenderer;
-    renderer->setPresentationTime(monotonicClockNanos);
-}

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

@@ -259,14 +259,14 @@ Java_com_google_android_filament_View_nSetAmbientOcclusionOptions(JNIEnv*, jclas
 
 extern "C" JNIEXPORT void JNICALL
 Java_com_google_android_filament_View_nSetSSCTOptions(JNIEnv *, jclass, jlong nativeView,
-        jfloat ssctLightConeRad, jfloat ssctShadowDistance, jfloat ssctContactDistanceMax,
+        jfloat ssctLightConeRad, jfloat ssctStartTraceDistance, jfloat ssctContactDistanceMax,
         jfloat ssctIntensity, jfloat ssctLightDirX, jfloat ssctLightDirY, jfloat ssctLightDirZ,
         jfloat ssctDepthBias, jfloat ssctDepthSlopeBias, jint ssctSampleCount,
         jint ssctRayCount, jboolean ssctEnabled) {
     View* view = (View*) nativeView;
     View::AmbientOcclusionOptions options = view->getAmbientOcclusionOptions();
     options.ssct.lightConeRad = ssctLightConeRad;
-    options.ssct.shadowDistance = ssctShadowDistance;
+    options.ssct.shadowDistance = ssctStartTraceDistance;
     options.ssct.contactDistanceMax = ssctContactDistanceMax;
     options.ssct.intensity = ssctIntensity;
     options.ssct.lightDirection = math::float3{ ssctLightDirX, ssctLightDirY, ssctLightDirZ };
@@ -461,11 +461,3 @@ Java_com_google_android_filament_View_nPick(JNIEnv* env, jclass,
         JniCallback::postToJavaAndDestroy(callback);
     }, callback->getHandler());
 }
-
-extern "C"
-JNIEXPORT void JNICALL
-Java_com_google_android_filament_View_nSetGuardBandOptions(JNIEnv *, jclass,
-        jlong nativeView, jboolean enabled) {
-    View* view = (View*) nativeView;
-    view->setGuardBandOptions({ .enabled = (bool)enabled });
-}

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

@@ -104,12 +104,12 @@ public class MorphTargetBuffer {
     }
 
     /**
-     * Updates float4 positions for the given morph target.
+     * Updates positions of morph target at the index.
      *
      * @param engine {@link Engine} instance
      * @param targetIndex The index of morph target to be updated
-     * @param positions An array with at least count*4 floats
-     * @param count Number of float4 vectors in positions to be consumed
+     * @param positions Pointer to at least count positions
+     * @param count Number of position elements in positions
      */
     public void setPositionsAt(@NonNull Engine engine,
             @IntRange(from = 0) int targetIndex,
@@ -122,15 +122,12 @@ public class MorphTargetBuffer {
     }
 
     /**
-     * Updates tangents for the given morph target.
-     *
-     * These quaternions must be represented as signed shorts, where real numbers in the [-1,+1]
-     * range multiplied by 32767.
+     * Updates tangents of morph target at the index.
      *
      * @param engine {@link Engine} instance
      * @param targetIndex The index of morph target to be updated
-     * @param tangents An array with at least "count*4" shorts
-     * @param count number of short4 quaternions in tangents
+     * @param tangents Pointer to at least count tangents
+     * @param count Number of tangent elements in tangents
      */
     public void setTangentsAt(@NonNull Engine engine,
             @IntRange(from = 0) int targetIndex,
@@ -143,14 +140,14 @@ public class MorphTargetBuffer {
     }
 
     /**
-     * @return number of vertices in this {@link MorphTargetBuffer}
+     * @return number of vertex count in this {@link MorphTargetBuffer}
      */
     public int getVertexCount() {
         return nGetVertexCount(mNativeObject);
     }
 
     /**
-     * @return number of morph targets in this {@link MorphTargetBuffer}
+     * @return number of target count in this {@link MorphTargetBuffer}
      */
     public int getCount() {
         return nGetCount(mNativeObject);

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

@@ -61,17 +61,13 @@ public class Renderer {
         /**
          * How far in advance a buffer must be queued for presentation at a given time in ns
          * On Android you can use {@link android.view.Display#getPresentationDeadlineNanos()}.
-         * @deprecated this value is ignored
          */
-        @Deprecated
         public long presentationDeadlineNanos = 0;
 
         /**
          * Offset by which vsyncSteadyClockTimeNano provided in beginFrame() is offset in ns
          * On Android you can use {@link android.view.Display#getAppVsyncOffsetNanos()}.
-         * @deprecated this value is ignored
          */
-        @Deprecated
         public long vsyncOffsetNanos = 0;
     };
 
@@ -179,7 +175,8 @@ public class Renderer {
      */
     public void setDisplayInfo(@NonNull DisplayInfo info) {
         mDisplayInfo = info;
-        nSetDisplayInfo(getNativeObject(), info.refreshRate);
+        nSetDisplayInfo(getNativeObject(),
+                info.refreshRate, info.presentationDeadlineNanos, info.vsyncOffsetNanos);
     }
 
     /**
@@ -250,23 +247,6 @@ public class Renderer {
         return mEngine;
     }
 
-    /**
-     * Set the time at which the frame must be presented to the display.
-     * <p>
-     * This must be called between {@link #beginFrame} and {@link #endFrame}.
-     * </p>
-     *
-     * @param monotonicClockNanos  The time in nanoseconds corresponding to the system monotonic
-     *                             up-time clock. The presentation time is typically set in the
-     *                             middle of the period of interest and cannot be too far in the
-     *                             future as it is limited by how many buffers are available in
-     *                             the display sub-system. Typically it is set to 1 or 2 vsync
-     *                             periods away.
-     */
-    public void setPresentationTime(long monotonicClockNanos) {
-        nSetPresentationTime(getNativeObject(), monotonicClockNanos);
-    }
-
     /**
      * Sets up a frame for this <code>Renderer</code>.
      * <p><code>beginFrame</code> manages frame pacing, and returns whether or not a frame should be
@@ -684,7 +664,6 @@ public class Renderer {
         mNativeObject = 0;
     }
 
-    private static native void nSetPresentationTime(long nativeObject, long monotonicClockNanos);
     private static native boolean nBeginFrame(long nativeRenderer, long nativeSwapChain, long frameTimeNanos);
     private static native void nEndFrame(long nativeRenderer);
     private static native void nRender(long nativeRenderer, long nativeView);
@@ -706,7 +685,8 @@ public class Renderer {
             Object handler, Runnable callback);
     private static native double nGetUserTime(long nativeRenderer);
     private static native void nResetUserTime(long nativeRenderer);
-    private static native void nSetDisplayInfo(long nativeRenderer, float refreshRate);
+    private static native void nSetDisplayInfo(long nativeRenderer,
+            float refreshRate, long presentationDeadlineNanos, long vsyncOffsetNanos);
     private static native void nSetFrameRateOptions(long nativeRenderer,
             float interval, float headRoomRatio, float scaleRate, int history);
     private static native void nSetClearOptions(long nativeRenderer,

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

@@ -157,6 +157,8 @@ public class DisplayHelper {
             info = new Renderer.DisplayInfo();
         }
         info.refreshRate = DisplayHelper.getRefreshRate(display);
+        info.presentationDeadlineNanos = DisplayHelper.getPresentationDeadlineNanos(display);
+        info.vsyncOffsetNanos = DisplayHelper.getAppVsyncOffsetNanos(display);
         return info;
     }
 

android/filament-utils-android/build.gradle

@@ -1,8 +1,6 @@
 apply plugin: 'kotlin-android'
 
 android {
-    namespace 'com.google.android.filament.utils'
-
     sourceSets {
         main {
             kotlin.srcDirs += "src/main/java"
@@ -12,14 +10,14 @@ android {
     defaultConfig {
         missingDimensionStrategy 'functionality', 'full'
     }
+
+    // No need to package up the following shared libs, which arise as a side effect of our
+    // externalNativeBuild dependencies. When clients pick and choose from project-level gradle
+    // dependencies, these shared libs already get pulled in, so we need to avoid the error:
+    // "More than one file was found with OS independent path ..."
     packagingOptions {
-        // No need to package up the following shared libs, which arise as a side effect of our
-        // externalNativeBuild dependencies. When clients pick and choose from project-level gradle
-        // dependencies, these shared libs already get pulled in, so we need to avoid the error:
-        // "More than one file was found with OS independent path ..."
-        jniLibs {
-            excludes += ['lib/*/libfilament-jni.so', 'lib/*/libgltfio-jni.so']
-        }
+        exclude 'lib/*/libfilament-jni.so'
+        exclude 'lib/*/libgltfio-jni.so'
     }
 }
 

android/gltfio-android/CMakeLists.txt

@@ -40,33 +40,30 @@ set(CMAKE_SHARED_LINKER_FLAGS_RELEASE "${CMAKE_SHARED_LINKER_FLAGS_RELEASE} -Wl,
 set(GLTFIO_SRCS
         ${GLTFIO_DIR}/include/gltfio/Animator.h
         ${GLTFIO_DIR}/include/gltfio/AssetLoader.h
+        ${GLTFIO_DIR}/include/gltfio/MaterialProvider.h
+        ${GLTFIO_DIR}/include/gltfio/ResourceLoader.h
         ${GLTFIO_DIR}/include/gltfio/FilamentAsset.h
         ${GLTFIO_DIR}/include/gltfio/FilamentInstance.h
-        ${GLTFIO_DIR}/include/gltfio/MaterialProvider.h
-        ${GLTFIO_DIR}/include/gltfio/NodeManager.h
-        ${GLTFIO_DIR}/include/gltfio/ResourceLoader.h
         ${GLTFIO_DIR}/include/gltfio/TextureProvider.h
         ${GLTFIO_DIR}/include/gltfio/math.h
 
         ${GLTFIO_DIR}/src/Animator.cpp
         ${GLTFIO_DIR}/src/AssetLoader.cpp
-        ${GLTFIO_DIR}/src/DependencyGraph.cpp
-        ${GLTFIO_DIR}/src/DependencyGraph.h
         ${GLTFIO_DIR}/src/DracoCache.cpp
         ${GLTFIO_DIR}/src/DracoCache.h
+        ${GLTFIO_DIR}/src/DependencyGraph.cpp
+        ${GLTFIO_DIR}/src/DependencyGraph.h
         ${GLTFIO_DIR}/src/FFilamentAsset.h
-        ${GLTFIO_DIR}/src/FFilamentInstance.h
         ${GLTFIO_DIR}/src/FilamentAsset.cpp
+        ${GLTFIO_DIR}/src/FFilamentInstance.h
         ${GLTFIO_DIR}/src/FilamentInstance.cpp
-        ${GLTFIO_DIR}/src/FNodeManager.h
         ${GLTFIO_DIR}/src/GltfEnums.h
         ${GLTFIO_DIR}/src/Ktx2Provider.cpp
         ${GLTFIO_DIR}/src/MaterialProvider.cpp
-        ${GLTFIO_DIR}/src/NodeManager.cpp
         ${GLTFIO_DIR}/src/ResourceLoader.cpp
         ${GLTFIO_DIR}/src/StbProvider.cpp
-        ${GLTFIO_DIR}/src/TangentsJob.cpp
         ${GLTFIO_DIR}/src/TangentsJob.h
+        ${GLTFIO_DIR}/src/TangentsJob.cpp
         ${GLTFIO_DIR}/src/UbershaderLoader.cpp
         ${GLTFIO_DIR}/src/Wireframe.cpp
         ${GLTFIO_DIR}/src/Wireframe.h

android/gltfio-android/build.gradle

@@ -1,6 +1,4 @@
 android {
-    namespace 'com.google.android.filament.gltfio'
-
     flavorDimensions "functionality"
     productFlavors {
         full {
@@ -17,14 +15,13 @@ android {
             }
         }
     }
+
+    // No need to package up the following shared libs, which arise as a side effect of our
+    // externalNativeBuild dependencies. When clients pick and choose from project-level gradle
+    // dependencies, these shared libs already get pulled in, so we need to avoid the error:
+    // "More than one file was found with OS independent path ..."
     packagingOptions {
-        // No need to package up the following shared libs, which arise as a side effect of our
-        // externalNativeBuild dependencies. When clients pick and choose from project-level gradle
-        // dependencies, these shared libs already get pulled in, so we need to avoid the error:
-        // "More than one file was found with OS independent path ..."
-        jniLibs {
-            excludes += ['lib/*/libfilament-jni.so']
-        }
+        exclude 'lib/*/libfilament-jni.so'
     }
 }
 

android/gltfio-android/src/main/cpp/Animator.cpp

@@ -36,13 +36,6 @@ Java_com_google_android_filament_gltfio_Animator_nUpdateBoneMatrices(JNIEnv*, jc
     animator->updateBoneMatrices();
 }
 
-extern "C" JNIEXPORT void JNICALL
-Java_com_google_android_filament_gltfio_Animator_nApplyCrossFade(JNIEnv*, jclass, jlong nativeAnimator,
-        jint previousAnimIndex, jfloat previousAnimTime, jfloat alpha) {
-    Animator* animator = (Animator*) nativeAnimator;
-    animator->applyCrossFade(previousAnimIndex, previousAnimTime, alpha);
-}
-
 extern "C" JNIEXPORT void JNICALL
 Java_com_google_android_filament_gltfio_Animator_nResetBoneMatrices(JNIEnv*, jclass, jlong nativeAnimator) {
     Animator* animator = (Animator*) nativeAnimator;

android/gltfio-android/src/main/cpp/FilamentAsset.cpp

@@ -120,34 +120,10 @@ extern "C" JNIEXPORT void JNICALL
 Java_com_google_android_filament_gltfio_FilamentAsset_nGetLightEntities(JNIEnv* env, jclass,
         jlong nativeAsset, jintArray result) {
     FilamentAsset* asset = (FilamentAsset*) nativeAsset;
-    const jsize available = env->GetArrayLength(result);
-    const size_t minCount = std::min(available, (jsize) asset->getLightEntityCount());
-    if (minCount == 0) {
-        return;
-    }
+    jsize available = env->GetArrayLength(result);
     Entity* entities = (Entity*) env->GetIntArrayElements(result, nullptr);
-    std::copy_n(asset->getLightEntities(), minCount, entities);
-    env->ReleaseIntArrayElements(result, (jint*) entities, 0);
-}
-
-extern "C" JNIEXPORT jint JNICALL
-Java_com_google_android_filament_gltfio_FilamentAsset_nGetRenderableEntityCount(JNIEnv*, jclass,
-        jlong nativeAsset) {
-    FilamentAsset* asset = (FilamentAsset*) nativeAsset;
-    return asset->getRenderableEntityCount();
-}
-
-extern "C" JNIEXPORT void JNICALL
-Java_com_google_android_filament_gltfio_FilamentAsset_nGetRenderableEntities(JNIEnv* env, jclass,
-        jlong nativeAsset, jintArray result) {
-    FilamentAsset* asset = (FilamentAsset*) nativeAsset;
-    const jsize available = env->GetArrayLength(result);
-    const size_t minCount = std::min(available, (jsize) asset->getRenderableEntityCount());
-    if (minCount == 0) {
-        return;
-    }
-    Entity* entities = (Entity*) env->GetIntArrayElements(result, nullptr);
-    std::copy_n(asset->getRenderableEntities(), minCount, entities);
+    std::copy_n(asset->getLightEntities(),
+            std::min(available, (jsize) asset->getLightEntityCount()), entities);
     env->ReleaseIntArrayElements(result, (jint*) entities, 0);
 }
 
@@ -155,13 +131,10 @@ extern "C" JNIEXPORT void JNICALL
 Java_com_google_android_filament_gltfio_FilamentAsset_nGetCameraEntities(JNIEnv* env, jclass,
         jlong nativeAsset, jintArray result) {
     FilamentAsset* asset = (FilamentAsset*) nativeAsset;
-    const jsize available = env->GetArrayLength(result);
-    const size_t minCount = std::min(available, (jsize) asset->getCameraEntityCount());
-    if (minCount == 0) {
-        return;
-    }
+    jsize available = env->GetArrayLength(result);
     Entity* entities = (Entity*) env->GetIntArrayElements(result, nullptr);
-    std::copy_n(asset->getCameraEntities(), minCount, entities);
+    std::copy_n(asset->getCameraEntities(),
+            std::min(available, (jsize) asset->getCameraEntityCount()), entities);
     env->ReleaseIntArrayElements(result, (jint*) entities, 0);
 }
 

android/gltfio-android/src/main/java/com/google/android/filament/gltfio/Animator.java

@@ -66,27 +66,6 @@ public class Animator {
         nUpdateBoneMatrices(getNativeObject());
     }
 
-    /**
-     * Applies a blended transform to the union of nodes affected by two animations.
-     * Used for cross-fading from a previous skinning-based animation or rigid body animation.
-     *
-     * First, this stashes the current transform hierarchy into a transient memory buffer.
-     *
-     * Next, this applies previousAnimIndex / previousAnimTime to the actual asset by internally
-     * calling applyAnimation().
-     *
-     * Finally, the stashed local transforms are lerped (via the scale / translation / rotation
-     * components) with their live counterparts, and the results are pushed to the asset.
-     *
-     * To achieve a cross fade effect with skinned models, clients will typically call animator
-     * methods in this order: (1) applyAnimation (2) applyCrossFade (3) updateBoneMatrices. The
-     * animation that clients pass to applyAnimation is the "current" animation corresponding to
-     * alpha=1, while the "previous" animation passed to applyCrossFade corresponds to alpha=0.
-     */
-    public void applyCrossFade(int previousAnimIndex, float previousAnimTime, float alpha) {
-        nApplyCrossFade(getNativeObject(), previousAnimIndex, previousAnimTime, alpha);
-    }
-
     /**
      * Pass the identity matrix into all bone nodes, useful for returning to the T pose.
      *
@@ -139,7 +118,6 @@ public class Animator {
 
     private static native void nApplyAnimation(long nativeAnimator, int index, float time);
     private static native void nUpdateBoneMatrices(long nativeAnimator);
-    private static native void nApplyCrossFade(long nativeAnimator, int animIndex, float animTime, float alpha);
     private static native void nResetBoneMatrices(long nativeAnimator);
     private static native int nGetAnimationCount(long nativeAnimator);
     private static native float nGetAnimationDuration(long nativeAnimator, int index);

android/gltfio-android/src/main/java/com/google/android/filament/gltfio/FilamentAsset.java

@@ -30,9 +30,9 @@ import com.google.android.filament.MaterialInstance;
  *
  * <p>For usage instructions, see the documentation for {@link AssetLoader}.</p>
  *
- * <p>This class owns a hierarchy of entities that have been loaded from a glTF asset. Every entity
- * has a <code>TransformManager</code> component, and some entities also have compnents managed by
- * <code>NameComponentManager</code>, <code>RenderableManager</code>, and others.</p>
+ * <p>This class owns a hierarchy of entities that have been loaded from a glTF asset. Every entity has
+ * a <code>TransformManager</code> component, and some entities also have
+ * <code>NameComponentManager</code> and/or <code>RenderableManager</code> components.</p>
  *
  * <p>In addition to the aforementioned entities, an asset has strong ownership over a list of
  * <code>VertexBuffer</code>, <code>IndexBuffer</code>, <code>MaterialInstance</code>, and
@@ -113,15 +113,6 @@ public class FilamentAsset {
         return result;
     }
 
-    /**
-     * Gets only the entities that have renderable components.
-     */
-    public @NonNull @Entity int[] getRenderableEntities() {
-        int[] result = new int[nGetRenderableEntityCount(mNativeObject)];
-        nGetRenderableEntities(mNativeObject, result);
-        return result;
-    }
-
     /**
      * Gets only the entities that have camera components.
      *
@@ -330,9 +321,6 @@ public class FilamentAsset {
     private static native int nGetLightEntityCount(long nativeAsset);
     private static native void nGetLightEntities(long nativeAsset, int[] result);
 
-    private static native int nGetRenderableEntityCount(long nativeAsset);
-    private static native void nGetRenderableEntities(long nativeAsset, int[] result);
-
     private static native int nGetCameraEntityCount(long nativeAsset);
     private static native void nGetCameraEntities(long nativeAsset, int[] result);
 

android/gradle.properties

@@ -1,5 +1,5 @@
 GROUP=com.google.android.filament
-VERSION_NAME=1.23.0
+VERSION_NAME=1.21.3
 
 POM_DESCRIPTION=Real-time physically based rendering engine for Android.
 

