Merge branch 'rc/1.56.4' into release

README.md

@@ -31,7 +31,7 @@ repositories {
 }
 
 dependencies {
-    implementation 'com.google.android.filament:filament-android:1.56.3'
+    implementation 'com.google.android.filament:filament-android:1.56.4'
 }
 ```
 
@@ -51,7 +51,7 @@ Here are all the libraries available in the group `com.google.android.filament`:
 iOS projects can use CocoaPods to install the latest release:
 
 ```shell
-pod 'Filament', '~> 1.56.3'
+pod 'Filament', '~> 1.56.4'
 ```
 
 ## Documentation

RELEASE_NOTES.md

@@ -7,6 +7,9 @@ A new header is inserted each time a *tag* is created.
 Instead, if you are authoring a PR for the main branch, add your release note to
 [NEW_RELEASE_NOTES.md](./NEW_RELEASE_NOTES.md).
 
+## v1.56.4
+
+
 ## v1.56.3
 
 

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

@@ -38,7 +38,7 @@ Java_com_google_android_filament_EntityManager_nCreateArray(JNIEnv* env, jclass,
     // (which it is), but still.
     em->create((size_t) n, reinterpret_cast<Entity *>(entities));
 
-    env->ReleaseIntArrayElements(entities_, entities, JNI_ABORT);
+    env->ReleaseIntArrayElements(entities_, entities, 0);
 }
 
 extern "C" JNIEXPORT jint JNICALL

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

@@ -167,7 +167,7 @@ Java_com_google_android_filament_IndirectLight_nGetDirectionEstimateStatic(JNIEn
     jfloat *outDirection = env->GetFloatArrayElements(outDirection_, NULL);
     *reinterpret_cast<filament::math::float3*>(outDirection) = IndirectLight::getDirectionEstimate((filament::math::float3*)sh);
     env->ReleaseFloatArrayElements(outDirection_, outDirection, 0);
-    env->ReleaseFloatArrayElements(sh_, sh, JNI_ABORT);
+    env->ReleaseFloatArrayElements(sh_, sh, 0);
 }
 
 extern "C" JNIEXPORT void JNICALL
@@ -178,5 +178,5 @@ Java_com_google_android_filament_IndirectLight_nGetColorEstimateStatic(JNIEnv *e
     *reinterpret_cast<filament::math::float4*>(outColor) =
             IndirectLight::getColorEstimate((filament::math::float3*)sh, math::float3{x, y, z});
     env->ReleaseFloatArrayElements(outColor_, outColor, 0);
-    env->ReleaseFloatArrayElements(sh_, sh, JNI_ABORT);
+    env->ReleaseFloatArrayElements(sh_, sh, 0);
 }

android/gradle.properties

@@ -1,5 +1,5 @@
 GROUP=com.google.android.filament
-VERSION_NAME=1.56.3
+VERSION_NAME=1.56.4
 
 POM_DESCRIPTION=Real-time physically based rendering engine for Android.
 

docs/Materials.html

@@ -1787,7 +1787,12 @@
       declare a variable called <code>eyeDirection</code> you can access it in the fragment shader using
       <code>variable_eyeDirection</code>. In the vertex shader, the interpolant name is simply a member of
       the <code>MaterialVertexInputs</code> structure (<code>material.eyeDirection</code> in your example). Each
-      interpolant is of type <code>float4</code> (<code>vec4</code>) in the shaders.
+      interpolant is of type <code>float4</code> (<code>vec4</code>) in the shaders. By default the precision of the
+      interpolant is <code>highp</code> in <em class="asterisk">both</em> the vertex and fragment shaders.
+      An alternate syntax can be used to specify both the name and precision of the interpolant.
+      In this case the specified precision is used as-is in both fragment and vertex stages, in
+      particular if <code>default</code> is specified the default precision is used is the fragment shader
+      (<code>mediump</code>) and in the vertex shader (<code>highp</code>).
 
 </p></dd></dl><p></p><pre class="listing tilde"><code><span class="line">material {</span>
 <span class="line">    name : Skybox,</span>
@@ -1798,7 +1803,11 @@
 <span class="line">        }</span>
 <span class="line">    ],</span>
 <span class="line">    variables : [</span>
-<span class="line">         eyeDirection</span>
+<span class="line">         eyeDirection,</span>
+<span class="line">         {</span>
+<span class="line">            name : eyeColor,</span>
+<span class="line">            precision : medium</span>
+<span class="line">         }</span>
 <span class="line">    ],</span>
 <span class="line">    vertexDomain : device,</span>
 <span class="line">    depthWrite : <span class="hljs-literal">false</span>,</span>

filament/backend/src/metal/MetalContext.h

@@ -135,6 +135,7 @@ struct MetalContext {
     bool supportsTextureSwizzling = false;
     bool supportsAutoDepthResolve = false;
     bool supportsMemorylessRenderTargets = false;
+    bool supportsDepthClamp = false;
     uint8_t maxColorRenderTargets = 4;
     struct {
         uint8_t common;

filament/backend/src/metal/MetalDriver.mm

@@ -151,6 +151,11 @@
             mContext->highestSupportedGpuFamily.mac   >= 2;     // newer macOS GPUs
     }
 
+    mContext->supportsDepthClamp = false;