android/gradle/wrapper/gradle-wrapper.properties

@@ -1,6 +1,6 @@
 #Wed Nov 17 10:40:18 PST 2021
 distributionBase=GRADLE_USER_HOME
 distributionPath=wrapper/dists
-distributionUrl=https\://services.gradle.org/distributions/gradle-7.3.3-bin.zip
+distributionUrl=https\://services.gradle.org/distributions/gradle-7.2-bin.zip
 zipStoreBase=GRADLE_USER_HOME
 zipStorePath=wrapper/dists

android/samples/sample-gltf-viewer/build.gradle

@@ -25,8 +25,6 @@ clean.doFirst {
 }
 
 android {
-    namespace 'com.google.android.filament.gltf'
-
     compileSdkVersion versions.compileSdk
     defaultConfig {
         applicationId "com.google.android.filament.gltf"
@@ -34,7 +32,6 @@ android {
         targetSdkVersion versions.targetSdk
         missingDimensionStrategy 'functionality', 'full'
     }
-
     // NOTE: This is a workaround required because the AGP task collectReleaseDependencies
     //       is not configuration-cache friendly yet; this is only useful for Play publication
     dependenciesInfo {

android/samples/sample-hello-camera/build.gradle

@@ -16,8 +16,6 @@ clean.doFirst {
 }
 
 android {
-    namespace 'com.google.android.filament.hellocam'
-
     compileSdkVersion versions.compileSdk
     defaultConfig {
         applicationId "com.google.android.filament.hellocamera"

android/samples/sample-hello-triangle/build.gradle

@@ -16,8 +16,6 @@ clean.doFirst {
 }
 
 android {
-    namespace 'com.google.android.filament.hellotriangle'
-
     compileSdkVersion versions.compileSdk
     defaultConfig {
         applicationId "com.google.android.filament.hellotriangle"

android/samples/sample-image-based-lighting/build.gradle

@@ -22,8 +22,6 @@ clean.doFirst {
 }
 
 android {
-    namespace 'com.google.android.filament.ibl'
-
     compileSdkVersion versions.compileSdk
     defaultConfig {
         applicationId "com.google.android.filament.ibl"

android/samples/sample-lit-cube/build.gradle

@@ -15,8 +15,6 @@ clean.doFirst {
     delete "src/main/assets"
 }
 android {
-    namespace 'com.google.android.filament.litcube'
-
     compileSdkVersion versions.compileSdk
     defaultConfig {
         applicationId "com.google.android.filament.litcube"

android/samples/sample-live-wallpaper/build.gradle

@@ -4,8 +4,6 @@ apply plugin: 'kotlin-android'
 project.ext.isSample = true
 
 android {
-    namespace 'com.google.android.filament.livewallpaper'
-
     compileSdkVersion versions.compileSdk
     defaultConfig {
         applicationId "com.google.android.filament.livewallpaper"

android/samples/sample-material-builder/build.gradle

@@ -19,8 +19,6 @@ clean.doFirst {
 }
 
 android {
-    namespace 'com.google.android.filament.material_builder'
-
     compileSdkVersion versions.compileSdk
     defaultConfig {
         applicationId "com.google.android.filament.material_builder"

android/samples/sample-multi-view/build.gradle

@@ -15,8 +15,6 @@ clean.doFirst {
     delete "src/main/assets"
 }
 android {
-    namespace 'com.google.android.filament.multiview'
-
     compileSdkVersion versions.compileSdk
     defaultConfig {
         applicationId "com.google.android.filament.multiview"

android/samples/sample-page-curl/build.gradle

@@ -18,8 +18,6 @@ clean.doFirst {
 }
 
 android {
-    namespace 'com.google.android.filament.pagecurl'
-
     compileSdkVersion versions.compileSdk
     defaultConfig {
         applicationId "com.google.android.filament.pagecurl"

android/samples/sample-stream-test/build.gradle

@@ -16,8 +16,6 @@ clean.doFirst {
 }
 
 android {
-    namespace 'com.google.android.filament.streamtest'
-
     compileSdkVersion versions.compileSdk
     defaultConfig {
         applicationId "com.google.android.filament.streamtest"

android/samples/sample-texture-view/build.gradle

@@ -16,8 +16,6 @@ clean.doFirst {
 }
 
 android {
-    namespace 'com.google.android.filament.textureview'
-
     compileSdkVersion versions.compileSdk
     defaultConfig {
         applicationId "com.google.android.filament.textureview"

android/samples/sample-textured-object/build.gradle

@@ -22,8 +22,6 @@ clean.doFirst {
 }
 
 android {
-    namespace 'com.google.android.filament.textured'
-
     compileSdkVersion versions.compileSdk
     defaultConfig {
         applicationId "com.google.android.filament.textured"

android/samples/sample-transparent-view/build.gradle

@@ -16,8 +16,6 @@ clean.doFirst {
 }
 
 android {
-    namespace 'com.google.android.filament.transparentrendering'
-
     compileSdkVersion versions.compileSdk
     defaultConfig {
         applicationId "com.google.android.filament.textureview"

build/android/build.sh

@@ -49,8 +49,7 @@ FILAMENT_NDK_VERSION=${FILAMENT_NDK_VERSION:-$(cat `dirname $0`/ndk.version)}
 
 # Install the required NDK version specifically (if not present)
 if [[ ! -d "${ANDROID_HOME}/ndk/$FILAMENT_NDK_VERSION" ]]; then
-    # NOTE: We MUST use Java 1.8 to run sdkmanager currently, it fails starting with Java 11
-    JAVA_HOME=${JAVA_HOME_8_X64} ${ANDROID_HOME}/tools/bin/sdkmanager "ndk;$FILAMENT_NDK_VERSION" > /dev/null
+    ${ANDROID_HOME}/tools/bin/sdkmanager "ndk;$FILAMENT_NDK_VERSION" > /dev/null
 fi
 
 # Only build 1 64 bit target during presubmit to cut down build times during presubmit

build/android/ndk.version

@@ -1 +1 @@
-24.0.8215888
+23.1.7779620

docs/Materials.html

@@ -118,7 +118,6 @@ counter-increment: h6;margin-right:10px}
 &nbsp;&nbsp;&nbsp;&nbsp;<a href="#materialdefinitions/materialblock/shading:customsurfaceshading" class="level3"><span class="tocNumber">4.2.31&nbsp; </span>Shading: customSurfaceShading</a><br>
 &nbsp;&nbsp;<a href="#materialdefinitions/vertexblock" class="level2"><span class="tocNumber">4.3&nbsp; </span>Vertex block</a><br>
 &nbsp;&nbsp;&nbsp;&nbsp;<a href="#materialdefinitions/vertexblock/materialvertexinputs" class="level3"><span class="tocNumber">4.3.1&nbsp; </span>Material vertex inputs</a><br>
-&nbsp;&nbsp;&nbsp;&nbsp;<a href="#materialdefinitions/vertexblock/customvertexattributes" class="level3"><span class="tocNumber">4.3.2&nbsp; </span>Custom vertex attributes</a><br>
 &nbsp;&nbsp;<a href="#materialdefinitions/fragmentblock" class="level2"><span class="tocNumber">4.4&nbsp; </span>Fragment block</a><br>
 &nbsp;&nbsp;&nbsp;&nbsp;<a href="#materialdefinitions/fragmentblock/preparematerialfunction" class="level3"><span class="tocNumber">4.4.1&nbsp; </span>prepareMaterial function</a><br>
 &nbsp;&nbsp;&nbsp;&nbsp;<a href="#materialdefinitions/fragmentblock/materialfragmentinputs" class="level3"><span class="tocNumber">4.4.2&nbsp; </span>Material fragment inputs</a><br>
@@ -1537,7 +1536,7 @@ non-shader data.
 
 </p><dl><dt>Type</dt><dd><p>     array of <code>string</code>
 
-</p></dd><dt>Value</dt><dd><p>      Each entry must be any of <code>dynamicLighting</code>, <code>directionalLighting</code>, <code>shadowReceiver</code>,<code>skinning</code> or <code>ssr</code>.
+</p></dd><dt>Value</dt><dd><p>      Each entry must be any of <code>dynamicLighting</code>, <code>directionalLighting</code>, <code>shadowReceiver</code> or <code>skinning</code>.
 
 </p></dd><dt>Description</dt><dd><p>      Used to specify a list of shader variants that the application guarantees will never be
       needed. These shader variants are skipped during the code generation phase, thus reducing
@@ -1562,9 +1561,7 @@ non-shader data.
 </li>
 <li class="minus"><code>fog</code>, used when global fog is applied to the scene
 </li>
-<li class="minus"><code>vsm</code>, used when VSM shadows are enabled and the object is a shadow receiver
-</li>
-<li class="minus"><code>ssr</code>, used when screen-space reflections are enabled in the View</li></ul>
+<li class="minus"><code>vsm</code>, used when VSM shadows are enabled and the object is a shadow receiver</li></ul>
 
 <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>
@@ -1610,8 +1607,7 @@ non-shader data.
 
 </p><dl><dt>Type</dt><dd><p>     array of <code>string</code>
 
-</p></dd><dt>Value</dt><dd><p>      Each entry must be any of <code>uv0</code>, <code>uv1</code>, <code>color</code>, <code>position</code>, <code>tangents</code>, <code>custom0</code>
-      through <code>custom7</code>.
+</p></dd><dt>Value</dt><dd><p>      Each entry must be any of <code>uv0</code>, <code>uv1</code>, <code>color</code>, <code>position</code>, <code>tangents</code>.
 
 </p></dd><dt>Description</dt><dd><p>      Lists the vertex attributes required by the material. The <code>position</code> attribute is always
       required and does not need to be specified. The <code>tangents</code> attribute is automatically required
@@ -1626,8 +1622,7 @@ non-shader data.
 <span class="line">        },</span>
 <span class="line">    ],</span>
 <span class="line">    requires : [</span>
-<span class="line">        uv0,</span>
-<span class="line">        custom0</span>
+<span class="line">        uv0</span>
 <span class="line">    ],</span>
 <span class="line">    shadingModel : lit,</span>
 <span class="line">}</span>
@@ -1636,7 +1631,6 @@ non-shader data.
 <span class="line">    void material(inout MaterialInputs material) {</span>
 <span class="line">        prepareMaterial(material);</span>
 <span class="line">        material.baseColor = texture(materialParams_texture, getUV0());</span>
-<span class="line">        material.baseColor.rgb *= getCustom0().rgb;</span>
 <span class="line">    }</span>
 <span class="line">}</span></code></pre>
 <a class="target" name="vertexandattributes:variables">&nbsp;</a><a class="target" name="materialdefinitions/materialblock/vertexandattributes:variables">&nbsp;</a><a class="target" name="toc4.2.8">&nbsp;</a><h3>Vertex and attributes: variables</h3>
@@ -1765,9 +1759,8 @@ non-shader data.
           render target is brightened.
 </li>
     <li class="minus"><strong class="asterisk">Masked</strong>: blending is disabled. This blending mode enables alpha masking. The alpha channel
-          of the material's output defines whether a fragment is discarded or not. Additionally,
-          ALPHA_TO_COVERAGE is enabled for non-translucent views. See the maskThreshold section for more
-          information.</li></ul>
+          of the material's output defines whether a fragment is discarded or not. See the maskThreshold
+          section for more information.</li></ul>
 
 <p></p></dd></dl><p></p><pre class="listing tilde"><code><span class="line">material {</span>
 <span class="line">    blending : transparent</span>
@@ -1858,9 +1851,8 @@ and sorting issues are minimized or eliminated</span></center></div></center>
 </p></dd><dt>Value</dt><dd><p>      A value between <code>0.0</code> and <code>1.0</code>. Defaults to <code>0.4</code>.
 
 </p></dd><dt>Description</dt><dd><p>      Sets the minimum alpha value a fragment must have to not be discarded when the <code>blending</code> mode
-      is set to <code>masked</code>. If the fragment is not discarded, its source alpha is set to 1. When the
-      blending mode is not <code>masked</code>, this value is ignored. This value can be used to controlled the
-      appearance of alpha-masked objects.
+      is set to <code>masked</code>. When the blending mode is not <code>masked</code>, this value is ignored. This value
+      can be used to controlled the appearance of alpha-masked objects.
 
 </p></dd></dl><p></p><pre class="listing tilde"><code><span class="line">material {</span>
 <span class="line">    blending : masked,</span>
@@ -2282,31 +2274,15 @@ APIs listed in the <a href="#shaderpublicapis">Shader public APIs</a> section.
     this behavior using the <code>flipUV</code> property and setting it to <code>false</code>.</p></div>
 
 <p></p>
-<a class="target" name="customvertexattributes">&nbsp;</a><a class="target" name="materialdefinitions/vertexblock/customvertexattributes">&nbsp;</a><a class="target" name="toc4.3.2">&nbsp;</a><h3>Custom vertex attributes</h3>
-<p>
-
-
-You can use up to 8 custom vertex attributes, all of type <code>float4</code>. These attributes can be accessed
-using the vertex block shader functions <code>getCustom0()</code> to <code>getCustom7()</code>. However, before using
-custom attributes, you <em class="asterisk">must</em> declare those attributes as required in the <code>requires</code> property of
-the material:
-
-</p><pre class="listing tilde"><code><span class="line">material {</span>
-<span class="line">    requires : [</span>
-<span class="line">        custom0,</span>
-<span class="line">        custom1,</span>
-<span class="line">        custom2</span>
-<span class="line">    ]</span>
-<span class="line">}</span></code></pre>
 <a class="target" name="fragmentblock">&nbsp;</a><a class="target" name="materialdefinitions/fragmentblock">&nbsp;</a><a class="target" name="toc4.4">&nbsp;</a><h2>Fragment block</h2>
 <p>
 
 
-The fragment block must be used to control the fragment shading stage of the material. The fragment
+The fragment block must be used to control the fragment shading stage of the material. The vertex
 block must contain valid
 <a href="https://www.khronos.org/registry/OpenGL/specs/es/3.0/GLSL_ES_Specification_3.00.pdf">ESSL 3.0</a>
 code (the version of GLSL supported in OpenGL ES 3.0). You are free to create multiple functions
-inside the fragment block but you <strong class="asterisk">must</strong> declare the <code>material</code> function:
+inside the vertex block but you <strong class="asterisk">must</strong> declare the <code>material</code> function:
 
 </p><pre class="listing tilde"><code><span class="line">fragment {</span>
 <span class="line">    <span class="hljs-type">void</span> material(<span class="hljs-keyword">inout</span> MaterialInputs material) {</span>
@@ -2340,7 +2316,7 @@ standard lit shading model:
 
 
 Note that you <strong class="asterisk">must</strong> call <code>prepareMaterial(material)</code> before exiting the  <code>material()</code> function.
-This <code>prepareMaterial</code> function sets up the internal state of the material model. Some of the APIs
+This <code>prepareMaterial</code> function sets up the internal state of the material mdoel. Some of the APIs
 described in the Fragment APIs section - like <code>shading_normal</code> for instance - can only be accessed
 <em class="underscore">after</em> invoking <code>prepareMaterial()</code>.
 
@@ -2640,7 +2616,7 @@ type aliases:
 <p>
 
 </p><div class="table"><table class="table"><tbody><tr><th style="text-align:left"> Name </th><th style="text-align:center"> Type </th><th style="text-align:left"> Description </th></tr>
-<tr><td style="text-align:left"> <strong class="asterisk">getResolution()</strong> </td><td style="text-align:center"> float4 </td><td style="text-align:left"> Dimensions of the view's effective viewport in pixels: <code>width</code>, <code>height</code>, <code>1 / width</code>, <code>1 / height</code>. This might be different from <code>View::getViewport()</code> for instance because of added rendering guard-bands. This can be used in conjunction with <code>getNormalizedViewportCoord()</code> to generate pixel coordinates. </td></tr>
+<tr><td style="text-align:left"> <strong class="asterisk">getResolution()</strong> </td><td style="text-align:center"> float4 </td><td style="text-align:left"> Resolution of the view in pixels: <code>width</code>, <code>height</code>, <code>1 / width</code>, <code>1 / height</code> </td></tr>
 <tr><td style="text-align:left"> <strong class="asterisk">getWorldCameraPosition()</strong> </td><td style="text-align:center"> float3 </td><td style="text-align:left"> Position of the camera/eye in world space </td></tr>
 <tr><td style="text-align:left"> <strong class="asterisk">getWorldOffset()</strong> </td><td style="text-align:center"> float3 </td><td style="text-align:left"> The shift required to obtain API-level world space </td></tr>
 <tr><td style="text-align:left"> <strong class="asterisk">getTime()</strong> </td><td style="text-align:center"> float </td><td style="text-align:left"> Current time as a remainder of 1 second. Yields a value between 0 and 1 </td></tr>
@@ -2845,9 +2821,7 @@ variants:
 </li>
 <li class="minus"><code>fog</code>, used when global fog is applied to the scene
 </li>
-<li class="minus"><code>vsm</code>, used when VSM shadows are enabled and the object is a shadow receiver
-</li>
-<li class="minus"><code>ssr</code>, used when screen-space reflections are enabled in the View</li></ul>
+<li class="minus"><code>vsm</code>, used when VSM shadows are enabled and the object is a shadow receiver</li></ul>
 
 <p></p><p>
 

docs/Materials.md.html

@@ -1125,8 +1125,7 @@ Type
 :    array of `string`
 
 Value
-:     Each entry must be any of `uv0`, `uv1`, `color`, `position`, `tangents`, `custom0`
-      through `custom7`.
+:     Each entry must be any of `uv0`, `uv1`, `color`, `position`, `tangents`.
 
 Description
 :     Lists the vertex attributes required by the material. The `position` attribute is always
@@ -1143,8 +1142,7 @@ material {
         },
     ],
     requires : [
-        uv0,
-        custom0
+        uv0
     ],
     shadingModel : lit,
 }
@@ -1153,7 +1151,6 @@ fragment {
     void material(inout MaterialInputs material) {
         prepareMaterial(material);
         material.baseColor = texture(materialParams_texture, getUV0());
-        material.baseColor.rgb *= getCustom0().rgb;
     }
 }
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
@@ -1844,30 +1841,13 @@ struct MaterialVertexInputs {
     of the current mesh: `material.uv0 = vec2(mesh_uv0.x, 1.0 - mesh_uv0.y)`. You can control
     this behavior using the `flipUV` property and setting it to `false`.
 
-### Custom vertex attributes
-
-You can use up to 8 custom vertex attributes, all of type `float4`. These attributes can be accessed
-using the vertex block shader functions `getCustom0()` to `getCustom7()`. However, before using
-custom attributes, you *must* declare those attributes as required in the `requires` property of
-the material:
-
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ JSON
-material {
-    requires : [
-        custom0,
-        custom1,
-        custom2
-    ]
-}
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-
 ## Fragment block
 
-The fragment block must be used to control the fragment shading stage of the material. The fragment
+The fragment block must be used to control the fragment shading stage of the material. The vertex
 block must contain valid
 [ESSL 3.0](https://www.khronos.org/registry/OpenGL/specs/es/3.0/GLSL_ES_Specification_3.00.pdf)
 code (the version of GLSL supported in OpenGL ES 3.0). You are free to create multiple functions
-inside the fragment block but you **must** declare the `material` function:
+inside the vertex block but you **must** declare the `material` function:
 
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ GLSL
 fragment {
@@ -1902,7 +1882,7 @@ fragment {
 ### prepareMaterial function
 
 Note that you **must** call `prepareMaterial(material)` before exiting the  `material()` function.
-This `prepareMaterial` function sets up the internal state of the material model. Some of the APIs
+This `prepareMaterial` function sets up the internal state of the material mdoel. Some of the APIs
 described in the Fragment APIs section - like `shading_normal` for instance - can only be accessed
 _after_ invoking `prepareMaterial()`.
 

docs/viewer/filament-viewer.js

@@ -15,17 +15,22 @@
 */
 
 // If you are bundling this with rollup, webpack, or esbuild, the following URL should be trimmed.
-import { LitElement, html, css } from "https://unpkg.com/lit?module";
+import { LitElement, html, css } from "https://unpkg.com/lit-element?module";
 
-// This little utility checks if the Filament module is ready for action.
-// If so, it immediately calls the given function. If not, it asks the Filament
-// loader to call it as soon as the module becomes ready.
+// To allow the DOM to render before the Filament WASM module is ready, we maintain a little
+// queue of tasks that get invoked as soon as the module is done loading.
 class FilamentTasks {
+    constructor() {
+        this.tasks = []
+        if (!Filament.Engine) {
+            Filament.init([], () => { for (const task of this.tasks) task(); });
+        }
+    }
     add(callback) {
-        if (Filament.isReady) {
-            callback();
+        if (!Filament.Engine) {
+            this.tasks.push(callback);
         } else {
-            Filament.init([], callback);
+            callback();
         }
     }
 }
@@ -63,7 +68,6 @@ class FilamentViewer extends LitElement {
         this.sky = null;          // Path to skybox ktx.
         this.enableDrop = null;   // Enables drag and drop.
         this.intensity = 30000;   // Intensity of the image based light.
-        this.materialVariant = 0; // Index of material variant.
 