+    if (@available(macOS 10.11, iOS 11.0, *)) {
+        mContext->supportsDepthClamp = true;
+    }
+
     // In order to support resolve store action on depth attachment, the GPU needs to support it.
     // Note that support for depth resolve implies support for stencil resolve using .sample0 resolve filter.
     // (Other resolve filters are supported starting .apple5 and .mac2 families).
@@ -1063,7 +1068,7 @@
 }
 
 bool MetalDriver::isDepthClampSupported() {
-    return true;
+    return mContext->supportsDepthClamp;
 }
 
 bool MetalDriver::isWorkaroundNeeded(Workaround workaround) {
@@ -1258,6 +1263,7 @@
     mContext->cullModeState.invalidate();
     mContext->windingState.invalidate();
     mContext->scissorRectState.invalidate();
+    mContext->depthClampState.invalidate();
     mContext->currentPolygonOffset = {0.0f, 0.0f};
 
     mContext->finalizedDescriptorSets.clear();
@@ -1729,10 +1735,12 @@
     }
 
     // depth clip mode
-    MTLDepthClipMode depthClipMode = rs.depthClamp ? MTLDepthClipModeClamp : MTLDepthClipModeClip;
-    mContext->depthClampState.updateState(depthClipMode);
-    if (mContext->depthClampState.stateChanged()) {
-        [mContext->currentRenderPassEncoder setDepthClipMode:depthClipMode];
+    if (mContext->supportsDepthClamp) {
+        MTLDepthClipMode depthClipMode = rs.depthClamp ? MTLDepthClipModeClamp : MTLDepthClipModeClip;
+        mContext->depthClampState.updateState(depthClipMode);
+        if (mContext->depthClampState.stateChanged()) {
+            [mContext->currentRenderPassEncoder setDepthClipMode:depthClipMode];
+        }
     }
 
     // Set the depth-stencil state, if a state change is needed.

filament/backend/src/metal/MetalShaderCompiler.mm

@@ -226,9 +226,11 @@ bool isReady() const noexcept {
             CompilerPriorityQueue const priorityQueue = program.getPriorityQueue();
             mCompilerThreadPool.queue(priorityQueue, token,
                     [this, name, device = mDevice, program = std::move(program), token]() {
-                        MetalFunctionBundle compiledProgram = compileProgram(program, device);
-                        token->set(compiledProgram);
-                        mCallbackManager.put(token->handle);
+                        @autoreleasepool {
+                            MetalFunctionBundle compiledProgram = compileProgram(program, device);
+                            token->set(compiledProgram);
+                            mCallbackManager.put(token->handle);
+                        }
                     });
 
             break;

filament/backend/src/opengl/OpenGLContext.cpp

@@ -500,11 +500,13 @@ void OpenGLContext::initBugs(Bugs* bugs, Extensions const& exts,
             // qualcomm seems to have no problem with this (which is good for us)
             bugs->allow_read_only_ancillary_feedback_loop = true;
 
+#ifndef __EMSCRIPTEN__
             // Older Adreno devices that support ES3.0 only tend to be extremely buggy, so we
             // fall back to ES2.0.
             if (major == 3 && minor == 0) {
                 bugs->force_feature_level0 = true;
             }
+#endif
         } else if (strstr(renderer, "Mali")) {
             // ARM GPU
             bugs->vao_doesnt_store_element_array_buffer_binding = true;

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

@@ -64,10 +64,14 @@ struct OSMesaAPI {
                 break;
             }
         }
-        FILAMENT_CHECK_PRECONDITION(mLib)
-                << "Unable to dlopen libOSMesa to create a software GL context";
+        if (mLib) {
+            OSMesaGetProcAddress = (GetProcAddressFunc) dlsym(mLib, "OSMesaGetProcAddress");
+        } else {
+            OSMesaGetProcAddress = (GetProcAddressFunc) dlsym(RTLD_LOCAL, "OSMesaGetProcAddress");
+        }
 
-        OSMesaGetProcAddress = (GetProcAddressFunc) dlsym(mLib, "OSMesaGetProcAddress");
+        FILAMENT_CHECK_PRECONDITION(OSMesaGetProcAddress)
+                << "Unable to against libOSMesa to create a software GL context";
 
         OSMesaCreateContext = (CreateContextFunc) OSMesaGetProcAddress("OSMesaCreateContext");
         OSMesaDestroyContext =
@@ -76,7 +80,9 @@ struct OSMesaAPI {
     }
 
     ~OSMesaAPI() {
-        dlclose(mLib);
+        if (mLib) {
+            dlclose(mLib);
+        }
     }
 private:
     void* mLib = nullptr;

filament/include/filament/Camera.h

@@ -61,7 +61,7 @@ namespace filament {
  *  filament::Camera* myCamera = engine->createCamera(myCameraEntity);
  *  myCamera->setProjection(45, 16.0/9.0, 0.1, 1.0);
  *  myCamera->lookAt({0, 1.60, 1}, {0, 0, 0});
- *  engine->destroyCameraComponent(myCamera);
+ *  engine->destroyCameraComponent(myCameraEntity);
  * ~~~~~~~~~~~
  *
  *