         // Private properties:
         this.filamentTasks = new FilamentTasks();
@@ -80,7 +84,6 @@ class FilamentViewer extends LitElement {
             sky: { type: String },
             enableDrop: { type: Boolean },
             intensity: { type: Number },
-            materialVariant: { type: Number },
         }
     }
 
@@ -113,7 +116,6 @@ class FilamentViewer extends LitElement {
         if (props.has("intensity") && this.indirectLight) {
             this.indirectLight.setIntensity(this.intensity);
         }
-        if (props.has("materialVariant") && this.asset) this._applyMaterialVariant();
     }
 
     static get styles() {
@@ -246,7 +248,7 @@ class FilamentViewer extends LitElement {
             return response.arrayBuffer();
         }).then(arrayBuffer => {
             const ktxData = new Uint8Array(arrayBuffer);
-            this.indirectLight = this.engine.createIblFromKtx1(ktxData);
+            this.indirectLight = this.engine.createIblFromKtx(ktxData);
             this.indirectLight.setIntensity(this.intensity);
             this.scene.setIndirectLight(this.indirectLight);
         });
@@ -264,7 +266,7 @@ class FilamentViewer extends LitElement {
             return response.arrayBuffer();
         }).then(arrayBuffer => {
             const ktxData = new Uint8Array(arrayBuffer);
-            this.skybox = this.engine.createSkyFromKtx1(ktxData);
+            this.skybox = this.engine.createSkyFromKtx(ktxData);
             this.scene.setSkybox(this.skybox);
         });
     }
@@ -309,7 +311,7 @@ class FilamentViewer extends LitElement {
                 asyncInterval: 30
             };
 
-            const doneAddingResources = (resourceLoader, stbProvider, ktx2Provider) => {
+            const doneAddingResources = resourceLoader => {
                 this.srcBlobResources = {};
                 resourceLoader.asyncBeginLoad(this.asset);
                 const timer = setInterval(() => {
@@ -318,8 +320,6 @@ class FilamentViewer extends LitElement {
                     if (progress >= 1) {
                         clearInterval(timer);
                         resourceLoader.delete();
-                        stbProvider.delete();
-                        ktx2Provider.delete();
                         this.animator = this.asset.getAnimator();
                         this.animationStartTime = Date.now();
                     }
@@ -337,20 +337,13 @@ class FilamentViewer extends LitElement {
                     config.recomputeBoundingBoxes,
                     config.ignoreBindTransform);
 
-                const stbProvider = new Filament.gltfio$StbProvider(this.engine);
-                const ktx2Provider = new Filament.gltfio$Ktx2Provider(this.engine);
-
-                resourceLoader.addStbProvider("image/jpeg", stbProvider);
-                resourceLoader.addStbProvider("image/png", stbProvider);
-                resourceLoader.addKtx2Provider("image/ktx2", ktx2Provider);
-
                 let remaining = Object.keys(this.srcBlobResources).length;
                 for (const name in this.srcBlobResources) {
                     this.srcBlobResources[name].arrayBuffer().then(buffer => {
                         const desc = getBufferDescriptor(new Uint8Array(buffer));
                         resourceLoader.addResourceData(name, getBufferDescriptor(desc));
                         if (--remaining === 0) {
-                            doneAddingResources(resourceLoader, stbProvider, ktx2Provider);
+                            doneAddingResources(resourceLoader);
                         }
                     });
                 }
@@ -382,7 +375,6 @@ class FilamentViewer extends LitElement {
             this.asset.loadResources(() => {
                 this.animator = this.asset.getAnimator();
                 this.animationStartTime = Date.now();
-                this._applyMaterialVariant();
             }, null, basePath);
 
             this._updateOverlay();
@@ -391,11 +383,9 @@ class FilamentViewer extends LitElement {
 
     _updateAsset() {
         // Invoke the first glTF animation if it exists.
-        if (this.animator) {
-            if (this.animator.getAnimationCount() > 0) {
-                const ms = Date.now() - this.animationStartTime;
-                this.animator.applyAnimation(0, ms / 1000);
-            }
+        if (this.animator && this.animator.getAnimationCount() > 0) {
+            const ms = Date.now() - this.animationStartTime;
+            this.animator.applyAnimation(0, ms / 1000);
             this.animator.updateBoneMatrices();
         }
 
@@ -436,20 +426,6 @@ class FilamentViewer extends LitElement {
 
         window.requestAnimationFrame(this._renderFrame.bind(this));
     }
-
-    _applyMaterialVariant() {
-        if (!this.hasAttribute("materialVariant")) {
-            return;
-        }
-        const names = this.asset.getMaterialVariantNames();
-        const index = this.materialVariant;
-        if (index < 0 || index >= names.length) {
-            console.error(`Material variant ${index} does not exist in this asset.`);
-            return;
-        }
-        console.info(this.src, `Applying material variant: ${names[index]}`);
-        this.asset.applyMaterialVariant(index);
-    }
 }
 
 customElements.define("filament-viewer", FilamentViewer);

docs/viewer/index.html

@@ -43,7 +43,7 @@ filament-viewer::part(canvas) {
 </p>
 
 </main>
-<script src="https://unpkg.com/filament@1.22.0/filament.js"></script>
+<script src="https://unpkg.com/filament/filament.js"></script>
 <script src="https://unpkg.com/gltumble"></script>
 <script src="filament-viewer.js" type="module"></script>
 </body>

filament/CMakeLists.txt

@@ -456,34 +456,6 @@ add_custom_command(
         APPEND
 )
 
-add_custom_command(
-        OUTPUT "${MATERIAL_DIR}/separableGaussianBlur1L.filamat"
-        DEPENDS src/materials/separableGaussianBlur.vs
-        DEPENDS src/materials/separableGaussianBlur.fs
-        APPEND
-)
-
-add_custom_command(
-        OUTPUT "${MATERIAL_DIR}/separableGaussianBlur2L.filamat"
-        DEPENDS src/materials/separableGaussianBlur.vs
-        DEPENDS src/materials/separableGaussianBlur.fs
-        APPEND
-)
-
-add_custom_command(
-        OUTPUT "${MATERIAL_DIR}/separableGaussianBlur3L.filamat"
-        DEPENDS src/materials/separableGaussianBlur.vs
-        DEPENDS src/materials/separableGaussianBlur.fs
-        APPEND
-)
-
-add_custom_command(
-        OUTPUT "${MATERIAL_DIR}/separableGaussianBlur4L.filamat"
-        DEPENDS src/materials/separableGaussianBlur.vs
-        DEPENDS src/materials/separableGaussianBlur.fs
-        APPEND
-)
-
 add_custom_command(
         OUTPUT ${RESGEN_OUTPUTS}
         COMMAND resgen ${RESGEN_FLAGS} ${MATERIAL_BINS}

filament/backend/CMakeLists.txt

@@ -389,7 +389,6 @@ if (APPLE)
         test/test_MRT.cpp
         test/test_LoadImage.cpp
         test/test_RenderExternalImage.cpp
-        test/test_StencilBuffer.cpp
         )
 
     target_link_libraries(backend_test PRIVATE

filament/backend/include/backend/DriverEnums.h

@@ -719,7 +719,7 @@ enum class SamplerCompareMode : uint8_t {
     COMPARE_TO_TEXTURE = 1
 };
 
-//! comparison function for the depth / stencil sampler
+//! comparison function for the depth sampler
 enum class SamplerCompareFunc : uint8_t {
     // don't change the enums values
     LE = 0,     //!< Less or equal
@@ -728,8 +728,8 @@ enum class SamplerCompareFunc : uint8_t {
     G,          //!< Strictly greater than
     E,          //!< Equal
     NE,         //!< Not equal
-    A,          //!< Always. Depth / stencil testing is deactivated.
-    N           //!< Never. The depth / stencil test always fails.
+    A,          //!< Always. Depth testing is deactivated.
+    N           //!< Never. The depth test always fails.
 };
 
 //! Sampler paramters
@@ -785,18 +785,6 @@ enum class BlendFunction : uint8_t {
     SRC_ALPHA_SATURATE      //!< f(src, dst) = (1,1,1) * min(src.a, 1 - dst.a), 1
 };
 
-//! stencil operation
-enum class StencilOperation : uint8_t {
-    KEEP,                   //!< Keeps the current value.
-    ZERO,                   //!< Sets the value to 0.
-    REPLACE,                //!< Sets the value to the stencil reference value.
-    INCR,                   //!< Increments the current value. Clamps to the maximum representable unsigned value.
-    INCR_WRAP,              //!< Increments the current value. Wraps value to zero when incrementing the maximum representable unsigned value.
-    DECR,                   //!< Decrements the current value. Clamps to 0.
-    DECR_WRAP,              //!< Decrements the current value. Wraps value to the maximum representable unsigned value when decrementing a value of zero.
-    INVERT,                 //!< Bitwise inverts the current value.
-};
-
 //! Stream for external textures
 enum class StreamType {
     NATIVE,     //!< Not synchronized but copy-free. Good for video.
@@ -829,11 +817,9 @@ struct RasterState {
     using DepthFunc = backend::SamplerCompareFunc;
     using BlendEquation = backend::BlendEquation;
     using BlendFunction = backend::BlendFunction;
-    using StencilFunction = backend::SamplerCompareFunc;
-    using StencilOperation = backend::StencilOperation;
 
     RasterState() noexcept { // NOLINT
-        static_assert(sizeof(RasterState) == sizeof(uint64_t),
+        static_assert(sizeof(RasterState) == sizeof(uint32_t),
                 "RasterState size not what was intended");
         culling = CullingMode::BACK;
         blendEquationRGB = BlendEquation::ADD;
@@ -842,10 +828,6 @@ struct RasterState {
         blendFunctionSrcAlpha = BlendFunction::ONE;
         blendFunctionDstRGB = BlendFunction::ZERO;
         blendFunctionDstAlpha = BlendFunction::ZERO;
-        stencilFunc = StencilFunction::A;
-        stencilOpStencilFail = StencilOperation::KEEP;
-        stencilOpDepthFail = StencilOperation::KEEP;
-        stencilOpDepthStencilPass = StencilOperation::KEEP;
     }
 
     bool operator == (RasterState rhs) const noexcept { return u == rhs.u; }
@@ -874,56 +856,40 @@ struct RasterState {
     union {
         struct {
             //! culling mode
-            CullingMode culling                         : 2;        //  2
+            CullingMode culling                 : 2;        //  2
 
             //! blend equation for the red, green and blue components
-            BlendEquation blendEquationRGB              : 3;        //  5
+            BlendEquation blendEquationRGB      : 3;        //  5
             //! blend equation for the alpha component
-            BlendEquation blendEquationAlpha            : 3;        //  8
+            BlendEquation blendEquationAlpha    : 3;        //  8
 
             //! blending function for the source color
-            BlendFunction blendFunctionSrcRGB           : 4;        // 12
+            BlendFunction blendFunctionSrcRGB   : 4;        // 12
             //! blending function for the source alpha
-            BlendFunction blendFunctionSrcAlpha         : 4;        // 16
+            BlendFunction blendFunctionSrcAlpha : 4;        // 16
             //! blending function for the destination color
-            BlendFunction blendFunctionDstRGB           : 4;        // 20
+            BlendFunction blendFunctionDstRGB   : 4;        // 20
             //! blending function for the destination alpha
-            BlendFunction blendFunctionDstAlpha         : 4;        // 24
+            BlendFunction blendFunctionDstAlpha : 4;        // 24
 
             //! Whether depth-buffer writes are enabled
-            bool depthWrite                             : 1;        // 25
+            bool depthWrite                     : 1;        // 25
             //! Depth test function
-            DepthFunc depthFunc                         : 3;        // 28
+            DepthFunc depthFunc                 : 3;        // 28
 
             //! Whether color-buffer writes are enabled
-            bool colorWrite                             : 1;        // 29
+            bool colorWrite                     : 1;        // 29
 
             //! use alpha-channel as coverage mask for anti-aliasing
-            bool alphaToCoverage                        : 1;        // 30
+            bool alphaToCoverage                : 1;        // 30
 
             //! whether front face winding direction must be inverted
-            bool inverseFrontFaces                      : 1;        // 31
+            bool inverseFrontFaces              : 1;        // 31
 
-            //! Whether stencil-buffer writes are enabled
-            bool stencilWrite                           : 1;        // 32
-            //! Stencil reference value
-            uint8_t stencilRef                          : 8;        // 40
-            //! Stencil test function
-            StencilFunction stencilFunc                 : 3;        // 43
-            //! Stencil operation when stencil test fails
-            StencilOperation stencilOpStencilFail       : 3;        // 46
             //! padding, must be 0
-            uint8_t padding0                            : 2;        // 48
-            //! Stencil operation when stencil test passes but depth test fails
-            StencilOperation stencilOpDepthFail         : 3;        // 51
-            //! Stencil operation when both stencil and depth test pass
-            StencilOperation stencilOpDepthStencilPass  : 3;        // 54
-            //! padding, must be 0
-            uint8_t padding1                            : 2;        // 56
-            //! padding, must be 0
-            uint8_t padding2                            : 8;        // 64
+            uint8_t padding                     : 1;        // 32
         };
-        uint64_t u = 0;
+        uint32_t u = 0;
     };
 };
 

filament/backend/include/backend/Handle.h

@@ -26,7 +26,6 @@
 #include <stdint.h>
 
 #include <limits>
-#include <type_traits>
 
 namespace filament::backend {
 

filament/backend/include/private/backend/CommandStream.h

@@ -34,7 +34,6 @@
 #include <utils/compiler.h>
 #include <utils/ThreadUtils.h>
 
-#include <cstddef>
 #include <functional>
 #include <tuple>
 #include <thread>

filament/backend/include/private/backend/Program.h

@@ -67,8 +67,6 @@ public:
             utils::Invocable<utils::io::ostream&(utils::io::ostream& out)>&& logger);
 
     // sets one of the program's shader (e.g. vertex, fragment)
-    // string-based shaders are null terminated, consequently the size parameter must include the
-    // null terminating character.
     Program& shader(Shader shader, void const* data, size_t size) noexcept;
 
     // sets the 'bindingPoint' uniform block's name for this program.
@@ -86,14 +84,10 @@ public:
     Program& setSamplerGroup(size_t bindingPoint, ShaderStageFlags stageFlags,
             Sampler const* samplers, size_t count) noexcept;
 
-    // string-based shaders are null terminated, consequently the size parameter must include the
-    // null terminating character.
     Program& withVertexShader(void const* data, size_t size) {
         return shader(Shader::VERTEX, data, size);
     }
 
-    // string-based shaders are null terminated, consequently the size parameter must include the
-    // null terminating character.
     Program& withFragmentShader(void const* data, size_t size) {
         return shader(Shader::FRAGMENT, data, size);
     }

filament/backend/src/metal/MetalHandles.mm

@@ -346,11 +346,8 @@ MetalProgram::MetalProgram(id<MTLDevice> device, const Program& program) noexcep
             continue;
         }
 
-        assert_invariant( source[source.size() - 1] == '\0' );
-
-        // the shader string is null terminated and the length includes the null character
         NSString* objcSource = [[NSString alloc] initWithBytes:source.data()
-                                                        length:source.size() - 1
+                                                        length:source.size()
                                                       encoding:NSUTF8StringEncoding];
         NSError* error = nil;
         // When options is nil, Metal uses the most recent language version available.

filament/backend/src/opengl/GLUtils.h

@@ -255,23 +255,6 @@ constexpr inline GLenum getDepthFunc(SamplerCompareFunc func) noexcept {
     return getTextureCompareFunc(func);
 }
 
-constexpr inline GLenum getStencilFunc(SamplerCompareFunc func) noexcept {
-    return getTextureCompareFunc(func);
-}
-
-constexpr inline GLenum getStencilOp(StencilOperation op) noexcept {
-    switch (op) {
-        case StencilOperation::KEEP:        return GL_KEEP;
-        case StencilOperation::ZERO:        return GL_ZERO;
-        case StencilOperation::REPLACE:     return GL_REPLACE;
-        case StencilOperation::INCR:        return GL_INCR;
-        case StencilOperation::INCR_WRAP:   return GL_INCR_WRAP;
-        case StencilOperation::DECR:        return GL_DECR;
-        case StencilOperation::DECR_WRAP:   return GL_DECR_WRAP;
-        case StencilOperation::INVERT:      return GL_INVERT;
-    }
-}
-
 constexpr inline GLenum getFormat(PixelDataFormat format) noexcept {
     using PixelDataFormat = PixelDataFormat;
     switch (format) {

filament/backend/src/opengl/OpenGLContext.h

@@ -101,9 +101,6 @@ public:
     inline void colorMask(GLboolean flag) noexcept;
     inline void depthMask(GLboolean flag) noexcept;
     inline void depthFunc(GLenum func) noexcept;
-    inline void stencilFunc(GLenum func, GLint ref, GLuint mask) noexcept;
-    inline void stencilOp(GLenum sfail, GLenum dpfail, GLenum dppass) noexcept;
-    inline void stencilMask(GLuint mask) noexcept;
     inline void polygonOffset(GLfloat factor, GLfloat units) noexcept;
     inline void beginQuery(GLenum target, GLuint query) noexcept;
     inline void endQuery(GLenum target) noexcept;
@@ -292,23 +289,6 @@ private:
             GLboolean colorMask         = GL_TRUE;
             GLboolean depthMask         = GL_TRUE;
             GLenum depthFunc            = GL_LESS;
-            struct StencilFunc {
-                GLenum func             = GL_ALWAYS;
-                GLint ref               = 0;
-                GLuint mask             = ~GLuint(0);
-                bool operator != (StencilFunc const& rhs) const noexcept {
-                    return func != rhs.func || ref != rhs.ref || mask != rhs.mask;
-                }
-            } stencilFunc;
-            struct StencilOp {
-                GLenum sfail            = GL_KEEP;
-                GLenum dpfail           = GL_KEEP;
-                GLenum dppass           = GL_KEEP;
-                bool operator != (StencilOp const& rhs) const noexcept {
-                    return sfail != rhs.sfail || dpfail != rhs.dpfail || dppass != rhs.dppass;
-                }
-            } stencilOp;
-            GLuint stencilMask          = ~GLuint(0);
         } raster;
 
         struct PolygonOffset {
@@ -654,27 +634,6 @@ void OpenGLContext::depthFunc(GLenum func) noexcept {
     });
 }
 
-void OpenGLContext::stencilFunc(GLenum func, GLint ref, GLuint mask) noexcept {
-    // WARNING: don't call this without updating mRasterState
-    update_state(state.raster.stencilFunc, {func, ref, mask}, [&]() {
-        glStencilFunc(func, ref, mask);
-    });
-}
-
-void OpenGLContext::stencilOp(GLenum sfail, GLenum dpfail, GLenum dppass) noexcept {
-    // WARNING: don't call this without updating mRasterState
-    update_state(state.raster.stencilOp, {sfail, dpfail, dppass}, [&]() {
-        glStencilOp(sfail, dpfail, dppass);
-    });
-}
-
-void OpenGLContext::stencilMask(GLuint mask) noexcept {
-    // WARNING: don't call this without updating mRasterState
-    update_state(state.raster.stencilMask, mask, [&]() {
-        glStencilMask(mask);
-    });
-}
-
 void OpenGLContext::polygonOffset(GLfloat factor, GLfloat units) noexcept {
     update_state(state.polygonOffset, { factor, units }, [&]() {
         if (factor != 0 || units != 0) {

filament/backend/src/opengl/OpenGLDriver.cpp

@@ -294,19 +294,6 @@ void OpenGLDriver::setRasterStateSlow(RasterState rs) noexcept {
         gl.depthMask(GLboolean(rs.depthWrite));
     }
 
-    // stencil test / operation
-    if (rs.stencilFunc == RasterState::StencilFunction::A && !rs.stencilWrite) {
-        gl.disable(GL_STENCIL_TEST);
-    } else {
-        gl.enable(GL_STENCIL_TEST);
-        gl.stencilFunc(getStencilFunc(rs.stencilFunc), rs.stencilRef, ~GLuint(0));
-        gl.stencilOp(getStencilOp(rs.stencilOpStencilFail),
-                getStencilOp(rs.stencilOpDepthFail),
-                getStencilOp(rs.stencilOpDepthStencilPass));
-        GLuint stencilMask = rs.stencilWrite ? ~GLuint(0) : 0x00;
-        gl.stencilMask(stencilMask);
-    }
-
     // write masks
     gl.colorMask(GLboolean(rs.colorWrite));
 

filament/backend/src/opengl/OpenGLProgram.cpp

@@ -186,11 +186,7 @@ highp uint packHalf2x16(vec2 v) {
             GLuint shaderId = glCreateShader(glShaderType);
             { // scope for source/length (we don't want them to leak out)
                 const char* const source = shaderView.data();
-
-                // the shader string is null terminated and the length includes the null character
-                const GLint length = (GLint)shaderView.length() - 1;
-                assert_invariant( source[length] == '\0' );
-
+                const GLint length = (GLint)shaderView.length();
                 glShaderSource(shaderId, 1, &source, &length);
                 glCompileShader(shaderId);
 #ifndef NDEBUG

filament/backend/src/vulkan/VulkanDriver.cpp

@@ -1633,21 +1633,19 @@ void VulkanDriver::blit(TargetBufferFlags buffers, Handle<HwRenderTarget> dst, V
 
     VkFilter vkfilter = filter == SamplerMagFilter::NEAREST ? VK_FILTER_NEAREST : VK_FILTER_LINEAR;
 
-    // The Y inversion below makes it so that Vk matches GL and Metal.
-
     const VkExtent2D srcExtent = srcTarget->getExtent();
-    const int32_t srcLeft = srcRect.left;
-    const int32_t srcTop = srcExtent.height - srcRect.bottom - srcRect.height;
-    const int32_t srcRight = srcRect.left + srcRect.width;
-    const int32_t srcBottom = srcTop + srcRect.height;
-    const VkOffset3D srcOffsets[2] = { { srcLeft, srcTop, 0 }, { srcRight, srcBottom, 1 }};
+    const int32_t srcLeft = std::min(srcRect.left, (int32_t) srcExtent.width);
+    const int32_t srcBottom = std::min(srcRect.bottom, (int32_t) srcExtent.height);
+    const int32_t srcRight = std::min(srcRect.left + srcRect.width, srcExtent.width);
+    const int32_t srcTop = std::min(srcRect.bottom + srcRect.height, srcExtent.height);
+    const VkOffset3D srcOffsets[2] = { { srcLeft, srcBottom, 0 }, { srcRight, srcTop, 1 }};
 
     const VkExtent2D dstExtent = dstTarget->getExtent();
-    const int32_t dstLeft = dstRect.left;
-    const int32_t dstTop = dstExtent.height - dstRect.bottom - dstRect.height;
-    const int32_t dstRight = dstRect.left + dstRect.width;
-    const int32_t dstBottom = dstTop + dstRect.height;
-    const VkOffset3D dstOffsets[2] = { { dstLeft, dstTop, 0 }, { dstRight, dstBottom, 1 }};
+    const int32_t dstLeft = std::min(dstRect.left, (int32_t) dstExtent.width);
+    const int32_t dstBottom = std::min(dstRect.bottom, (int32_t) dstExtent.height);
+    const int32_t dstRight = std::min(dstRect.left + dstRect.width, dstExtent.width);
+    const int32_t dstTop = std::min(dstRect.bottom + dstRect.height, dstExtent.height);
+    const VkOffset3D dstOffsets[2] = { { dstLeft, dstBottom, 0 }, { dstRight, dstTop, 1 }};
 
     if (any(buffers & TargetBufferFlags::DEPTH) && srcTarget->hasDepth() && dstTarget->hasDepth()) {
         mBlitter.blitDepth({dstTarget, dstOffsets, srcTarget, srcOffsets});

filament/backend/src/vulkan/VulkanUtility.cpp

@@ -546,7 +546,6 @@ void transitionImageLayout(VkCommandBuffer cmdbuffer, VulkanLayoutTransition tra
     if (transition.oldLayout == transition.newLayout) {
         return;
     }
-    assert_invariant(transition.image != VK_NULL_HANDLE && "Please call bindToSwapChain.");
     VkImageMemoryBarrier barrier = {};
     barrier.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER;
     barrier.oldLayout = transition.oldLayout;

filament/backend/test/BackendTest.cpp

@@ -20,26 +20,17 @@
 
 #include <utils/Hash.h>
 
-#include <fstream>
-
 #include "ShaderGenerator.h"
 #include "TrianglePrimitive.h"
 
 static constexpr size_t CONFIG_MIN_COMMAND_BUFFERS_SIZE = 1 * 1024 * 1024;
 static constexpr size_t CONFIG_COMMAND_BUFFERS_SIZE     = 3 * CONFIG_MIN_COMMAND_BUFFERS_SIZE;
 