filament/src/RenderPass.cpp

@@ -864,21 +864,44 @@ backend::Viewport RenderPass::Executor::applyScissorViewport(
         backend::Viewport const& scissorViewport, backend::Viewport const& scissor) noexcept {
     // scissor is set, we need to apply the offset/clip
     // clang vectorizes this!
-    constexpr int32_t maxvali = std::numeric_limits<int32_t>::max();
-    // compute new left/bottom, assume no overflow
-    int32_t const l = scissor.left + scissorViewport.left;
-    int32_t const b = scissor.bottom + scissorViewport.bottom;
-    // compute right/top without overflowing, scissor.width/height guaranteed
-    // to convert to int32
-    int32_t r = (l > maxvali - int32_t(scissor.width)) ? maxvali : l + int32_t(scissor.width);
-    int32_t t = (b > maxvali - int32_t(scissor.height)) ? maxvali : b + int32_t(scissor.height);
-    // clip to the viewport
-    assert_invariant(l == std::max(l, scissorViewport.left));
-    assert_invariant(b == std::max(b, scissorViewport.bottom));
-    r = std::min(r, scissorViewport.left + int32_t(scissorViewport.width));
-    t = std::min(t, scissorViewport.bottom + int32_t(scissorViewport.height));
-    assert_invariant(r >= l && t >= b);
-    return { l, b, uint32_t(r - l), uint32_t(t - b) };
+    constexpr int64_t maxvali = std::numeric_limits<int32_t>::max();
+
+    // we do all the offsetting/clipping math in 64 bits below to avoid overflows
+    struct {
+        int64_t l;
+        int64_t b;
+        int64_t r;
+        int64_t t;
+    } svp, s;
+
+    // convert scissorViewport to {left,bottom,right,top} format
+    svp.l = scissorViewport.left;
+    svp.b = scissorViewport.bottom;
+    svp.r = svp.l + scissorViewport.width;
+    svp.t = svp.b + scissorViewport.height;
+
+    // convert scissor to {left,bottom,right,top} format and offset by scissorViewport's left,bottom
+    s.l = svp.l + scissor.left;
+    s.b = svp.b + scissor.bottom;
+    s.r = s.l + scissor.width;
+    s.t = s.b + scissor.height;
+
+    // clip to the scissorViewport
+    s.l = std::max(s.l, svp.l);
+    s.b = std::max(s.b, svp.b);
+    s.r = std::min(s.r, svp.r);
+    s.t = std::min(s.t, svp.t);
+
+    // clip to positive int32_t
+    s.l = std::max(s.l, int64_t(0));
+    s.b = std::max(s.b, int64_t(0));
+    s.r = std::min(s.r, maxvali);
+    s.t = std::min(s.t, maxvali);
+
+    assert_invariant(s.r >= s.l && s.t >= s.b);
+
+    // convert back to Viewport format
+    return { int32_t(s.l), int32_t(s.b), uint32_t(s.r - s.l), uint32_t(s.t - s.b) };
 }
 