+namespace test {
+
 using namespace filament;
 using namespace filament::backend;
 
-#ifndef IOS
-#include <imageio/ImageEncoder.h>
-#include <image/ColorTransform.h>
-
-using namespace image;
-#endif
-
-namespace test {
-
 Backend BackendTest::sBackend = Backend::NOOP;
 bool BackendTest::sIsMobilePlatform = false;
 
@@ -140,38 +131,20 @@ void BackendTest::renderTriangle(Handle<HwRenderTarget> renderTarget,
 }
 
 void BackendTest::readPixelsAndAssertHash(const char* testName, size_t width, size_t height,
-        Handle<HwRenderTarget> rt, uint32_t expectedHash, bool exportScreenshot) {
+        Handle<HwRenderTarget> rt, uint32_t expectedHash) {
     void* buffer = calloc(1, width * height * 4);
 
     struct Capture {
         uint32_t expectedHash;
         char* name;
-        bool exportScreenshot;
-        size_t width, height;
     };
-    auto* c = new Capture();
+    Capture* c = new Capture();
     c->expectedHash = expectedHash;
     c->name = strdup(testName);
-    c->exportScreenshot = exportScreenshot;
-    c->width = width;
-    c->height = height;
 
     PixelBufferDescriptor pbd(buffer, width * height * 4, PixelDataFormat::RGBA, PixelDataType::UBYTE,
             1, 0, 0, width, [](void* buffer, size_t size, void* user) {
-                auto* c = (Capture*)user;
-
-                // Export a screenshot, if requested.
-                if (c->exportScreenshot) {
-#ifndef IOS
-                    LinearImage image(c->width, c->height, 4);
-                    image = toLinearWithAlpha<uint8_t>(c->width, c->height, c->width * 4,
-                            (uint8_t*) buffer);
-                    const std::string png = std::string(c->name) + ".png";
-                    std::ofstream outputStream(png.c_str(), std::ios::binary | std::ios::trunc);
-                    ImageEncoder::encode(outputStream, ImageEncoder::Format::PNG, image, "",
-                            png);
-#endif
-                }
+                Capture* c = (Capture*)user;
 
                 // Hash the contents of the buffer and check that they match.
                 uint32_t hash = utils::hash::murmur3((const uint32_t*) buffer, size / 4, 0);

filament/backend/test/BackendTest.h

@@ -56,8 +56,7 @@ protected:
             filament::backend::Handle<filament::backend::HwProgram> program);
 
     void readPixelsAndAssertHash(const char* testName, size_t width, size_t height,
-            filament::backend::Handle<filament::backend::HwRenderTarget> rt, uint32_t expectedHash,
-            bool exportScreenshot = false);
+            filament::backend::Handle<filament::backend::HwRenderTarget> rt, uint32_t expectedHash);
 
     filament::backend::DriverApi& getDriverApi() { return *commandStream; }
     filament::backend::Driver& getDriver() { return *driver; }

filament/backend/test/test_Blit.cpp

@@ -866,157 +866,4 @@ TEST_F(BackendTest, Blit2DTextureArray) {
     api.stopCapture(0);
 }
 
-TEST_F(BackendTest, BlitRegion) {
-    auto& api = getDriverApi();
-
-    constexpr int kSrcTexWidth = 1024;
-    constexpr int kSrcTexHeight = 1024;
-    constexpr auto kSrcTexFormat = TextureFormat::RGBA8;
-    constexpr int kDstTexWidth = 384;
-    constexpr int kDstTexHeight = 384;
-    constexpr auto kDstTexFormat = TextureFormat::RGBA8;
-    constexpr int kNumLevels = 3;
-
-    // Create a SwapChain and make it current. We don't really use it so the res doesn't matter.
-    auto swapChain = api.createSwapChainHeadless(256, 256, 0);
-    api.makeCurrent(swapChain, swapChain);
-
-    // Create a source texture.
-    Handle<HwTexture> srcTexture = api.createTexture(
-        SamplerType::SAMPLER_2D, kNumLevels, kSrcTexFormat, 1, kSrcTexWidth, kSrcTexHeight, 1,
-        TextureUsage::SAMPLEABLE | TextureUsage::UPLOADABLE | TextureUsage::COLOR_ATTACHMENT);
-    const bool flipY = sBackend == Backend::OPENGL;
-    createBitmap(api, srcTexture, kSrcTexWidth, kSrcTexHeight, 0, float3(0.5, 0, 0), flipY);
-    createBitmap(api, srcTexture, kSrcTexWidth, kSrcTexHeight, 1, float3(0, 0, 0.5), flipY);
-
-    // Create a destination texture.
-    Handle<HwTexture> dstTexture = api.createTexture(
-        SamplerType::SAMPLER_2D, kNumLevels, kDstTexFormat, 1, kDstTexWidth, kDstTexHeight, 1,
-        TextureUsage::SAMPLEABLE | TextureUsage::COLOR_ATTACHMENT);
-
-    // Create a RenderTarget for each texture's miplevel.
-    Handle<HwRenderTarget> srcRenderTargets[kNumLevels];
-    Handle<HwRenderTarget> dstRenderTargets[kNumLevels];
-    for (uint8_t level = 0; level < kNumLevels; level++) {
-        srcRenderTargets[level] = api.createRenderTarget( TargetBufferFlags::COLOR,
-                kSrcTexWidth >> level, kSrcTexHeight >> level, 1, { srcTexture, level, 0 }, {}, {});
-        dstRenderTargets[level] = api.createRenderTarget( TargetBufferFlags::COLOR,
-                kDstTexWidth >> level, kDstTexHeight >> level, 1, { dstTexture, level, 0 }, {}, {});
-    }
-
-    // Blit one-quarter of src level 1 to dst level 0.
-    const int srcLevel = 1;
-    Viewport srcRect = {
-        .left = (kSrcTexWidth >> srcLevel) / 2,
-        .bottom = (kSrcTexHeight >> srcLevel) / 2,
-        .width = (kSrcTexWidth >> srcLevel) / 2,
-        .height = (kSrcTexHeight >> srcLevel) / 2,
-    };
-    Viewport dstRect = {
-        .left = 10,
-        .bottom = 10,
-        .width = kDstTexWidth - 10,
-        .height = kDstTexHeight - 10,
-    };
-
-    api.blit(TargetBufferFlags::COLOR0, dstRenderTargets[0],
-            dstRect, srcRenderTargets[srcLevel],
-            srcRect, SamplerMagFilter::LINEAR);
-
-    // Push through an empty frame to allow the texture to upload and the blit to execute.
-    api.beginFrame(0, 0);
-    api.commit(swapChain);
-    api.endFrame(0);
-
-    // Grab a screenshot.
-    ScreenshotParams params { kDstTexWidth, kDstTexHeight, "BlitRegion.png" };
-    api.beginFrame(0, 0);
-    dumpScreenshot(api, dstRenderTargets[0], &params);
-    api.commit(swapChain);
-    api.endFrame(0);
-
-    // Wait for the ReadPixels result to come back.
-    api.finish();
-    executeCommands();
-    getDriver().purge();
-
-    // Cleanup.
-    api.destroyTexture(srcTexture);
-    api.destroyTexture(dstTexture);
-    api.destroySwapChain(swapChain);
-    for (auto rt : srcRenderTargets)  api.destroyRenderTarget(rt);
-    for (auto rt : dstRenderTargets)  api.destroyRenderTarget(rt);
-    executeCommands();
-}
-
-TEST_F(BackendTest, BlitRegionToSwapChain) {
-    auto& api = getDriverApi();
-
-    constexpr int kSrcTexWidth = 1024;
-    constexpr int kSrcTexHeight = 1024;
-    constexpr auto kSrcTexFormat = TextureFormat::RGBA8;
-    constexpr int kDstTexWidth = 512;
-    constexpr int kDstTexHeight = 512;
-    constexpr int kNumLevels = 3;
-
-    // Create a SwapChain and make it current.
-    auto swapChain = createSwapChain();
-    Handle<HwRenderTarget> dstRenderTarget = api.createDefaultRenderTarget();
-    api.makeCurrent(swapChain, swapChain);
-
-    // Create a source texture.
-    Handle<HwTexture> srcTexture = api.createTexture(
-        SamplerType::SAMPLER_2D, kNumLevels, kSrcTexFormat, 1, kSrcTexWidth, kSrcTexHeight, 1,
-        TextureUsage::SAMPLEABLE | TextureUsage::UPLOADABLE | TextureUsage::COLOR_ATTACHMENT);
-    const bool flipY = sBackend == Backend::OPENGL;
-    createBitmap(api, srcTexture, kSrcTexWidth, kSrcTexHeight, 0, float3(0.5, 0, 0), flipY);
-    createBitmap(api, srcTexture, kSrcTexWidth, kSrcTexHeight, 1, float3(0, 0, 0.5), flipY);
-
-    // Create a RenderTarget for each texture's miplevel.
-    Handle<HwRenderTarget> srcRenderTargets[kNumLevels];
-    for (uint8_t level = 0; level < kNumLevels; level++) {
-        srcRenderTargets[level] = api.createRenderTarget( TargetBufferFlags::COLOR,
-                kSrcTexWidth >> level, kSrcTexHeight >> level, 1, { srcTexture, level, 0 }, {}, {});
-    }
-
-    // Blit one-quarter of src level 1 to dst level 0.
-    const int srcLevel = 1;
-    Viewport srcRect = {
-        .left = (kSrcTexWidth >> srcLevel) / 2,
-        .bottom = (kSrcTexHeight >> srcLevel) / 2,
-        .width = (kSrcTexWidth >> srcLevel) / 2,
-        .height = (kSrcTexHeight >> srcLevel) / 2,
-    };
-    Viewport dstRect = {
-        .left = 10,
-        .bottom = 10,
-        .width = kDstTexWidth - 10,
-        .height = kDstTexHeight - 10,
-    };
-
-    api.beginFrame(0, 0);
-
-    api.blit(TargetBufferFlags::COLOR0, dstRenderTarget,
-            dstRect, srcRenderTargets[srcLevel],
-            srcRect, SamplerMagFilter::LINEAR);
-
-    ScreenshotParams params { kDstTexWidth, kDstTexHeight, "BlitRegionToSwapChain.png" };
-    dumpScreenshot(api, dstRenderTarget, &params);
-
-    api.commit(swapChain);
-
-    api.endFrame(0);
-
-    // Wait for the ReadPixels result to come back.
-    api.finish();
-    executeCommands();
-    getDriver().purge();
-
-    // Cleanup.
-    api.destroyTexture(srcTexture);
-    api.destroySwapChain(swapChain);
-    for (auto rt : srcRenderTargets)  api.destroyRenderTarget(rt);
-    executeCommands();
-}
-
 } // namespace test

filament/backend/test/test_StencilBuffer.cpp

@@ -1,150 +0,0 @@
-/*
- * 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 "BackendTest.h"
-
-#include "ShaderGenerator.h"
-#include "TrianglePrimitive.h"
-
-using namespace filament;
-using namespace filament::backend;
-
-namespace {
-
-////////////////////////////////////////////////////////////////////////////////////////////////////
-// Shaders
-////////////////////////////////////////////////////////////////////////////////////////////////////
-
-std::string vertex (R"(#version 450 core
-
-layout(location = 0) in vec4 mesh_position;
-
-void main() {
-    gl_Position = vec4(mesh_position.xy, 0.0, 1.0);
-#if defined(TARGET_VULKAN_ENVIRONMENT)
-    // In Vulkan, clip space is Y-down. In OpenGL and Metal, clip space is Y-up.
-    gl_Position.y = -gl_Position.y;
-#endif
-}
-)");
-
-std::string fragment (R"(#version 450 core
-
-layout(location = 0) out vec4 fragColor;
-
-void main() {
-    fragColor = vec4(1.0);
-}
-
-)");
-
-}
-
-namespace test {
-
-// 1. Clear the stencil buffer to all zeroes.
-// 2. Render a small triangle only to the stencil buffer and set the operation to write ones.
-// 3. Enable a "== 0" stencil test and enable color writes.
-// 4. Render a larger triangle to the SwapChain.
-// 5. The larger triangle should have a "hole" in the middle.
-TEST_F(BackendTest, StencilBuffer) {
-    auto& api = getDriverApi();
-
-    // The test is executed within this block scope to force destructors to run before
-    // executeCommands().
-    {
-        // Create a platform-specific SwapChain and make it current.
-        auto swapChain = createSwapChain();
-        api.makeCurrent(swapChain, swapChain);
-
-        // Create a program.
-        ShaderGenerator shaderGen(vertex, fragment, sBackend, sIsMobilePlatform);
-        Program p = shaderGen.getProgram();
-        ProgramHandle program = api.createProgram(std::move(p));
-
-        // We'll be using a triangle as geometry.
-        TrianglePrimitive smallTriangle(api);
-        static filament::math::float2 vertices[3] = {
-                { -0.5, -0.5 },
-                {  0.5, -0.5 },
-                { -0.5,  0.5 }
-        };
-        smallTriangle.updateVertices(vertices);
-        TrianglePrimitive triangle(api);
-
-        // Create two textures: a color and a stencil, and an associated RenderTarget.
-        auto colorTexture = getDriverApi().createTexture(SamplerType::SAMPLER_2D, 1,
-                TextureFormat::RGBA8, 1, 512, 512, 1, TextureUsage::COLOR_ATTACHMENT);
-        auto stencilTexture = getDriverApi().createTexture(SamplerType::SAMPLER_2D, 1,
-                TextureFormat::STENCIL8, 1, 512, 512, 1, TextureUsage::STENCIL_ATTACHMENT);
-        auto renderTarget = getDriverApi().createRenderTarget(
-                TargetBufferFlags::COLOR0 | TargetBufferFlags::STENCIL, 512, 512, 1,
-                {{colorTexture}}, {}, {{stencilTexture}});
-
-        // Step 1: Clear the stencil buffer to all zeroes and the color buffer to blue.
-        // Render a small triangle only to the stencil buffer, increasing the stencil buffer to 1.
-        RenderPassParams params = {};
-        params.flags.clear = TargetBufferFlags::COLOR0 | TargetBufferFlags::STENCIL;
-        params.viewport = {0, 0, 512, 512};
-        params.clearColor = math::float4(0.0f, 0.0f, 1.0f, 1.0f);
-        params.clearStencil = 0u;
-        params.flags.discardStart = TargetBufferFlags::ALL;
-        params.flags.discardEnd = TargetBufferFlags::NONE;
-
-        PipelineState ps = {};
-        ps.program = program;
-        ps.rasterState.colorWrite = false;
-        ps.rasterState.depthWrite = false;
-        ps.rasterState.stencilWrite = true;
-        ps.rasterState.stencilOpDepthStencilPass = StencilOperation::INCR;
-
-        api.makeCurrent(swapChain, swapChain);
-        api.beginFrame(0, 0);
-
-        api.beginRenderPass(renderTarget, params);
-        api.draw(ps, smallTriangle.getRenderPrimitive(), 1);
-        api.endRenderPass();
-
-        // Step 2: Render a larger triangle with the stencil test enabled.
-        params.flags.clear = TargetBufferFlags::NONE;
-        params.flags.discardStart = TargetBufferFlags::NONE;
-        ps.rasterState.colorWrite = true;
-        ps.rasterState.stencilWrite = false;
-        ps.rasterState.stencilOpDepthStencilPass = StencilOperation::KEEP;
-        ps.rasterState.stencilFunc = RasterState::StencilFunction::E;
-        ps.rasterState.stencilRef = 0u;
-
-        api.beginRenderPass(renderTarget, params);
-        api.draw(ps, triangle.getRenderPrimitive(), 1);
-        api.endRenderPass();
-
-        readPixelsAndAssertHash("StencilBuffer", 512, 512, renderTarget, 0x3B1AEF0F, true);
-
-        api.commit(swapChain);
-        api.endFrame(0);
-
-        api.destroyProgram(program);
-        api.destroySwapChain(swapChain);
-        api.destroyTexture(colorTexture);
-        api.destroyTexture(stencilTexture);
-        api.destroyRenderTarget(renderTarget);
-    }
-
-    flushAndWait();
-    getDriver().purge();
-}
-
-} // namespace test

filament/include/filament/MorphTargetBuffer.h

@@ -78,49 +78,43 @@ public:
     };
 
     /**
-     * Updates positions for the given morph target.
-     *
-     * This is equivalent to the float4 method, but uses 1.0 for the 4th component.
+     * Updates the position of morph target at the index.
      *
      * Both positions and tangents must be provided.
      *
      * @param engine Reference to the filament::Engine associated with this MorphTargetBuffer.
      * @param targetIndex the index of morph target to be updated.
-     * @param positions pointer to at least "count" positions
-     * @param count number of float3 vectors in positions
-     * @param offset offset into the target buffer, expressed as a number of float4 vectors
+     * @param weights pointer to at least count positions
+     * @param count number of position elements in positions
      * @see setTangentsAt
      */
     void setPositionsAt(Engine& engine, size_t targetIndex,
             math::float3 const* positions, size_t count, size_t offset = 0);
 
     /**
-     * Updates positions for the given morph target.
+     * Updates the position of morph target at the index.
      *
      * Both positions and tangents must be provided.
      *
      * @param engine Reference to the filament::Engine associated with this MorphTargetBuffer.
      * @param targetIndex the index of morph target to be updated.
-     * @param positions pointer to at least "count" positions
-     * @param count number of float4 vectors in positions
-     * @param offset offset into the target buffer, expressed as a number of float4 vectors
-     * @see setTangentsAt
+     * @param weights pointer to at least count positions
+     * @param count number of position elements in positions
+     * @see setPositionsAt
      */
     void setPositionsAt(Engine& engine, size_t targetIndex,
             math::float4 const* positions, size_t count, size_t offset = 0);
 
     /**
-     * Updates tangents for the given morph target.
+     * Updates the position of morph target at the index.
      *
-     * These quaternions must be represented as signed shorts, where real numbers in the [-1,+1]
-     * range multiplied by 32767.
+     * Both positions and tangents must be provided.
      *
      * @param engine Reference to the filament::Engine associated with this MorphTargetBuffer.
      * @param targetIndex the index of morph target to be updated.
-     * @param tangents pointer to at least "count" tangents
-     * @param count number of short4 quaternions in tangents
-     * @param offset offset into the target buffer, expressed as a number of short4 vectors
-     * @see setPositionsAt
+     * @param tangents pointer to at least count tangents
+     * @param count number of tangent elements in tangents
+     * @see setTangentsAt
      */
     void setTangentsAt(Engine& engine, size_t targetIndex,
             math::short4 const* tangents, size_t count, size_t offset = 0);

filament/include/filament/Options.h

@@ -25,9 +25,6 @@ namespace filament {
 
 class Texture;
 