 UTILS_NOINLINE // no need to be inlined

filament/src/details/Engine.cpp

@@ -1159,8 +1159,13 @@ bool FEngine::destroy(const FMaterial* ptr) {
             if (!ASSERT_PRECONDITION_NON_FATAL(pos->second.empty(),
                     "destroying material \"%s\" but %u instances still alive",
                     ptr->getName().c_str(), (*pos).second.size())) {
-                return false;
+                // We return true here, because the material has been destroyed. However, this
+                // leaks this material's ResourceList<FMaterialInstance> until the engine
+                // is shut down. Note that it's still possible for the MaterialInstances to be
+                // destroyed manually.
+                return true;
             }
+            mMaterialInstances.erase(pos);
         }
     }
     return success;

filament/src/details/Renderer.cpp

@@ -963,6 +963,11 @@ void FRenderer::renderJob(RootArenaScope& rootArenaScope, FView& view) {
                             uint32_t(float(xvp.width ) * aoOptions.resolution),
                             uint32_t(float(xvp.height) * aoOptions.resolution)});
 
+                // this needs to reset the sampler that are only set in RendererUtils::colorPass(), because
+                // this descriptor-set is also used for ssr/picking/structure and these could be stale
+                // it would be better to use a separate desriptor-set for those two cases so that we don't
+                // have to do this
+                view.unbindSamplers(driver);
                 view.commitUniformsAndSamplers(driver);
             });
 

filament/src/details/View.cpp

@@ -912,6 +912,10 @@ void FView::commitUniformsAndSamplers(DriverApi& driver) const noexcept {
     mColorPassDescriptorSet.commit(driver);
 }
 
+void FView::unbindSamplers(DriverApi& driver) noexcept {
+    mColorPassDescriptorSet.unbindSamplers(driver);
+}
+
 void FView::commitFroxels(DriverApi& driverApi) const noexcept {
     if (mHasDynamicLighting) {
         mFroxelizer.commit(driverApi);

filament/src/details/View.h

@@ -174,6 +174,7 @@ class FView : public View {
 
     void commitFroxels(backend::DriverApi& driverApi) const noexcept;
     void commitUniformsAndSamplers(backend::DriverApi& driver) const noexcept;
+    void unbindSamplers(backend::DriverApi& driver) noexcept;
 
     utils::JobSystem::Job* getFroxelizerSync() const noexcept { return mFroxelizerSync; }
     void setFroxelizerSync(utils::JobSystem::Job* sync) noexcept { mFroxelizerSync = sync; }

filament/src/ds/ColorPassDescriptorSet.cpp

@@ -541,6 +541,17 @@ void ColorPassDescriptorSet::commit(backend::DriverApi& driver) noexcept {
     }
 }
 