-/**
- * Generic quality level.
- */
 enum class QualityLevel : uint8_t {
     LOW,
     MEDIUM,
@@ -72,8 +69,8 @@ enum class BlendMode : uint8_t {
  *
  */
 struct DynamicResolutionOptions {
-    math::float2 minScale = {0.5f, 0.5f};           //!< minimum scale factors in x and y %codegen_java_float%
-    math::float2 maxScale = {1.0f, 1.0f};           //!< maximum scale factors in x and y %codegen_java_float%
+    math::float2 minScale = math::float2(0.5f);     //!< minimum scale factors in x and y
+    math::float2 maxScale = math::float2(1.0f);     //!< maximum scale factors in x and y
     float sharpness = 0.9f;                         //!< sharpness when QualityLevel::MEDIUM or higher is used [0 (disabled), 1 (sharpest)]
     bool enabled = false;                           //!< enable or disable dynamic resolution
     bool homogeneousScaling = false;                //!< set to true to force homogeneous scaling
@@ -128,8 +125,8 @@ struct BloomOptions {
         ADD,           //!< Bloom is modulated by the strength parameter and added to the scene
         INTERPOLATE    //!< Bloom is interpolated with the scene using the strength parameter
     };
-    Texture* dirt = nullptr;                //!< user provided dirt texture %codegen_skip_json% %codegen_skip_javascript%
-    float dirtStrength = 0.2f;              //!< strength of the dirt texture %codegen_skip_json% %codegen_skip_javascript%
+    Texture* dirt = nullptr;                //!< user provided dirt texture
+    float dirtStrength = 0.2f;              //!< strength of the dirt texture
     float strength = 0.10f;                 //!< bloom's strength between 0.0 and 1.0
     uint32_t resolution = 360;              //!< resolution of vertical axis (2^levels to 2048)
     float anamorphism = 1.0f;               //!< bloom x/y aspect-ratio (1/32 to 32)
@@ -154,16 +151,16 @@ struct BloomOptions {
  * Options to control fog in the scene
  */
 struct FogOptions {
-    float distance = 0.0f;                 //!< distance in world units from the camera where the fog starts ( >= 0.0 )
-    float maximumOpacity = 1.0f;           //!< fog's maximum opacity between 0 and 1
-    float height = 0.0f;                   //!< fog's floor in world units
-    float heightFalloff = 1.0f;            //!< how fast fog dissipates with altitude
-    LinearColor color = {0.5f, 0.5f, 0.5f};//!< fog's color (linear), see fogColorFromIbl
-    float density = 0.1f;                  //!< fog's density at altitude given by 'height'
-    float inScatteringStart = 0.0f;        //!< distance in world units from the camera where in-scattering starts
-    float inScatteringSize = -1.0f;        //!< size of in-scattering (>0 to activate). Good values are >> 1 (e.g. ~10 - 100).
-    bool fogColorFromIbl = false;          //!< Fog color will be modulated by the IBL color in the view direction.
-    bool enabled = false;                  //!< enable or disable fog
+    float distance = 0.0f;              //!< distance in world units from the camera where the fog starts ( >= 0.0 )
+    float maximumOpacity = 1.0f;        //!< fog's maximum opacity between 0 and 1
+    float height = 0.0f;                //!< fog's floor in world units
+    float heightFalloff = 1.0f;         //!< how fast fog dissipates with altitude
+    LinearColor color{0.5f};            //!< fog's color (linear), see fogColorFromIbl
+    float density = 0.1f;               //!< fog's density at altitude given by 'height'
+    float inScatteringStart = 0.0f;     //!< distance in world units from the camera where in-scattering starts
+    float inScatteringSize = -1.0f;     //!< size of in-scattering (>0 to activate). Good values are >> 1 (e.g. ~10 - 100).
+    bool fogColorFromIbl = false;       //!< Fog color will be modulated by the IBL color in the view direction.
+    bool enabled = false;               //!< enable or disable fog
 };
 
 /**
@@ -177,9 +174,8 @@ struct FogOptions {
  */
 struct DepthOfFieldOptions {
     enum class Filter : uint8_t {
-        NONE,
-        UNUSED,
-        MEDIAN
+        NONE = 0,
+        MEDIAN = 2
     };
     float cocScale = 1.0f;              //!< circle of confusion scale factor (amount of blur)
     float maxApertureDiameter = 0.01f;  //!< maximum aperture diameter in meters (zero to disable rotation)
@@ -231,7 +227,7 @@ struct VignetteOptions {
     float midPoint = 0.5f;                      //!< high values restrict the vignette closer to the corners, between 0 and 1
     float roundness = 0.5f;                     //!< controls the shape of the vignette, from a rounded rectangle (0.0), to an oval (0.5), to a circle (1.0)
     float feather = 0.5f;                       //!< softening amount of the vignette effect, between 0 and 1
-    LinearColorA color = {0.0f, 0.0f, 0.0f, 1.0f}; //!< color of the vignette effect, alpha is currently ignored
+    LinearColorA color{0.0f, 0.0f, 0.0f, 1.0f}; //!< color of the vignette effect, alpha is currently ignored
     bool enabled = false;                       //!< enables or disables the vignette effect
 };
 
@@ -275,18 +271,17 @@ struct AmbientOcclusionOptions {
      * Ambient shadows from dominant light
      */
     struct Ssct {
-        float lightConeRad = 1.0f;       //!< full cone angle in radian, between 0 and pi/2
-        float shadowDistance = 0.3f;     //!< how far shadows can be cast
-        float contactDistanceMax = 1.0f; //!< max distance for contact
-        float intensity = 0.8f;          //!< intensity
-        math::float3 lightDirection = { 0, -1, 0 };    //!< light direction
-        float depthBias = 0.01f;         //!< depth bias in world units (mitigate self shadowing)
-        float depthSlopeBias = 0.01f;    //!< depth slope bias (mitigate self shadowing)
-        uint8_t sampleCount = 4;         //!< tracing sample count, between 1 and 255
-        uint8_t rayCount = 1;            //!< # of rays to trace, between 1 and 255
-        bool enabled = false;            //!< enables or disables SSCT
-    };
-    Ssct ssct;                           // %codegen_skip_javascript% %codegen_java_flatten%
+        float lightConeRad = 1.0f;          //!< full cone angle in radian, between 0 and pi/2
+        float shadowDistance = 0.3f;        //!< how far shadows can be cast
+        float contactDistanceMax = 1.0f;    //!< max distance for contact
+        float intensity = 0.8f;             //!< intensity
+        math::float3 lightDirection{ 0, -1, 0 };    //!< light direction
+        float depthBias = 0.01f;        //!< depth bias in world units (mitigate self shadowing)
+        float depthSlopeBias = 0.01f;   //!< depth slope bias (mitigate self shadowing)
+        uint8_t sampleCount = 4;        //!< tracing sample count, between 1 and 255
+        uint8_t rayCount = 1;           //!< # of rays to trace, between 1 and 255
+        bool enabled = false;           //!< enables or disables SSCT
+    } ssct;
 };
 
 /**
@@ -333,31 +328,21 @@ struct ScreenSpaceReflectionsOptions {
     bool enabled = false;
 };
 
-/**
- * Options for the  screen-space guard band.
- * A guard band can be enabled to avoid some artifacts towards the edge of the screen when
- * using screen-space effects such as SSAO. Enabling the guard band reduces performance slightly.
- * Currently the guard band can only be enabled or disabled.
- */
-struct GuardBandOptions {
-    bool enabled = false;
-};
-
 /**
  * List of available post-processing anti-aliasing techniques.
  * @see setAntiAliasing, getAntiAliasing, setSampleCount
  */
 enum class AntiAliasing : uint8_t {
-    NONE,   //!< no anti aliasing performed as part of post-processing
-    FXAA    //!< FXAA is a low-quality but very efficient type of anti-aliasing. (default).
+    NONE = 0,   //!< no anti aliasing performed as part of post-processing
+    FXAA = 1    //!< FXAA is a low-quality but very efficient type of anti-aliasing. (default).
 };
 
 /**
  * List of available post-processing dithering techniques.
  */
 enum class Dithering : uint8_t {
-    NONE,       //!< No dithering
-    TEMPORAL    //!< Temporal dithering (default)
+    NONE = 0,       //!< No dithering
+    TEMPORAL = 1    //!< Temporal dithering (default)
 };
 
 /**

filament/include/filament/Renderer.h

@@ -80,8 +80,11 @@ public:
         // refresh-rate of the display in Hz. set to 0 for offscreen or turn off frame-pacing.
         float refreshRate = 60.0f;
 
-        [[deprecated]] uint64_t presentationDeadlineNanos = 0;
-        [[deprecated]] uint64_t vsyncOffsetNanos = 0;
+        // how far in advance a buffer must be queued for presentation at a given time in ns
+        uint64_t presentationDeadlineNanos = 0;
+
+        // offset by which vsyncSteadyClockTimeNano provided in beginFrame() is offset in ns
+        uint64_t vsyncOffsetNanos = 0;
     };
 
     /**
@@ -240,20 +243,6 @@ public:
     bool beginFrame(SwapChain* swapChain,
             uint64_t vsyncSteadyClockTimeNano = 0u);
 
-    /**
-     * Set the time at which the frame must be presented to the display.
-     *
-     * This must be called between beginFrame() and endFrame().
-     *
-     * @param monotonic_clock_ns  the time in nanoseconds corresponding to the system monotonic up-time clock.
-     *                            the presentation time is typically set in the middle of the period
-     *                            of interest. The presentation time cannot be too far in the
-     *                            future because it is limited by how many buffers are available in
-     *                            the display sub-system. Typically it is set to 1 or 2 vsync periods
-     *                            away.
-     */
-    void setPresentationTime(int64_t monotonic_clock_ns);
-
     /**
      * Render a View into this renderer's window.
      *

filament/include/filament/View.h

@@ -84,7 +84,6 @@ public:
     using VsmShadowOptions = VsmShadowOptions;
     using SoftShadowOptions = SoftShadowOptions;
     using ScreenSpaceReflectionsOptions = ScreenSpaceReflectionsOptions;
-    using GuardBandOptions = GuardBandOptions;
 
     /**
      * Sets the View's name. Only useful for debugging.
@@ -346,20 +345,6 @@ public:
      */
     ScreenSpaceReflectionsOptions const& getScreenSpaceReflectionsOptions() const noexcept;
 
-    /**
-     * Enables or disable screen-space guard band. Disabled by default.
-     *
-     * @param options guard band options
-     */
-    void setGuardBandOptions(GuardBandOptions options) noexcept;
-
-    /**
-     * Returns screen-space guard band options.
-     *
-     * @return guard band options
-     */
-    GuardBandOptions const& getGuardBandOptions() const noexcept;
-
     /**
      * Enables or disable multi-sample anti-aliasing (MSAA). Disabled by default.
      *

filament/src/PostProcessManager.cpp

@@ -1108,6 +1108,8 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::gaussianBlurPass(FrameGraph&
                 mi->setParameter("level", float(inSubDesc.level));
                 mi->setParameter("layer", float(inSubDesc.layer));
                 mi->setParameter("reinhard", reinhard ? uint32_t(1) : uint32_t(0));
+                mi->setParameter("resolution", float4{
+                        tempDesc.width, tempDesc.height, 1.0f / tempDesc.width, 1.0f / tempDesc.height });
                 mi->setParameter("axis",float2{ 1.0f / inDesc.width, 0 });
                 mi->setParameter("count", (int32_t)m);
                 mi->setParameter("kernel", kernel, m);
@@ -1129,6 +1131,8 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::gaussianBlurPass(FrameGraph&
                 });
                 mi->setParameter("level", 0.0f);
                 mi->setParameter("layer", 0.0f);
+                mi->setParameter("resolution",
+                        float4{ width, height, 1.0f / width, 1.0f / height });
                 mi->setParameter("axis", float2{ 0, 1.0f / tempDesc.height });
                 mi->commit(driver);
                 // we don't need to call use() here, since it's the same material
@@ -1526,7 +1530,8 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::dof(FrameGraph& fg,
                 mi->setParameter("cocClamp", float2{
                     -(dofOptions.maxForegroundCOC ? dofOptions.maxForegroundCOC : DOF_DEFAULT_MAX_COC),
                       dofOptions.maxBackgroundCOC ? dofOptions.maxBackgroundCOC : DOF_DEFAULT_MAX_COC});
-                mi->setParameter("texelSize", float2{ 1.0f / colorDesc.width, 1.0f / colorDesc.height });
+                mi->setParameter("uvscale", float4{ width, height,
+                        1.0f / colorDesc.width, 1.0f / colorDesc.height });
                 commitAndRender(out, material, driver);
             });
 
@@ -1581,15 +1586,15 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::dof(FrameGraph& fg,
                 const PipelineState pipeline(material.getPipelineState(mEngine, variant));
 
                 for (size_t level = 0 ; level < mipmapCount - 1u ; level++) {
-                    const float w = FTexture::valueForLevel(level, desc.width);
-                    const float h = FTexture::valueForLevel(level, desc.height);
+                    uint32_t w = FTexture::valueForLevel(level, desc.width);
+                    uint32_t h = FTexture::valueForLevel(level, desc.height);
 
                     auto const& out = resources.getRenderPassInfo(data.rp[level]);
                     driver.setMinMaxLevels(inOutColor, level, level);
                     driver.setMinMaxLevels(inOutCoc, level, level);
                     mi->setParameter("mip", uint32_t(level));
                     mi->setParameter("weightScale", 0.5f / float(1u<<level));   // FIXME: halfres?
-                    mi->setParameter("texelSize", float2{ 1.0f / w, 1.0f / h });
+                    mi->setParameter("pixelSize", 1.0f / float2{w, h});
                     mi->commit(driver);
                     render(out, pipeline, driver);
                 }
@@ -1608,12 +1613,10 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::dof(FrameGraph& fg,
     //       the DoF effect resolution?
     // Size of a tile in full-resolution pixels -- must match TILE_SIZE in dofDilate.mat
     const size_t tileSize = 16;
-
-    // we assume the width/height is already multiple of 16
-    assert_invariant(!(colorDesc.width  & 0xF) && !(colorDesc.height & 0xF));
-    const uint32_t tileBufferWidth  = colorDesc.width  / dofResolution;
-    const uint32_t tileBufferHeight = colorDesc.height / dofResolution;
-    const size_t tileReductionCount = ctz(tileSize / dofResolution);
+    // round the width/height to 16 (tile size), divide by scale (1 or 2) for halfres
+    const uint32_t tileBufferWidth  = ((colorDesc.width  + (tileSize - 1u)) & ~(tileSize - 1u)) / dofResolution;
+    const uint32_t tileBufferHeight = ((colorDesc.height + (tileSize - 1u)) & ~(tileSize - 1u)) / dofResolution;
+    const size_t tileReductionCount = std::log2(tileSize / dofResolution);
 
     struct PostProcessDofTiling1 {
         FrameGraphId<FrameGraphTexture> inCocMinMax;
@@ -1624,11 +1627,8 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::dof(FrameGraph& fg,
     for (size_t i = 0; i < tileReductionCount; i++) {
         auto& ppDoFTiling = fg.addPass<PostProcessDofTiling1>("DoF Tiling",
                 [&](FrameGraph::Builder& builder, auto& data) {
-
-                    // this must be true by construction
                     assert_invariant(((tileBufferWidth  >> i) & 1u) == 0);
                     assert_invariant(((tileBufferHeight >> i) & 1u) == 0);
-
                     data.inCocMinMax = builder.sample(inTilesCocMinMax);
                     data.outTilesCocMinMax = builder.createTexture("dof tiles output", {
                             .width  = tileBufferWidth  >> (i + 1u),
@@ -1640,6 +1640,7 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::dof(FrameGraph& fg,
                 [=](FrameGraphResources const& resources,
                         auto const& data, DriverApi& driver) {
                     auto const& inputDesc = resources.getDescriptor(data.inCocMinMax);
+                    auto const& outputDesc = resources.getDescriptor(data.outTilesCocMinMax);
                     auto const& out = resources.getRenderPassInfo();
                     auto inCocMinMax = resources.getTexture(data.inCocMinMax);
                     auto const& material = (!textureSwizzleSupported && (i == 0)) ?
@@ -1647,7 +1648,9 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::dof(FrameGraph& fg,
                             getPostProcessMaterial("dofTiles");
                     FMaterialInstance* const mi = material.getMaterialInstance(mEngine);
                     mi->setParameter("cocMinMax", inCocMinMax, { .filterMin = SamplerMinFilter::NEAREST });
-                    mi->setParameter("texelSize", float2{ 1.0f / inputDesc.width, 1.0f / inputDesc.height });
+                    mi->setParameter("uvscale", float4{
+                        outputDesc.width, outputDesc.height,
+                        1.0f / inputDesc.width, 1.0f / inputDesc.height });
                     commitAndRender(out, material, driver);
                 });
         inTilesCocMinMax = ppDoFTiling->outTilesCocMinMax;
@@ -1709,14 +1712,16 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::dof(FrameGraph& fg,
                 data.coc            = builder.sample(ppDoFMipmap->inOutCoc);
                 data.tilesCocMinMax = builder.sample(dilated);
 
+                // The DoF buffer (output) doesn't need to be a multiple of 8 because it's not
+                // mipmapped. We just need to adjust the uv properly.
                 data.outColor = builder.createTexture("dof color output", {
-                        .width  = colorDesc.width  / dofResolution,
-                        .height = colorDesc.height / dofResolution,
+                        .width  = (colorDesc.width  + (dofResolution / 2u)) / dofResolution,
+                        .height = (colorDesc.height + (dofResolution / 2u)) / dofResolution,
                         .format = fg.getDescriptor(data.color).format
                 });
                 data.outAlpha = builder.createTexture("dof alpha output", {
-                        .width  = colorDesc.width  / dofResolution,
-                        .height = colorDesc.height / dofResolution,
+                        .width  = builder.getDescriptor(data.outColor).width,
+                        .height = builder.getDescriptor(data.outColor).height,
                         .format = TextureFormat::R8
                 });
                 data.outColor  = builder.write(data.outColor, FrameGraphTexture::Usage::COLOR_ATTACHMENT);
@@ -1733,6 +1738,8 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::dof(FrameGraph& fg,
                 auto tilesCocMinMax = resources.getTexture(data.tilesCocMinMax);
 
                 auto const& inputDesc = resources.getDescriptor(data.coc);
+                auto const& outputDesc = resources.getDescriptor(data.outColor);
+                auto const& tilesDesc = resources.getDescriptor(data.tilesCocMinMax);
 
                 auto const& material = getPostProcessMaterial("dof");
                 FMaterialInstance* const mi = material.getMaterialInstance(mEngine);
@@ -1745,11 +1752,19 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::dof(FrameGraph& fg,
                         { .filterMin = SamplerMinFilter::NEAREST_MIPMAP_NEAREST });
                 mi->setParameter("tiles", tilesCocMinMax,
                         { .filterMin = SamplerMinFilter::NEAREST });
+
                 mi->setParameter("cocToTexelScale", float2{
                         bokehAspectRatio / (inputDesc.width  * dofResolution),
                                      1.0 / (inputDesc.height * dofResolution)
                 });
+
                 mi->setParameter("cocToPixelScale", (1.0f / dofResolution));
+                mi->setParameter("uvscale", float4{
+                    outputDesc.width  / float(inputDesc.width),
+                    outputDesc.height / float(inputDesc.height),
+                    outputDesc.width  / float(tileSize / dofResolution * tilesDesc.width),
+                    outputDesc.height / float(tileSize / dofResolution * tilesDesc.height)
+                });
                 mi->setParameter("ringCounts", float4{
                     dofOptions.foregroundRingCount ? dofOptions.foregroundRingCount : DOF_DEFAULT_RING_COUNT,
                     dofOptions.backgroundRingCount ? dofOptions.backgroundRingCount : DOF_DEFAULT_RING_COUNT,
@@ -1764,6 +1779,7 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::dof(FrameGraph& fg,
      * DoF median
      */
 
+
     struct PostProcessDofMedian {
         FrameGraphId<FrameGraphTexture> inColor;
         FrameGraphId<FrameGraphTexture> inAlpha;
@@ -1794,11 +1810,18 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::dof(FrameGraph& fg,
                 auto inAlpha        = resources.getTexture(data.inAlpha);
                 auto tilesCocMinMax = resources.getTexture(data.tilesCocMinMax);
 
+                auto const& outputDesc = resources.getDescriptor(data.outColor);
+                auto const& tilesDesc = resources.getDescriptor(data.tilesCocMinMax);
+
                 auto const& material = getPostProcessMaterial("dofMedian");
                 FMaterialInstance* const mi = material.getMaterialInstance(mEngine);
                 mi->setParameter("dof",   inColor,        { .filterMin = SamplerMinFilter::NEAREST_MIPMAP_NEAREST });
                 mi->setParameter("alpha", inAlpha,        { .filterMin = SamplerMinFilter::NEAREST_MIPMAP_NEAREST });
                 mi->setParameter("tiles", tilesCocMinMax, { .filterMin = SamplerMinFilter::NEAREST });
+                mi->setParameter("uvscale", float2{
+                        outputDesc.width  / float(tileSize / dofResolution * tilesDesc.width),
+                        outputDesc.height / float(tileSize / dofResolution * tilesDesc.height)
+                });
                 commitAndRender(out, material, driver);
             });
 
@@ -1834,6 +1857,8 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::dof(FrameGraph& fg,
             },
             [=](FrameGraphResources const& resources,
                     auto const& data, DriverApi& driver) {
+                auto const& dofDesc = resources.getDescriptor(data.dof);
+                auto const& tilesDesc = resources.getDescriptor(data.tilesCocMinMax);
                 auto const& out = resources.getRenderPassInfo();
 
                 auto color      = resources.getTexture(data.color);
@@ -1847,6 +1872,12 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::dof(FrameGraph& fg,
                 mi->setParameter("dof",   dof,   { .filterMag = SamplerMagFilter::NEAREST });
                 mi->setParameter("alpha", alpha, { .filterMag = SamplerMagFilter::NEAREST });
                 mi->setParameter("tiles", tilesCocMinMax, { .filterMin = SamplerMinFilter::NEAREST });
+                mi->setParameter("uvscale", float4{
+                    colorDesc.width  / (dofDesc.width    * float(dofResolution)),
+                    colorDesc.height / (dofDesc.height   * float(dofResolution)),
+                    colorDesc.width  / (tilesDesc.width  * float(tileSize)),
+                    colorDesc.height / (tilesDesc.height * float(tileSize))
+                });
                 commitAndRender(out, material, driver);
             });
 
@@ -2652,43 +2683,66 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::opaqueBlit(FrameGraph& fg,
     return ppBlit->output;
 }
 
-FrameGraphId<FrameGraphTexture> PostProcessManager::upscale(FrameGraph& fg, bool translucent,
-        DynamicResolutionOptions dsrOptions, FrameGraphId<FrameGraphTexture> input,
-        filament::Viewport const& vp, FrameGraphTexture::Descriptor const& outDesc,
-        backend::SamplerMagFilter filter) noexcept {
+FrameGraphId<FrameGraphTexture> PostProcessManager::blendBlit(
+        FrameGraph& fg, bool translucent, DynamicResolutionOptions dsrOptions,
+        FrameGraphId<FrameGraphTexture> input, filament::Viewport const& vp,
+        FrameGraphTexture::Descriptor const& outDesc) noexcept {
 