+void ColorPassDescriptorSet::unbindSamplers(DriverApi&) noexcept {
+    // this needs to reset the sampler that are only set in RendererUtils::colorPass(), because
+    // this descriptor-set is also used for ssr/picking/structure and these could be stale
+    // it would be better to use a separate descriptor-set for those two cases so that we don't
+    // have to do this
+    setSampler(+PerViewBindingPoints::STRUCTURE, {}, {});
+    setSampler(+PerViewBindingPoints::SHADOW_MAP, {}, {});
+    setSampler(+PerViewBindingPoints::SSAO, {}, {});
+    setSampler(+PerViewBindingPoints::SSR, {}, {});
+}
+
 void ColorPassDescriptorSet::setSampler(backend::descriptor_binding_t binding,
         TextureHandle th, SamplerParams params) noexcept {
     for (size_t i = 0; i < DESCRIPTOR_LAYOUT_COUNT; i++) {

filament/src/ds/ColorPassDescriptorSet.h

@@ -158,6 +158,8 @@ class ColorPassDescriptorSet {
     // update local data into GPU UBO
     void commit(backend::DriverApi& driver) noexcept;
 
+    void unbindSamplers(backend::DriverApi& driver) noexcept;
+
     // bind this UBO
     void bind(backend::DriverApi& driver, uint8_t index) const noexcept {
         mDescriptorSet[index].bind(driver, DescriptorSetBindingPoints::PER_VIEW);

ios/CocoaPods/Filament.podspec

@@ -1,12 +1,12 @@
 Pod::Spec.new do |spec|
   spec.name = "Filament"
-  spec.version = "1.56.3"
+  spec.version = "1.56.4"
   spec.license = { :type => "Apache 2.0", :file => "LICENSE" }
   spec.homepage = "https://google.github.io/filament"
   spec.authors = "Google LLC."
   spec.summary = "Filament is a real-time physically based rendering engine for Android, iOS, Windows, Linux, macOS, and WASM/WebGL."
   spec.platform = :ios, "11.0"
-  spec.source = { :http => "https://github.com/google/filament/releases/download/v1.56.3/filament-v1.56.3-ios.tgz" }
+  spec.source = { :http => "https://github.com/google/filament/releases/download/v1.56.4/filament-v1.56.4-ios.tgz" }
 
   # Fix linking error with Xcode 12; we do not yet support the simulator on Apple silicon.
   spec.pod_target_xcconfig = {

libs/bluegl/src/BlueGLLinuxOSMesa.cpp

@@ -38,12 +38,13 @@ bool initBinder() {
         }
     }
     if (!g_driver.library) {
-        return false;
+        // The library has been linked explicitly during compile.
+        g_driver.OSMesaGetProcAddress = (ProcAddressFunc) dlsym(RTLD_LOCAL, "OSMesaGetProcAddress");
+    } else {
+        g_driver.OSMesaGetProcAddress =
+                (ProcAddressFunc) dlsym(g_driver.library, "OSMesaGetProcAddress");
     }
 
-    g_driver.OSMesaGetProcAddress = (ProcAddressFunc)
-            dlsym(g_driver.library, "OSMesaGetProcAddress");
-
     return g_driver.OSMesaGetProcAddress;
 }
 
@@ -52,7 +53,9 @@ void* loadFunction(const char* name) {
 }
 
 void shutdownBinder() {
-    dlclose(g_driver.library);
+    if (g_driver.library) {
+        dlclose(g_driver.library);
+    }
     memset(&g_driver, 0, sizeof(g_driver));
 }
 

web/filament-js/package.json

@@ -1,6 +1,6 @@
 {
   "name": "filament",
-  "version": "1.56.3",
+  "version": "1.56.4",
   "description": "Real-time physically based rendering engine",
   "main": "filament.js",
   "module": "filament.js",