-    if (UTILS_LIKELY(!translucent && dsrOptions.quality == QualityLevel::LOW)) {
-        return opaqueBlit(fg, input, vp, outDesc, filter);
-    }
+    Handle<HwRenderPrimitive> fullScreenRenderPrimitive = mEngine.getFullScreenRenderPrimitive();
 
-    // The code below cannot handle sub-resources
-    assert_invariant(fg.getSubResourceDescriptor(input).layer == 0);
-    assert_invariant(fg.getSubResourceDescriptor(input).level == 0);
-
-    const bool lowQualityFallback = translucent && dsrOptions.quality != QualityLevel::LOW;
-    if (lowQualityFallback) {
+    bool lowQualityFallback = false;
+    if (translucent && dsrOptions.quality != QualityLevel::LOW) {
         // FidelityFX-FSR doesn't support the alpha channel currently
         dsrOptions.quality = QualityLevel::LOW;
+        lowQualityFallback = true;
     }
 
     const bool twoPassesEASU = mWorkaroundSplitEasu &&
             (dsrOptions.quality == QualityLevel::MEDIUM
                 || dsrOptions.quality == QualityLevel::HIGH);
 
+    // helper to set the EASU uniforms
+    auto setEasuUniforms = [vp](FMaterialInstance* mi,
+            FrameGraphTexture::Descriptor const& inputDesc,
+            FrameGraphTexture::Descriptor const& outputDesc) {
+        FSRUniforms uniforms{};
+        FSR_ScalingSetup(&uniforms, {
+                .input = vp,
+                .inputWidth = inputDesc.width,
+                .inputHeight = inputDesc.height,
+                .outputWidth = outputDesc.width,
+                .outputHeight = outputDesc.height,
+        });
+        mi->setParameter("EasuCon0", uniforms.EasuCon0);
+        mi->setParameter("EasuCon1", uniforms.EasuCon1);
+        mi->setParameter("EasuCon2", uniforms.EasuCon2);
+        mi->setParameter("EasuCon3", uniforms.EasuCon3);
+        mi->setParameter("textureSize",
+                float2{ inputDesc.width, inputDesc.height });
+    };
+
+    // helper to enable blending
+    auto enableTranslucentBlending = [](PipelineState& pipeline) {
+        pipeline.rasterState.blendFunctionSrcRGB = BlendFunction::ONE;
+        pipeline.rasterState.blendFunctionSrcAlpha = BlendFunction::ONE;
+        pipeline.rasterState.blendFunctionDstRGB = BlendFunction::ONE_MINUS_SRC_ALPHA;
+        pipeline.rasterState.blendFunctionDstAlpha = BlendFunction::ONE_MINUS_SRC_ALPHA;
+    };
+
     struct QuadBlitData {
         FrameGraphId<FrameGraphTexture> input;
         FrameGraphId<FrameGraphTexture> output;
         FrameGraphId<FrameGraphTexture> depth;
     };
 
-    auto& ppQuadBlit = fg.addPass<QuadBlitData>(dsrOptions.enabled ? "upscaling" : "compositing",
+    auto& ppQuadBlit = fg.addPass<QuadBlitData>(dsrOptions.enabled ? "quad scaling" : "blendBlit",
             [&](FrameGraph::Builder& builder, auto& data) {
                 data.input = builder.sample(input);
-                data.output = builder.createTexture("upscaled output", outDesc);
+                data.output = builder.createTexture("scaled output", outDesc);
 
                 if (twoPassesEASU) {
-                    // FIXME: it would be better to use the stencil buffer in this case (less bandwidth)
-                    data.depth = builder.createTexture("upscaled output depth", {
+                    // FIXME: it would be better to use the stencil buffer in this case
+                    data.depth = builder.createTexture("scaled output depth", {
                         .width = outDesc.width,
                         .height = outDesc.height,
                         .format = TextureFormat::DEPTH16
@@ -2701,38 +2755,9 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::upscale(FrameGraph& fg, bool
                         .attachments = { .color = { data.output }, .depth = { data.depth }},
                         .clearFlags = TargetBufferFlags::DEPTH });
             },
-            [this, twoPassesEASU, dsrOptions, vp, translucent, filter](FrameGraphResources const& resources,
+            [=](FrameGraphResources const& resources,
                     auto const& data, DriverApi& driver) {
 
-                // helper to set the EASU uniforms
-                auto setEasuUniforms = [vp, backend = mEngine.getBackend()](FMaterialInstance* mi,
-                        FrameGraphTexture::Descriptor const& inputDesc,
-                        FrameGraphTexture::Descriptor const& outputDesc) {
-                    FSRUniforms uniforms{};
-                    FSR_ScalingSetup(&uniforms, {
-                            .backend = backend,
-                            .input = vp,
-                            .inputWidth = inputDesc.width,
-                            .inputHeight = inputDesc.height,
-                            .outputWidth = outputDesc.width,
-                            .outputHeight = outputDesc.height,
-                    });
-                    mi->setParameter("EasuCon0", uniforms.EasuCon0);
-                    mi->setParameter("EasuCon1", uniforms.EasuCon1);
-                    mi->setParameter("EasuCon2", uniforms.EasuCon2);
-                    mi->setParameter("EasuCon3", uniforms.EasuCon3);
-                    mi->setParameter("textureSize",
-                            float2{ inputDesc.width, inputDesc.height });
-                };
-
-                // helper to enable blending
-                auto enableTranslucentBlending = [](PipelineState& pipeline) {
-                    pipeline.rasterState.blendFunctionSrcRGB = BlendFunction::ONE;
-                    pipeline.rasterState.blendFunctionSrcAlpha = BlendFunction::ONE;
-                    pipeline.rasterState.blendFunctionDstRGB = BlendFunction::ONE_MINUS_SRC_ALPHA;
-                    pipeline.rasterState.blendFunctionDstAlpha = BlendFunction::ONE_MINUS_SRC_ALPHA;
-                };
-
                 auto color = resources.getTexture(data.input);
                 auto const& inputDesc = resources.getDescriptor(data.input);
                 auto const& outputDesc = resources.getDescriptor(data.output);
@@ -2748,7 +2773,8 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::upscale(FrameGraph& fg, bool
                     auto* mi = splitEasuMaterial->getMaterialInstance(mEngine);
                     setEasuUniforms(mi, inputDesc, outputDesc);
                     mi->setParameter("color", color, {
-                        .filterMag = SamplerMagFilter::LINEAR
+                        .filterMag = SamplerMagFilter::LINEAR,
+                        .filterMin = SamplerMinFilter::LINEAR
                     });
                     mi->setParameter("resolution",
                             float4{ outputDesc.width, outputDesc.height,
@@ -2767,8 +2793,8 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::upscale(FrameGraph& fg, bool
                         setEasuUniforms(mi, inputDesc, outputDesc);
                     }
                     mi->setParameter("color", color, {
-                        .filterMag = (dsrOptions.quality == QualityLevel::LOW) ?
-                                filter : SamplerMagFilter::LINEAR
+                        .filterMag = SamplerMagFilter::LINEAR,
+                        .filterMin = SamplerMinFilter::LINEAR
                     });
                     mi->setParameter("resolution",
                             float4{ outputDesc.width, outputDesc.height,
@@ -2786,8 +2812,6 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::upscale(FrameGraph& fg, bool
                 // --------------------------------------------------------------------------------
                 // render pass with draw calls
 
-                auto fullScreenRenderPrimitive = mEngine.getFullScreenRenderPrimitive();
-
                 auto out = resources.getRenderPassInfo();
 
                 if (UTILS_UNLIKELY(twoPassesEASU)) {
@@ -2857,14 +2881,6 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::upscale(FrameGraph& fg, bool
     return output;
 }
 
-FrameGraphId<FrameGraphTexture> PostProcessManager::blit(FrameGraph& fg, bool translucent,
-        FrameGraphId<FrameGraphTexture> input, filament::Viewport const& vp,
-        FrameGraphTexture::Descriptor const& outDesc,
-        backend::SamplerMagFilter filter) noexcept {
-    // for now, we implement this by calling upscale() with the low-quality setting.
-    return upscale(fg, translucent, { .quality = QualityLevel::LOW }, input, vp, outDesc, filter);
-}
-
 FrameGraphId<FrameGraphTexture> PostProcessManager::resolveBaseLevel(FrameGraph& fg,
         const char* outputBufferName, FrameGraphId<FrameGraphTexture> input) noexcept {
     // Don't do anything if we're not a MSAA buffer

filament/src/PostProcessManager.h

@@ -211,15 +211,10 @@ public:
             FrameGraphTexture::Descriptor const& outDesc,
             backend::SamplerMagFilter filter = backend::SamplerMagFilter::LINEAR) noexcept;
 
-    FrameGraphId<FrameGraphTexture> upscale(FrameGraph& fg, bool translucent,
-            DynamicResolutionOptions dsrOptions, FrameGraphId<FrameGraphTexture> input,
-            filament::Viewport const& vp, FrameGraphTexture::Descriptor const& outDesc,
-            backend::SamplerMagFilter filter = backend::SamplerMagFilter::LINEAR) noexcept;
-
-    FrameGraphId<FrameGraphTexture> blit(FrameGraph& fg, bool translucent,
+    FrameGraphId<FrameGraphTexture> blendBlit(
+            FrameGraph& fg, bool translucent, DynamicResolutionOptions dsrOptions,
             FrameGraphId<FrameGraphTexture> input, filament::Viewport const& vp,
-            FrameGraphTexture::Descriptor const& outDesc,
-            backend::SamplerMagFilter filter = backend::SamplerMagFilter::LINEAR) noexcept;
+            FrameGraphTexture::Descriptor const& outDesc) noexcept;
 
     // resolve base level of input and outputs a 1-level texture
     FrameGraphId<FrameGraphTexture> resolveBaseLevel(FrameGraph& fg,

filament/src/RenderPass.cpp

@@ -51,7 +51,6 @@ RenderPass::RenderPass(RenderPass const& rhs) = default;
 RenderPass::~RenderPass() noexcept = default;
 
 RenderPass::Command* RenderPass::append(size_t count) noexcept {
-    // this is like a "in-place" realloc(). Works only with LinearAllocator.
     Command* const curr = mCommandArena.alloc<Command>(count);
     assert_invariant(mCommandBegin == nullptr || curr == mCommandEnd);
     if (mCommandBegin == nullptr) {
@@ -302,10 +301,9 @@ void RenderPass::generateCommandsImpl(uint32_t extraFlags,
     auto const* const UTILS_RESTRICT soaWorldAABBCenter = soa.data<FScene::WORLD_AABB_CENTER>();
     auto const* const UTILS_RESTRICT soaVisibility      = soa.data<FScene::VISIBILITY_STATE>();
     auto const* const UTILS_RESTRICT soaPrimitives      = soa.data<FScene::PRIMITIVES>();
-    auto const* const UTILS_RESTRICT soaSkinning        = soa.data<FScene::SKINNING_BUFFER>();
     auto const* const UTILS_RESTRICT soaMorphing        = soa.data<FScene::MORPHING_BUFFER>();
     auto const* const UTILS_RESTRICT soaVisibilityMask  = soa.data<FScene::VISIBLE_MASK>();
-    auto const* const UTILS_RESTRICT soaInstanceCount   = soa.data<FScene::INSTANCE_COUNT>();
+    auto const* const UTILS_RESTRICT soaInstanceCount  = soa.data<FScene::INSTANCE_COUNT>();
 
     const bool hasShadowing = renderFlags & HAS_SHADOWING;
     const bool viewInverseFrontFaces = renderFlags & HAS_INVERSE_FRONT_FACES;
@@ -398,7 +396,6 @@ void RenderPass::generateCommandsImpl(uint32_t extraFlags,
         const bool writeDepthForShadowCasters = depthContainsShadowCasters & shadowCaster;
 
         const Slice<FRenderPrimitive>& primitives = soaPrimitives[i];
-        const FRenderableManager::SkinningBindingInfo& skinning = soaSkinning[i];
         const FRenderableManager::MorphingBindingInfo& morphing = soaMorphing[i];
 
         /*
@@ -415,8 +412,6 @@ void RenderPass::generateCommandsImpl(uint32_t extraFlags,
                 cmdColor.primitive.primitiveHandle = primitive.getHwHandle();
                 RenderPass::setupColorCommand(cmdColor, variant, mi, inverseFrontFaces);
 
-                cmdColor.primitive.skinningHandle = skinning.handle;
-                cmdColor.primitive.skinningOffset = skinning.offset;
                 cmdColor.primitive.morphWeightBuffer = morphing.handle;
                 cmdColor.primitive.morphTargetBuffer = morphTargets.buffer->getHwHandle();
 
@@ -511,8 +506,6 @@ void RenderPass::generateCommandsImpl(uint32_t extraFlags,
                 cmdDepth.primitive.mi = mi;
                 cmdDepth.primitive.rasterState.culling = mi->getCullingMode();
 
-                cmdDepth.primitive.skinningHandle = skinning.handle;
-                cmdDepth.primitive.skinningOffset = skinning.offset;
                 cmdDepth.primitive.morphWeightBuffer = morphing.handle;
                 cmdDepth.primitive.morphTargetBuffer = morphTargets.buffer->getHwHandle();
 
@@ -559,18 +552,22 @@ void RenderPass::Executor::execute(const char* name,
     engine.flush();
 
     driver.beginRenderPass(renderTarget, params);
-    recordDriverCommands(engine, driver, mBegin, mEnd, params.readOnlyDepthStencil);
+    recordDriverCommands(engine, driver, mBegin, mEnd, mRenderableSoa, params.readOnlyDepthStencil);
     driver.endRenderPass();
 }
 
 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::recordDriverCommands(FEngine& engine,
+        backend::DriverApi& driver,
+        const Command* first, const Command* last,
+        FScene::RenderableSoa const& soa, uint16_t readOnlyDepthStencil) const noexcept {
     SYSTRACE_CALL();
 
     if (first != last) {
         SYSTRACE_VALUE32("commandCount", last - first);
 
+        auto const* const UTILS_RESTRICT soaSkinning = soa.data<FScene::SKINNING_BUFFER>();
+
         PolygonOffset dummyPolyOffset;
         PipelineState pipeline{ .polygonOffset = mPolygonOffset };
         PolygonOffset* const pPipelinePolygonOffset =
@@ -617,13 +614,14 @@ void RenderPass::Executor::recordDriverCommands(FEngine& engine, backend::Driver
             driver.bindUniformBufferRange(BindingPoints::PER_RENDERABLE,
                     uboHandle, offset, sizeof(PerRenderableUib));
 
-            if (UTILS_UNLIKELY(info.skinningHandle)) {
+            auto skinning = soaSkinning[info.index];
+            if (UTILS_UNLIKELY(skinning.handle)) {
                 // note: we can't bind less than CONFIG_MAX_BONE_COUNT due to glsl limitations
                 driver.bindUniformBufferRange(BindingPoints::PER_RENDERABLE_BONES,
-                        info.skinningHandle,
-                        info.skinningOffset * sizeof(PerRenderableUibBone),
+                        skinning.handle,
+                        skinning.offset * sizeof(PerRenderableUibBone),
                         CONFIG_MAX_BONE_COUNT * sizeof(PerRenderableUibBone));
-                // note: even if only skinning is enabled, binding morphTargetBuffer is needed.
+                // note: even if skinning is only enabled, binding morphTargetBuffer is needed.
                 driver.bindSamplers(BindingPoints::PER_RENDERABLE_MORPHING,
                         info.morphTargetBuffer);
             }
@@ -645,7 +643,7 @@ 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),
+        : mEngine(pass->mEngine), mBegin(b), mEnd(e), mRenderableSoa(*pass->mRenderableSoa),
           mCustomCommands(pass->mCustomCommands), mUboHandle(pass->mUboHandle),
           mPolygonOffset(pass->mPolygonOffset),
           mPolygonOffsetOverride(pass->mPolygonOffsetOverride) {

filament/src/RenderPass.h

@@ -213,28 +213,22 @@ public:
         return boolish ? std::numeric_limits<uint64_t>::max() : uint64_t(0);
     }
 
-    struct PrimitiveInfo { // 48 bytes
-        union {
-            FMaterialInstance const* mi;
-            uint64_t padding = {}; // ensures mi is 8 bytes on all archs
-        };                                                              // 8 bytes
-        backend::RasterState rasterState;                               // 8 bytes
+    struct PrimitiveInfo { // 32 bytes
+        FMaterialInstance const* mi = nullptr;                          // 8 bytes (4)
         backend::Handle<backend::HwRenderPrimitive> primitiveHandle;    // 4 bytes
-        backend::Handle<backend::HwBufferObject> skinningHandle;        // 4 bytes
         backend::Handle<backend::HwBufferObject> morphWeightBuffer;     // 4 bytes
         backend::Handle<backend::HwSamplerGroup> morphTargetBuffer;     // 4 bytes
-        uint32_t index = 0;                                             // 4 bytes
-        uint32_t skinningOffset = 0;                                    // 4 bytes
+        backend::RasterState rasterState;                               // 4 bytes
+        uint16_t index = 0;                                             // 2 bytes
         uint16_t instanceCount;                                         // 2 bytes
         Variant materialVariant;                                        // 1 byte
-        uint8_t reserved[5] = {};                                       // 5 bytes
+        uint8_t reserved[11 - sizeof(void*)] = {};                      // 3 bytes (7)
     };
-    static_assert(sizeof(PrimitiveInfo) == 48);
+    static_assert(sizeof(PrimitiveInfo) == 32);
 
-    struct alignas(8) Command {     // 64 bytes
+    struct alignas(8) Command {     // 40 bytes
         CommandKey key = 0;         //  8 bytes
-        PrimitiveInfo primitive;    // 48 bytes
-        uint64_t reserved = 0;      //  8 bytes
+        PrimitiveInfo primitive;    // 32 bytes
         bool operator < (Command const& rhs) const noexcept { return key < rhs.key; }
         // placement new declared as "throw" to avoid the compiler's null-check
         inline void* operator new (std::size_t, void* ptr) {
@@ -242,7 +236,7 @@ public:
             return ptr;
         }
     };
-    static_assert(sizeof(Command) == 64);
+    static_assert(sizeof(Command) == 40);
     static_assert(std::is_trivially_destructible_v<Command>,
             "Command isn't trivially destructible");
 
@@ -252,7 +246,7 @@ public:
 
     // Arena used for commands
     using Arena = utils::Arena<
-            utils::LinearAllocator,                 // note: can't change this allocator
+            utils::LinearAllocator,
             utils::LockingPolicy::NoLock,
             utils::TrackingPolicy::HighWatermark,
             utils::AreaPolicy::StaticArea>;
@@ -322,6 +316,7 @@ public:
         FEngine& mEngine;
         Command const* mBegin;
         Command const* mEnd;
+        FScene::RenderableSoa const& mRenderableSoa;
         const CustomCommandVector mCustomCommands;
         const backend::Handle<backend::HwBufferObject> mUboHandle;
         const backend::PolygonOffset mPolygonOffset;
@@ -330,7 +325,8 @@ public:
         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;
+                const Command* first, const Command* last,
+                FScene::RenderableSoa const& soa, uint16_t readOnlyDepthStencil) const noexcept;
 
     public:
         Executor(Executor const& rhs);

filament/src/Renderer.cpp

@@ -32,10 +32,6 @@ void Renderer::render(View const* view) {
     upcast(this)->render(upcast(view));
 }
 
-void Renderer::setPresentationTime(int64_t monotonic_clock_ns) {
-    upcast(this)->setPresentationTime(monotonic_clock_ns);
-}
-
 bool Renderer::beginFrame(SwapChain* swapChain, uint64_t vsyncSteadyClockTimeNano) {
     return upcast(this)->beginFrame(upcast(swapChain), vsyncSteadyClockTimeNano);
 }

filament/src/ShadowMap.cpp

@@ -22,6 +22,7 @@
 
 #include "details/Engine.h"
 #include "details/Scene.h"
+#include "details/View.h"
 
 #include <backend/DriverEnums.h>
 
@@ -30,11 +31,10 @@
 
 #include <limits>
 
+using namespace filament::math;
 using namespace utils;
 
 namespace filament {
-
-using namespace math;
 using namespace backend;
 
 // do this only if depth-clamp is available

filament/src/ShadowMap.h

@@ -30,43 +30,9 @@
 
 namespace filament {
 
+class FView;
 class RenderPass;
 
-// The value of the 'VISIBLE_MASK' after culling. Each bit represents visibility in a frustum
-// (either camera or light).
-//
-//                                    1
-// bits                               5 ... 7 6 5 4 3 2 1 0
-// +------------------------------------------------------+
-// VISIBLE_RENDERABLE                                     X
-// VISIBLE_DIR_SHADOW_RENDERABLE                        X
-// VISIBLE_SPOT_SHADOW_RENDERABLE_0                   X
-// VISIBLE_SPOT_SHADOW_RENDERABLE_1                 X
-// ...
-
-// A "shadow renderable" is a renderable rendered to the shadow map during a shadow pass:
-// PCF shadows: only shadow casters
-// VSM shadows: both shadow casters and shadow receivers
-
-static constexpr size_t VISIBLE_RENDERABLE_BIT              = 0u;
-static constexpr size_t VISIBLE_DIR_SHADOW_RENDERABLE_BIT   = 1u;
-
-static constexpr size_t VISIBLE_SPOT_SHADOW_RENDERABLE_N_BIT(size_t n) { return n + 2; }
-
-static constexpr Culler::result_type VISIBLE_DIR_SHADOW_RENDERABLE = 1u << VISIBLE_DIR_SHADOW_RENDERABLE_BIT;
-static constexpr Culler::result_type VISIBLE_SPOT_SHADOW_RENDERABLE_N(size_t n) {
-    return 1u << VISIBLE_SPOT_SHADOW_RENDERABLE_N_BIT(n);
-}
-
-// ORing of all the VISIBLE_SPOT_SHADOW_RENDERABLE bits
-static constexpr Culler::result_type VISIBLE_SPOT_SHADOW_RENDERABLE =
-        (0xFFu >> (sizeof(Culler::result_type) * 8u - CONFIG_MAX_SHADOW_CASTING_SPOTS)) << 2u;
-
-// Because we're using a uint16_t for the visibility mask, we're limited to 14 spot light shadows.
-// (2 of the bits are used for visible renderables + directional light shadow casters).
-static_assert(CONFIG_MAX_SHADOW_CASTING_SPOTS <= sizeof(Culler::result_type) * 8 - 2,
-        "CONFIG_MAX_SHADOW_CASTING_SPOTS cannot be higher than 14.");
-
 class ShadowMap {
 public:
     explicit ShadowMap(FEngine& engine) noexcept;
@@ -141,7 +107,7 @@ public:
             SceneInfo& sceneInfo) noexcept;
 
     void render(FScene const& scene, utils::Range<uint32_t> range,
-            FScene::VisibleMaskType visibilityMask, RenderPass* pass) noexcept;
+            FScene::VisibleMaskType visibilityMask, RenderPass* const pass) noexcept;
 
     // Do we have visible shadows. Valid after calling update().
     bool hasVisibleShadows() const noexcept { return mHasVisibleShadows; }

filament/src/ShadowMapManager.cpp

@@ -24,6 +24,8 @@
 
 #include <fg/FrameGraph.h>
 
+#include <private/filament/SibGenerator.h>
+
 #include <utils/debug.h>
 #include <utils/FixedCapacityVector.h>
 
@@ -115,7 +117,8 @@ void ShadowMapManager::addSpotShadowMap(size_t lightIndex,
     mSpotShadowMaps.emplace_back(shadowMap, lightIndex, options);
 }
 
-FrameGraphId<FrameGraphTexture> ShadowMapManager::render(FrameGraph& fg, FEngine& engine,
+FrameGraphId<FrameGraphTexture> ShadowMapManager::render(FrameGraph& fg,
+        FEngine& engine, backend::DriverApi& driver,
         RenderPass const& pass, FView& view) noexcept {
 
     constexpr size_t MAX_SHADOW_LAYERS =

filament/src/ShadowMapManager.h

@@ -29,6 +29,7 @@
 
 #include <private/backend/DriverApi.h>
 #include <private/backend/DriverApiForward.h>
+#include <private/backend/SamplerGroup.h>
 
 #include <backend/DriverEnums.h>
 #include <backend/Handle.h>
@@ -39,11 +40,13 @@
 
 #include <array>
 #include <memory>
+#include <vector>
 
 namespace filament {
 
 class FView;
 class FrameGraph;
+class ShadowMap;
 class RenderPass;
 
 struct ShadowMappingUniforms {
@@ -86,7 +89,7 @@ public:
             FScene::RenderableSoa& renderableData, FScene::LightSoa& lightData) noexcept;
 
     // Renders all the shadow maps.
-    FrameGraphId<FrameGraphTexture> render(FrameGraph& fg, FEngine& engine,
+    FrameGraphId<FrameGraphTexture> render(FrameGraph& fg, FEngine& engine, backend::DriverApi& driver,
             RenderPass const& pass, FView& view) noexcept;
 
     ShadowMap* getCascadeShadowMap(size_t cascade) noexcept {

filament/src/View.cpp

@@ -123,14 +123,6 @@ const View::ScreenSpaceReflectionsOptions& View::getScreenSpaceReflectionsOption
     return upcast(this)->getScreenSpaceReflectionsOptions();
 }
 
-void View::setGuardBandOptions(GuardBandOptions options) noexcept {
-    upcast(this)->setGuardBandOptions(options);
-}
-
-GuardBandOptions const& View::getGuardBandOptions() const noexcept {
-    return upcast(this)->getGuardBandOptions();
-}
-
 void View::setColorGrading(ColorGrading* colorGrading) noexcept {
     return upcast(this)->setColorGrading(upcast(colorGrading));
 }

filament/src/details/Engine.h

@@ -477,6 +477,13 @@ public:
             float dzn = -1.0f;
             float dzf =  1.0f;
         } shadowmap;
+        struct {
+            bool enabled = true;
+            int sampleCount = 7;
+            int spiralTurns = 1;
+            int kernelSize = 23;
+            float stddev = 8.0f;
+        } ssao;
         struct {
             bool camera_at_origin = true;
             struct {

filament/src/details/Renderer.cpp

@@ -66,6 +66,8 @@ FRenderer::FRenderer(FEngine& engine) :
         mPerRenderPassArena(engine.getPerRenderPassAllocator())
 {
     FDebugRegistry& debugRegistry = engine.getDebugRegistry();
+    debugRegistry.registerProperty("d.ssao.enabled",
+            &engine.debug.ssao.enabled);
     debugRegistry.registerProperty("d.renderer.doFrameCapture",
             &engine.debug.renderer.doFrameCapture);
 
@@ -175,10 +177,6 @@ void FRenderer::initializeClearFlags() {
                   | TargetBufferFlags::DEPTH_AND_STENCIL;
 }
 
-void FRenderer::setPresentationTime(int64_t monotonic_clock_ns) {
-    FEngine::DriverApi& driver = mEngine.getDriverApi();
-    driver.setPresentationTime(monotonic_clock_ns);
-}
 
 bool FRenderer::beginFrame(FSwapChain* swapChain, uint64_t vsyncSteadyClockTimeNano) {
     assert_invariant(swapChain);
@@ -192,6 +190,7 @@ bool FRenderer::beginFrame(FSwapChain* swapChain, uint64_t vsyncSteadyClockTimeN
     const time_point<steady_clock> appVsync(vsyncSteadyClockTimeNano ? userVsync : now);
 
     mFrameId++;
+    mViewRenderedCount = 0;
 
     { // scope for frame id trace
         char buf[64];
@@ -231,7 +230,7 @@ bool FRenderer::beginFrame(FSwapChain* swapChain, uint64_t vsyncSteadyClockTimeN
     * to ignore the return value and render the frame anyways -- which is perfectly fine.
     * The remaining work will be done when the first render() call is made.
     */
-    auto beginFrameInternal = [this, appVsync]() {
+    auto beginFrameInternal = [this, appVsync, vsyncSteadyClockTimeNano]() {
         FEngine& engine = mEngine;
         FEngine::DriverApi& driver = engine.getDriverApi();
 
@@ -244,6 +243,38 @@ bool FRenderer::beginFrame(FSwapChain* swapChain, uint64_t vsyncSteadyClockTimeN
                 .historySize = mFrameRateOptions.history
         }, mFrameId);
 
+        if (false && vsyncSteadyClockTimeNano) { // work in progress
+            const size_t interval = mFrameRateOptions.interval; // user requested swap-interval;
+            const steady_clock::duration refreshPeriod(uint64_t(1e9 / mDisplayInfo.refreshRate));
+            const steady_clock::duration presentationDeadline(mDisplayInfo.presentationDeadlineNanos);
+            const steady_clock::duration vsyncOffset(mDisplayInfo.vsyncOffsetNanos);
+
+            // hardware vsync timestamp
+            steady_clock::time_point hwVsync = appVsync - vsyncOffset;
+
+            // compute our desired presentation time. We can't pick a desired presentation time
+            // that's too far, or we won't be able to dequeue buffers.
+            steady_clock::time_point desiredPresentationTime = hwVsync + 2 * interval * refreshPeriod;
+
+            // Compute the deadline. This deadline is when the GPU must be finished.
+            // The deadline has 1ms backed in it on Android.
+            UTILS_UNUSED_IN_RELEASE
+            steady_clock::time_point deadline = desiredPresentationTime - presentationDeadline;
+
+            // one important thing is to make sure that the deadline is comfortably later than
+            // when the gpu will finish, otherwise we'll have inconsistent latency/frames.
+
+            // TODO: evaluate if we can make it in time, and if not why.
+            //   If the problem is cpu+gpu latency we can try to push the desired presentation time
+            //   further away, but this has limits, as only 2 buffers are dequeuable.
+            //   If the problem is the gpu is overwhelmed, then we need to
+            //    - see if there is more headroom in dynamic resolution
+            //    - or start skipping frames. Ideally lower the framerate to
+            // presentation time is set to the middle of the period we're interested in
+            steady_clock::time_point presentationTime = desiredPresentationTime - refreshPeriod / 2;
+            driver.setPresentationTime(presentationTime.time_since_epoch().count());
+        }
+
         // ask the engine to do what it needs to (e.g. updates light buffer, materials...)
         engine.prepare();
     };
@@ -420,9 +451,13 @@ void FRenderer::render(FView const* view) {
     }
 
     if (UTILS_LIKELY(view && view->getScene())) {
-        // NOTE: in the past we tried to kick the GPU here with a flush (b2cdf9f), but this
-        // was problematic on certain devices. b/232224942
+        if (mViewRenderedCount) {
+            // this is a good place to kick the GPU, since we've rendered a View before
+            // and we're about to render another one.
+            mEngine.getDriverApi().flush();
+        }
         renderInternal(view);
+        mViewRenderedCount++;
     }
 }
 
@@ -469,7 +504,6 @@ void FRenderer::renderJob(ArenaScope& arena, FView& view) {
     auto vignetteOptions = view.getVignetteOptions();
     auto colorGrading = view.getColorGrading();
     auto ssReflectionsOptions = view.getScreenSpaceReflectionsOptions();
-    auto guardBandOptions = view.getGuardBandOptions();
     const uint8_t msaaSampleCount = msaaOptions.enabled ? msaaOptions.sampleCount : 1u;
     if (!hasPostProcess) {
         // disable all effects that are part of post-processing
@@ -532,15 +566,12 @@ void FRenderer::renderJob(ArenaScope& arena, FView& view) {
     CameraInfo cameraInfo = view.computeCameraInfo(engine);
 
     // when colorgrading-as-subpass is active, we know that many other effects are disabled
-    // such as dof, bloom. Moreover, if fxaa and scaling are not enabled, we're essentially in
+    // such as dof, bloom. Moreover if fxaa and scaling are not enabled, we're essentially in
     // a very fast rendering path -- in this case, we would need an extra blit to "resolve" the
     // buffer padding (because there are no other pass that can do it as a side effect).
     // In this case, it is better to skip the padding, which won't be helping much.
     const bool noBufferPadding = colorGradingConfig.asSubpass && !hasFXAA && !scaled;
 
-    // guardBand must be a multiple of 16 to guarantee the same exact rendering up to 4 mip levels.
-    float guardBand = guardBandOptions.enabled ? 16.0f : 0.0f;
-
     if (hasPostProcess && !noBufferPadding) {
         // We always pad the rendering viewport to dimensions multiple of 16, this guarantees
         // that up to 4 mipmap levels are possible with an exact 1:2 scale. This also helps
@@ -555,14 +586,10 @@ void FRenderer::renderJob(ArenaScope& arena, FView& view) {
         //       Without post-processing, we usually draw directly into
         //       the SwapChain, and we might want to keep it this way.
 
-        auto round = [](uint32_t x) {
-            constexpr uint32_t rounding = 16u;
-            return (x + (rounding - 1u)) & ~(rounding - 1u);
-        };
-
         // compute the new rendering width and height, multiple of 16.
-        const float width  = float(round(svp.width )) + 2.0f * guardBand;
-        const float height = float(round(svp.height)) + 2.0f * guardBand;
+        constexpr uint32_t rounding = 16u;
+        const float width  = float( (svp.width  + (rounding - 1u)) & ~(rounding - 1u) );
+        const float height = float( (svp.height + (rounding - 1u)) & ~(rounding - 1u) );
 
         // scale the field-of-view up, so it covers exactly the extra pixels
         const float3 clipSpaceScaling{
@@ -577,8 +604,8 @@ void FRenderer::renderJob(ArenaScope& arena, FView& view) {
         // note: this creates an asymmetric frustum -- but we eventually copy only the
         // left/bottom part, which is a symmetric region.
         const float2 clipSpaceTranslation{
-                1.0f - clipSpaceScaling.x - 2.0f * guardBand / width,
-                1.0f - clipSpaceScaling.y - 2.0f * guardBand / height
+                1.0f - clipSpaceScaling.x,
+                1.0f - clipSpaceScaling.y
         };
 
         mat4f ts = mat4f::scaling(clipSpaceScaling);
@@ -590,8 +617,6 @@ void FRenderer::renderJob(ArenaScope& arena, FView& view) {
         // adjust svp to the new, larger, rendering dimensions
         svp.width  = uint32_t(width);
         svp.height = uint32_t(height);
-        xvp.left   = int32_t(guardBand);
-        xvp.bottom = int32_t(guardBand);
     }
 
     view.prepare(engine, driver, arena, svp, cameraInfo, getShaderUserTime(), needsAlphaChannel);
@@ -960,7 +985,7 @@ void FRenderer::renderJob(ArenaScope& arena, FView& view) {
     // --------------------------------------------------------------------------------------------
     // Post Processing...
 
-    UTILS_UNUSED_IN_RELEASE auto const& inputDesc = fg.getDescriptor(input);
+    auto const& inputDesc = fg.getDescriptor(input);
     assert_invariant(inputDesc.width == svp.width);
     assert_invariant(inputDesc.height == svp.height);
 
@@ -992,7 +1017,6 @@ void FRenderer::renderJob(ArenaScope& arena, FView& view) {
                         bloomOptions, vignetteOptions);
                 // the padded buffer is resolved now
                 xvp.left = xvp.bottom = 0;
-                svp = xvp;
             }
         }
 
@@ -1001,16 +1025,19 @@ void FRenderer::renderJob(ArenaScope& arena, FView& view) {
                     !hasColorGrading || needsAlphaChannel);
             // the padded buffer is resolved now
             xvp.left = xvp.bottom = 0;
-            svp = xvp;
         }
         if (scaled) {
             mightNeedFinalBlit = false;
             auto viewport = DEBUG_DYNAMIC_SCALING ? xvp : vp;
-            input = ppm.upscale(fg, blending, dsrOptions, input, xvp, {
-                    .width = viewport.width, .height = viewport.height,
-                    .format = colorGradingConfig.ldrFormat });
-            xvp.left = xvp.bottom = 0;
-            svp = xvp;
+            if (UTILS_LIKELY(!blending && dsrOptions.quality == QualityLevel::LOW)) {
+                input = ppm.opaqueBlit(fg, input, xvp, {
+                        .width = viewport.width, .height = viewport.height,
+                        .format = colorGradingConfig.ldrFormat }, SamplerMagFilter::LINEAR);
+            } else {
+                input = ppm.blendBlit(fg, blending, dsrOptions, input, xvp, {
+                        .width = viewport.width, .height = viewport.height,
+                        .format = colorGradingConfig.ldrFormat });
+            }
         }
     }
 
@@ -1036,14 +1063,22 @@ void FRenderer::renderJob(ArenaScope& arena, FView& view) {
             (outputIsSwapChain &&
                     (msaaSampleCount > 1 ||
                     colorGradingConfig.asSubpass ||
-                    ssReflectionsOptions.enabled))) {
+                    ssReflectionsOptions.enabled)))
+        {
             assert_invariant(!scaled);
-            input = ppm.blit(fg, blending, input, xvp, {
-                    .width = vp.width, .height = vp.height,
-                    .format = colorGradingConfig.ldrFormat }, SamplerMagFilter::NEAREST);
+            if (UTILS_LIKELY(!blending)) {
+                input = ppm.opaqueBlit(fg, input, xvp, {
+                        .width = vp.width, .height = vp.height,
+                        .format = colorGradingConfig.ldrFormat }, SamplerMagFilter::NEAREST);
+            } else {
+                input = ppm.blendBlit(fg, blending, { .quality = QualityLevel::LOW }, input, xvp, {
+                        .width = vp.width, .height = vp.height,
+                        .format = colorGradingConfig.ldrFormat});
+            }
         }
     }
 
+
 //    auto debug = structure
 //    fg.forwardResource(fgViewRenderTarget, debug ? debug : input);
 

filament/src/details/Renderer.h

@@ -75,9 +75,6 @@ public:
     // renders a single standalone view. The view must have a a custom rendertarget.
     void renderStandaloneView(FView const* view);
 
-
-    void setPresentationTime(int64_t monotonic_clock_ns);
-
     // start a frame
     bool beginFrame(FSwapChain* swapChain, uint64_t vsyncSteadyClockTimeNano);
 
@@ -102,7 +99,7 @@ public:
 
 
     void setDisplayInfo(DisplayInfo const& info) noexcept {
-        mDisplayInfo.refreshRate = info.refreshRate;
+        mDisplayInfo = info;
     }
 
     void setFrameRateOptions(FrameRateOptions const& options) noexcept {
@@ -164,6 +161,7 @@ private:
     FSwapChain* mSwapChain = nullptr;
     size_t mCommandsHighWatermark = 0;
     uint32_t mFrameId = 0;
+    uint32_t mViewRenderedCount = 0;
     FrameInfoManager mFrameInfoManager;
     backend::TextureFormat mHdrTranslucent;
     backend::TextureFormat mHdrQualityMedium;

filament/src/details/Scene.cpp

@@ -32,7 +32,6 @@
 
 #include <algorithm>
 
-using namespace filament::backend;
 using namespace filament::math;
 using namespace utils;
 
@@ -197,9 +196,7 @@ void FScene::prepare(const mat4& worldOriginTransform, bool shadowReceiversAreCa
     }
 }
 
-void FScene::updateUBOs(
-        Range<uint32_t> visibleRenderables,
-        Handle<HwBufferObject> renderableUbh) noexcept {
+void FScene::updateUBOs(utils::Range<uint32_t> visibleRenderables, backend::Handle<backend::HwBufferObject> renderableUbh) noexcept {
     FEngine::DriverApi& driver = mEngine.getDriverApi();
     FRenderableManager& rcm = mEngine.getRenderableManager();
 
@@ -222,7 +219,7 @@ void FScene::updateUBOs(
 
         // Using mat3f::getTransformForNormals handles non-uniform scaling, but DOESN'T guarantee that
         // the transformed normals will have unit-length, therefore they need to be normalized
-        // in the shader (that's already the case anyway, since normalization is needed after
+        // in the shader (that's already the case anyways, since normalization is needed after
         // interpolation).
         //
         // We pre-scale normals by the inverse of the largest scale factor to avoid
@@ -250,17 +247,20 @@ void FScene::updateUBOs(
         hasContactShadows = hasContactShadows || visibility.screenSpaceContactShadows;
 
         UniformBuffer::setUniform(buffer,
-                offset + offsetof(PerRenderableUib, flagsChannels),
-                PerRenderableUib::packFlagsChannels(
+                offset + offsetof(PerRenderableUib, flags),
+                PerRenderableUib::packFlags(
                         visibility.skinning,
                         visibility.morphing,
-                        visibility.screenSpaceContactShadows,
-                        sceneData.elementAt<CHANNELS>(i)));
+                        visibility.screenSpaceContactShadows));
 
         UniformBuffer::setUniform(buffer,
                 offset + offsetof(PerRenderableUib, morphTargetCount),
                 sceneData.elementAt<MORPHING_BUFFER>(i).count);
 
+        UniformBuffer::setUniform(buffer,
+                offset + offsetof(PerRenderableUib, channels),
+                (uint32_t)sceneData.elementAt<CHANNELS>(i));
+
         UniformBuffer::setUniform(buffer,
                 offset + offsetof(PerRenderableUib, objectId),
                 rcm.getEntity(ri).getId()); // we could also store the entity in sceneData
@@ -287,7 +287,7 @@ void FScene::terminate(FEngine& engine) {
 }
 
 void FScene::prepareDynamicLights(const CameraInfo& camera, ArenaScope& rootArena,
-        Handle<HwBufferObject> lightUbh) noexcept {
+        backend::Handle<backend::HwBufferObject> lightUbh) noexcept {
     FEngine::DriverApi& driver = mEngine.getDriverApi();
     FLightManager& lcm = mEngine.getLightManager();
     FScene::LightSoa& lightData = getLightData();
@@ -450,7 +450,7 @@ bool FScene::hasContactShadows() const noexcept {
 }
 
 UTILS_NOINLINE
-void FScene::forEach(Invocable<void(Entity)>&& functor) const noexcept {
+void FScene::forEach(Invocable<void(utils::Entity)>&& functor) const noexcept {
     std::for_each(mEntities.begin(), mEntities.end(), std::move(functor));
 }
 

filament/src/details/Scene.h

@@ -84,8 +84,8 @@ public:
     enum {
         RENDERABLE_INSTANCE,    //  4 | instance of the Renderable component
         WORLD_TRANSFORM,        // 16 | instance of the Transform component
-        VISIBILITY_STATE,       //  2 | visibility data of the component
-        SKINNING_BUFFER,        //  8 | bones uniform buffer handle, offset
+        VISIBILITY_STATE,       //  1 | visibility data of the component
+        SKINNING_BUFFER,        //  8 | bones uniform buffer handle, count, offset
         MORPHING_BUFFER,        // 16 | weights uniform buffer handle, count, morph targets
         WORLD_AABB_CENTER,      // 12 | world-space bounding box center of the renderable
         VISIBLE_MASK,           //  2 | each bit represents a visibility in a pass

filament/src/details/View.cpp

@@ -17,6 +17,7 @@
 #include "details/View.h"
 
 #include "Culler.h"
+#include "DFG.h"
 #include "Froxelizer.h"
 #include "RenderPrimitive.h"
 #include "ResourceAllocator.h"
@@ -886,7 +887,7 @@ void FView::updatePrimitivesLod(FEngine& engine, const CameraInfo&,
 FrameGraphId<FrameGraphTexture> FView::renderShadowMaps(FrameGraph& fg, FEngine& engine,
         FEngine::DriverApi& driver,
         RenderPass const& pass) noexcept {
-    return mShadowMapManager.render(fg, engine, pass, *this);
+    return mShadowMapManager.render(fg, engine, driver, pass, *this);
 }
 
 void FView::commitFrameHistory(FEngine& engine) noexcept {
@@ -953,10 +954,6 @@ void FView::setScreenSpaceReflectionsOptions(ScreenSpaceReflectionsOptions optio
     mScreenSpaceReflectionsOptions = options;
 }
 
-void FView::setGuardBandOptions(GuardBandOptions options) noexcept {
-    mGuardBandOptions = options;
-}
-
 void FView::setAmbientOcclusionOptions(AmbientOcclusionOptions options) noexcept {
     options.radius = math::max(0.0f, options.radius);
     options.power = std::max(0.0f, options.power);

filament/src/details/View.h

@@ -73,7 +73,40 @@ class FMaterialInstance;
 class FRenderer;
 class FScene;
 
+// The value of the 'VISIBLE_MASK' after culling. Each bit represents visibility in a frustum
+// (either camera or light).
+//
+//                                    1
+// bits                               5 ... 7 6 5 4 3 2 1 0
+// +------------------------------------------------------+
+// VISIBLE_RENDERABLE                                     X
+// VISIBLE_DIR_SHADOW_RENDERABLE                        X
+// VISIBLE_SPOT_SHADOW_RENDERABLE_0                   X
+// VISIBLE_SPOT_SHADOW_RENDERABLE_1                 X
+// ...
+
+// A "shadow renderable" is a renderable rendered to the shadow map during a shadow pass:
+// PCF shadows: only shadow casters
+// VSM shadows: both shadow casters and shadow receivers
+
+static constexpr size_t VISIBLE_RENDERABLE_BIT = 0u;
+static constexpr size_t VISIBLE_DIR_SHADOW_RENDERABLE_BIT = 1u;
+static constexpr size_t VISIBLE_SPOT_SHADOW_RENDERABLE_N_BIT(size_t n) { return n + 2; }
+
 static constexpr Culler::result_type VISIBLE_RENDERABLE = 1u << VISIBLE_RENDERABLE_BIT;
+static constexpr Culler::result_type VISIBLE_DIR_SHADOW_RENDERABLE = 1u << VISIBLE_DIR_SHADOW_RENDERABLE_BIT;
+static constexpr Culler::result_type VISIBLE_SPOT_SHADOW_RENDERABLE_N(size_t n) {
+    return 1u << VISIBLE_SPOT_SHADOW_RENDERABLE_N_BIT(n);
+}
+
+// ORing of all the VISIBLE_SPOT_SHADOW_RENDERABLE bits
+static constexpr Culler::result_type VISIBLE_SPOT_SHADOW_RENDERABLE =
+        (0xFFu >> (sizeof(Culler::result_type) * 8u - CONFIG_MAX_SHADOW_CASTING_SPOTS)) << 2u;
+
+// Because we're using a uint16_t for the visibility mask, we're limited to 14 spot light shadows.
+// (2 of the bits are used for visible renderables + directional light shadow casters).
+static_assert(CONFIG_MAX_SHADOW_CASTING_SPOTS <= sizeof(Culler::result_type) * 8 - 2,
+        "CONFIG_MAX_SHADOW_CASTING_SPOTS cannot be higher than 14.");
 
 // ------------------------------------------------------------------------------------------------
 
@@ -232,12 +265,6 @@ public:
         return mScreenSpaceReflectionsOptions;
     }
 
-    void setGuardBandOptions(GuardBandOptions options) noexcept;
-
-    GuardBandOptions const& getGuardBandOptions() const noexcept {
-        return mGuardBandOptions;
-    }
-
     void setColorGrading(FColorGrading* colorGrading) noexcept {
         mColorGrading = colorGrading == nullptr ? mDefaultColorGrading : colorGrading;
     }
@@ -500,7 +527,6 @@ private:
     TemporalAntiAliasingOptions mTemporalAntiAliasingOptions;
     MultiSampleAntiAliasingOptions mMultiSampleAntiAliasingOptions;
     ScreenSpaceReflectionsOptions mScreenSpaceReflectionsOptions;
-    GuardBandOptions mGuardBandOptions;
     BlendMode mBlendMode = BlendMode::OPAQUE;
     const FColorGrading* mColorGrading = nullptr;
     const FColorGrading* mDefaultColorGrading = nullptr;

filament/src/fg/FrameGraph.cpp

@@ -215,11 +215,6 @@ void FrameGraph::execute(backend::DriverApi& driver) noexcept {
 
         driver.popGroupMarker();
     }
-
-    // NOTE: in the past we tried to make this flush conditional (b2cdf9f), but this
-    // was problematic on certain devices. b/232224942
-    driver.flush();
-
     driver.popGroupMarker();
 }
 

filament/src/fsr.cpp

@@ -39,15 +39,6 @@ using namespace math;
 #pragma clang diagnostic pop
 
 void FSR_ScalingSetup(FSRUniforms* outUniforms, FSRScalingConfig config) noexcept {
-    // FsrEasu API claims it needs the left-top offset, however that's not true with OpenGL,
-    // in which case it uses the left-bottom offset.
-
-    auto yoffset = config.input.bottom;
-    if (config.backend == backend::Backend::METAL ||
-        config.backend == backend::Backend::VULKAN) {
-        yoffset = config.inputHeight - (config.input.bottom + config.input.height);
-    }
-
     FsrEasuConOffset( outUniforms->EasuCon0.v, outUniforms->EasuCon1.v,
                 outUniforms->EasuCon2.v, outUniforms->EasuCon3.v,
             // Viewport size (top left aligned) in the input image which is to be scaled.
@@ -57,7 +48,10 @@ void FSR_ScalingSetup(FSRUniforms* outUniforms, FSRScalingConfig config) noexcep
             // The output resolution.
             config.outputWidth, config.outputHeight,
             // Input image offset
-            config.input.left, yoffset);
+            config.input.left, config.input.bottom);
+    // FIXME: FsrEasu api claims it needs the top,left offset, but when using that we get
+    //        some bad artifacts. To investigate. For reference:
+    //           top = config.inputHeight - (config.input.bottom + config.input.height)
 }
 
 void FSR_SharpeningSetup(FSRUniforms* outUniforms, FSRSharpeningConfig config) noexcept {

filament/src/fsr.h

@@ -26,7 +26,6 @@
 namespace filament {
 
 struct FSRScalingConfig {
-    filament::backend::Backend backend;
     filament::Viewport input;   // region of source to be scaled
     uint32_t inputWidth;        // width of source
     uint32_t inputHeight;       // height of source

filament/src/materials/antiAliasing/fxaa.mat

@@ -29,7 +29,6 @@ vertex {
 
     void postProcessVertex(inout PostProcessVertexInputs postProcess) {
         postProcess.vertex.xy = materialParams.viewport.xy + postProcess.normalizedUV * materialParams.viewport.zw;
-        postProcess.vertex.xy = uvToRenderTargetUV(postProcess.vertex.xy);
     }
 
 }

filament/src/materials/antiAliasing/taa.mat

@@ -38,7 +38,7 @@ material {
 
 vertex {
     void postProcessVertex(inout PostProcessVertexInputs postProcess) {
-        postProcess.vertex.xy = uvToRenderTargetUV(postProcess.normalizedUV);
+        postProcess.vertex.xy = postProcess.normalizedUV;
     }
 }
 
@@ -204,15 +204,22 @@ void postProcess(inout PostProcessInputs postProcess) {
     float depth = textureLod(materialParams_depth, uv.xy, 0.0).r;
 
 #if HISTORY_REPROJECTION
-    uv.zw = uvToRenderTargetUV(uv.zw);
+#if defined(TARGET_METAL_ENVIRONMENT) || defined(TARGET_VULKAN_ENVIRONMENT)
+    // on metal/vulkan postProcess.normalizedUV has its origin at the left-top, but
+    // our projection matrix screen-space's origin is left-bottom
+    uv.w = 1.0 - uv.w;
+#endif
     // reproject history to current frame
     highp vec4 q = materialParams.reprojection * vec4(uv.zw, depth, 1.0);
     uv.zw = (q.xy * (1.0 / q.w)) * 0.5 + 0.5;
-    uv.zw = uvToRenderTargetUV(uv.zw);
+#if defined(TARGET_METAL_ENVIRONMENT) || defined(TARGET_VULKAN_ENVIRONMENT)
+    // on metal/vulkan after we applied the reprojection, we need to correct our uv back
+    uv.w = 1.0 - uv.w;
+#endif
 #endif
 
     // read center color and history samples
-    vec4 color = textureLod(materialParams_color, uv.xy, 0.0);
+    vec4 color = textureLodOffset(materialParams_color, uv.xy, 0.0, ivec2( 0,  0));
 #if FILTER_HISTORY
     vec4 history = sampleTextureCatmullRom(materialParams_history, uv.zw,
             vec2(textureSize(materialParams_history, 0)));

filament/src/materials/blitLow.mat

@@ -28,7 +28,6 @@ material {
 vertex {
     void postProcessVertex(inout PostProcessVertexInputs postProcess) {
         postProcess.vertex.xy = materialParams.viewport.xy + postProcess.normalizedUV * materialParams.viewport.zw;
-        postProcess.vertex.xy = uvToRenderTargetUV(postProcess.vertex.xy);
     }
 }
 

filament/src/materials/bloom/bloomDownsample.mat

@@ -29,7 +29,7 @@ material {
 
 vertex {
     void postProcessVertex(inout PostProcessVertexInputs postProcess) {
-        postProcess.vertex.xy = uvToRenderTargetUV(postProcess.normalizedUV);
+        postProcess.vertex.xy = postProcess.normalizedUV;
     }
 }
 

filament/src/materials/bloom/bloomUpsample.mat

@@ -26,7 +26,7 @@ material {
 
 vertex {
     void postProcessVertex(inout PostProcessVertexInputs postProcess) {
-        postProcess.vertex.xy = uvToRenderTargetUV(postProcess.normalizedUV);
+        postProcess.vertex.xy = postProcess.normalizedUV;
     }
 }
 

filament/src/materials/colorGrading/colorGrading.mat

@@ -77,7 +77,6 @@ material {
 vertex {
     void postProcessVertex(inout PostProcessVertexInputs postProcess) {
         postProcess.vertex.xy = materialParams.viewport.xy + postProcess.normalizedUV * materialParams.viewport.zw;
-        postProcess.vertex.xy = uvToRenderTargetUV(postProcess.vertex.xy);
     }
 }
 

filament/src/materials/colorGrading/customResolveAsSubpass.mat

@@ -32,6 +32,12 @@ material {
     framebufferFetch: true
 }
 
+vertex {
+    void postProcessVertex(inout PostProcessVertexInputs postProcess) {
+        postProcess.vertex.xy = postProcess.normalizedUV;
+    }
+}
+
 fragment {
 
     void dummy(){}

filament/src/materials/dof/dof.mat

@@ -35,6 +35,11 @@ material {
             type : float,
             name : bokehAngle,
         },
+        {
+           type : float4,
+           name : uvscale,
+           precision: high
+        },
         {
            type : float4,
            name : ringCounts,
@@ -63,7 +68,7 @@ material {
 
 vertex {
     void postProcessVertex(inout PostProcessVertexInputs postProcess) {
-        postProcess.vertex.xy = uvToRenderTargetUV(postProcess.normalizedUV);
+        postProcess.vertex = postProcess.normalizedUV.xyxy * materialParams.uvscale;
     }
 }
 
@@ -573,7 +578,7 @@ void backgroundTile(inout vec4 background, inout float bgOpacity,
 }
 
 void postProcess(inout PostProcessInputs postProcess) {
-    vec2 tiles = textureLod(materialParams_tiles, variable_vertex.xy, 0.0).rg;
+    vec2 tiles = textureLod(materialParams_tiles, variable_vertex.zw, 0.0).rg;
 
     /*
      * Tiles that are neither foreground or background (in focus) can be skipped

filament/src/materials/dof/dofCoc.mat

@@ -20,8 +20,8 @@ material {
             name : cocClamp
         },
         {
-            type : float2,
-            name : texelSize,
+            type : float4,
+            name : uvscale,
             precision: high
         }
     ],
@@ -50,7 +50,7 @@ vertex {
     void dummy(){}
 
     void postProcessVertex(inout PostProcessVertexInputs postProcess) {
-        postProcess.vertex.xy = uvToRenderTargetUV(postProcess.normalizedUV);
+        postProcess.vertex.xy = (postProcess.normalizedUV * materialParams.uvscale.xy) * materialParams.uvscale.zw;
     }
 }
 

filament/src/materials/dof/dofCombine.mat

@@ -20,6 +20,11 @@ material {
             type : sampler2d,
             name : alpha,
             precision: medium
+        },
+        {
+            type : float4,
+            name : uvscale,
+            precision: high
         }
     ],
     variables : [
@@ -32,7 +37,8 @@ material {
 
 vertex {
     void postProcessVertex(inout PostProcessVertexInputs postProcess) {
-        postProcess.vertex.xy = uvToRenderTargetUV(postProcess.normalizedUV);
+        postProcess.vertex.xy = postProcess.normalizedUV;
+        postProcess.vertex.zw = postProcess.normalizedUV * materialParams.uvscale.zw;
     }
 }
 
@@ -43,19 +49,23 @@ fragment {
 void dummy(){}
 
 void postProcess(inout PostProcessInputs postProcess) {
-    highp vec2 uv = variable_vertex.xy;
-    vec2 tiles = textureLod(materialParams_tiles, uv, 0.0).rg;
+    highp vec2 uv2 = variable_vertex.zw;
+    vec2 tiles = textureLod(materialParams_tiles, uv2, 0.0).rg;
 
     vec4 color;
     if (isTrivialTile(tiles)) {
-        color = textureLod(materialParams_color, uv, 0.0);
+        highp vec2 uv0 = variable_vertex.xy;
+        color = textureLod(materialParams_color, uv0, 0.0);
     } else if (isFastTile(tiles) && (abs(tiles.r) >= MAX_IN_FOCUS_COC + 1.0)) {
         // we can only handle fast tiles that don't have an alpha channel here
-        color = textureLod(materialParams_dof, uv, 0.0);
+        highp vec2 uv1 = variable_vertex.xy * materialParams.uvscale.xy;
+        color = textureLod(materialParams_dof, uv1, 0.0);
     } else {
-        float alpha = textureLod(materialParams_alpha, uv, 0.0).r;
-        vec4 dof = textureLod(materialParams_dof, uv, 0.0);
-        color = textureLod(materialParams_color, uv, 0.0);
+        highp vec2 uv0 = variable_vertex.xy;
+        highp vec2 uv1 = variable_vertex.xy * materialParams.uvscale.xy;
+        float alpha = textureLod(materialParams_alpha, uv1, 0.0).r;
+        vec4 dof = textureLod(materialParams_dof, uv1, 0.0);
+        color = textureLod(materialParams_color, uv0, 0.0);
         color = dof + (1.0 - alpha) * color;
     }
 

filament/src/materials/dof/dofDilate.mat

@@ -17,7 +17,7 @@ material {
 
 vertex {
     void postProcessVertex(inout PostProcessVertexInputs postProcess) {
-        postProcess.vertex.xy = uvToRenderTargetUV(postProcess.normalizedUV);
+        postProcess.vertex.xy = postProcess.normalizedUV;
     }
 }
 

filament/src/materials/dof/dofDownsample.mat

@@ -20,8 +20,8 @@ material {
             name : cocClamp
         },
         {
-            type : float2,
-            name : texelSize,
+            type : float4,
+            name : uvscale,
             precision: high
         }
     ],
@@ -50,21 +50,7 @@ vertex {
     void dummy(){}
 
     void postProcessVertex(inout PostProcessVertexInputs postProcess) {
-        // It works to fix-up the UV *before* applying the offsets because materialParams.texelSize
-        // corresponds to one texel.
-        // on GL:           v0 = v - 0.5 * texel                                                (1)
-        //                  v1 = v - 0.5 * texel + 1 * texel
-        //                     = v + 0.5 * texel                                                (2)
-        // on MTL/VK:       v0 = 1 - ((1 - v) - 0.5 * texel)
-        //                     = v + 0.5 * texel                                                (3)
-        //                  v1 = 1 - ((1 - v) - 0.5 * texel + 1 * texel)
-        //                     = v - 0.5 * texel                                                (4)
-        // Because the all the operations on the 4 texels in the fragment shader don't depend on
-        // the order they're fetched, it all works out.
-        postProcess.normalizedUV = uvToRenderTargetUV(postProcess.normalizedUV);
-
-        // offset UVs by -0.5 texel because we use textureLodOffset() in the fragment shader
-        postProcess.vertex.xy = postProcess.normalizedUV - 0.5 * materialParams.texelSize;
+        postProcess.vertex.xy = (postProcess.normalizedUV * materialParams.uvscale.xy * 2.0 - 0.5) * materialParams.uvscale.zw;
     }
 }
 

filament/src/materials/dof/dofMedian.mat

@@ -15,6 +15,11 @@ material {
             type : sampler2d,
             name : tiles,
             precision: medium
+        },
+        {
+            type : float2,
+            name : uvscale,
+            precision: high
         }
     ],
     variables : [
@@ -27,7 +32,8 @@ material {
 
 vertex {
     void postProcessVertex(inout PostProcessVertexInputs postProcess) {
-        postProcess.vertex.xy = uvToRenderTargetUV(postProcess.normalizedUV);
+        postProcess.vertex.xy = postProcess.normalizedUV;
+        postProcess.vertex.zw = postProcess.vertex.xy * materialParams.uvscale;
     }
 }
 
@@ -115,7 +121,7 @@ vec4 median(ivec2 ij) {
 }
 
 void postProcess(inout PostProcessInputs postProcess) {
-    vec2 tiles = textureLod(materialParams_tiles, variable_vertex.xy, 0.0).rg;
+    vec2 tiles = textureLod(materialParams_tiles, variable_vertex.zw, 0.0).rg;
     vec4 dof = vec4(0.0);
 
     if (!isTrivialTile(tiles)) {

filament/src/materials/dof/dofMipmap.mat

@@ -21,7 +21,7 @@ material {
         },
         {
             type : float2,
-            name : texelSize
+            name : pixelSize
         }
     ],
     variables : [
@@ -46,9 +46,7 @@ material {
 
 vertex {
     void postProcessVertex(inout PostProcessVertexInputs postProcess) {
-        // see explanation in dofDownSample.mat
-        postProcess.normalizedUV = uvToRenderTargetUV(postProcess.normalizedUV);
-        postProcess.vertex.xy = postProcess.normalizedUV - 0.5 * materialParams.texelSize;
+        postProcess.vertex.xy = postProcess.normalizedUV - materialParams.pixelSize * 0.5;
     }
 }
 

filament/src/materials/dof/dofTiles.mat

@@ -7,8 +7,8 @@ material {
             precision: medium
         },
         {
-            type : float2,
-            name : texelSize,
+            type : float4,
+            name : uvscale,
             precision: high
         }
     ],
@@ -22,9 +22,7 @@ material {
 
 vertex {
     void postProcessVertex(inout PostProcessVertexInputs postProcess) {
-        // see explanation in dofDownSample.mat
-        postProcess.normalizedUV = uvToRenderTargetUV(postProcess.normalizedUV);
-        postProcess.vertex.xy = postProcess.normalizedUV - 0.5 * materialParams.texelSize;
+        postProcess.vertex.xy = (postProcess.normalizedUV * materialParams.uvscale.xy * 2.0 - 0.5) * materialParams.uvscale.zw;
     }
 }
 

filament/src/materials/dof/dofTilesSwizzle.mat

@@ -7,8 +7,8 @@ material {
             precision: medium
         },
         {
-            type : float2,
-            name : texelSize,
+            type : float4,
+            name : uvscale,
             precision: high
         }
     ],
@@ -22,9 +22,7 @@ material {
 
 vertex {
     void postProcessVertex(inout PostProcessVertexInputs postProcess) {
-        // see explanation in dofDownSample.mat
-        postProcess.normalizedUV = uvToRenderTargetUV(postProcess.normalizedUV);
-        postProcess.vertex.xy = postProcess.normalizedUV - 0.5 * materialParams.texelSize;
+        postProcess.vertex.xy = (postProcess.normalizedUV * materialParams.uvscale.xy * 2.0 - 0.5) * materialParams.uvscale.zw;
     }
 }
 

filament/src/materials/flare/flare.mat

@@ -68,7 +68,13 @@ vertex {
 void dummy(){ }
 
 void postProcessVertex(inout PostProcessVertexInputs postProcess) {
-    postProcess.vertex.xy = uvToRenderTargetUV(postProcess.normalizedUV);
+    postProcess.vertex.xy = postProcess.normalizedUV;
+#if defined(TARGET_METAL_ENVIRONMENT) || defined(TARGET_VULKAN_ENVIRONMENT)
+    // On metal/vulkan postProcess.normalizedUV has its origin at the left-top, but we need a
+    // uniform coordinate space. So, we flip the y coordinate so we have bottom-left UV
+    // coordinates across all backends.
+    postProcess.vertex.y = 1.0 - postProcess.vertex.y;
+#endif
 }
 
 }

filament/src/materials/fsr/fsr_easu.mat

@@ -52,7 +52,6 @@ material {
 
 vertex {
     void postProcessVertex(inout PostProcessVertexInputs postProcess) {
-        postProcess.normalizedUV = uvToRenderTargetUV(postProcess.normalizedUV);
         postProcess.vertex.xy = postProcess.normalizedUV * materialParams.resolution.xy;
     }
 }