Fix Android build.

NEW_RELEASE_NOTES.md

@@ -8,5 +8,3 @@ appropriate header in [RELEASE_NOTES.md](./RELEASE_NOTES.md).
 ## Release notes for next branch cut
 
 - engine: add `View::getLastDynamicResolutionScale()` (b/457753622)
-
-- materials: Make Material Instances' UBO descriptor use dynamic offsets. [⚠️ **Recompile Materials**]

android/filament-utils-android/src/main/java/com/google/android/filament/utils/Utils.kt

@@ -22,7 +22,6 @@ object Utils {
     /**
      * Initializes the utils JNI layer. Must be called before using any utils functionality.
      */
-    @JvmStatic
     fun init() {
         // Load Filament first to ensure that the NioUtils Java class is available in the JNIEnv.
         Filament.init()

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

@@ -17,20 +17,12 @@ filamentTools {
 }
 
 // don't forget to update MainACtivity.kt when/if changing this.
-tasks.register('copyDamagedHelmetGltf', Copy) {
-    from file("../../../third_party/models/DamagedHelmet/DamagedHelmet.glb")
-    into file("src/main/assets/models")
-    rename {String fileName -> "helmet.glb"}
-}
-
-// don't forget to update MainACtivity.kt when/if changing this.
-tasks.register('copyBusterGltf', Copy) {
+tasks.register('copyMesh', Copy) {
     from "../../../third_party/models/BusterDrone"
     into "src/main/assets/models"
 }
 
-preBuild.dependsOn copyDamagedHelmetGltf
-preBuild.dependsOn copyBusterGltf
+preBuild.dependsOn copyMesh
 
 clean.doFirst {
     delete "src/main/assets"

filament/backend/CMakeLists.txt

@@ -214,6 +214,14 @@ if (FILAMENT_SUPPORTS_VULKAN)
             src/vulkan/VulkanMemory.h
             src/vulkan/VulkanPipelineCache.cpp
             src/vulkan/VulkanPipelineCache.h
+            src/vulkan/VulkanPipelineStateSerializer.cpp
+            src/vulkan/VulkanPipelineStateSerializer.h
+            src/vulkan/VulkanSamplerStateSerializer.h
+            src/vulkan/VulkanSamplerStateSerializer.cpp
+            src/vulkan/VulkanRenderPassStateSerializer.h
+            src/vulkan/VulkanRenderPassStateSerializer.cpp
+            src/vulkan/VulkanDescriptorSetLayoutSerializer.h
+            src/vulkan/VulkanDescriptorSetLayoutSerializer.cpp
             src/vulkan/VulkanPipelineLayoutCache.cpp
             src/vulkan/VulkanPipelineLayoutCache.h
             src/vulkan/VulkanQueryManager.cpp

filament/backend/include/backend/platforms/VulkanPlatform.h

@@ -444,11 +444,6 @@ protected:
 
     virtual VkExternalFenceHandleTypeFlagBits getFenceExportFlags() const noexcept;
 
-    /**
-     * Query if transient attachments are supported by the backend.
-     */
-    bool isTransientAttachmentSupported() const noexcept;
-
 private:
     friend struct VulkanPlatformPrivate;
 };

filament/backend/src/vulkan/VulkanConstants.h

@@ -214,10 +214,4 @@ constexpr static const int FVK_MAX_PIPELINE_AGE = FVK_MAX_COMMAND_BUFFERS;
 // destroying any unused pipeline object.
 static_assert(FVK_MAX_PIPELINE_AGE >= FVK_MAX_COMMAND_BUFFERS);
 
-// Indicates if the backend must be setup to allow doing a RenderDoc capture.
-//
-// If this is true, the features not supported by RenderDoc must be disabled, otherwise
-// when using RenderDoc the application will crash or will fail to do a capture.
-constexpr static const int FVK_RENDERDOC_CAPTURE_MODE = false;
-
 #endif

filament/backend/src/vulkan/VulkanContext.h

@@ -154,10 +154,6 @@ public:
         return mStagingBufferBypassEnabled;
     }
 
-    inline bool pipelineCreationFeedbackSupported() const noexcept {
-        return mPipelineCreationFeedbackSupported;
-    }
-
 private:
     VkPhysicalDeviceMemoryProperties mMemoryProperties = {};
     VkPhysicalDeviceProperties2 mPhysicalDeviceProperties = {
@@ -185,7 +181,6 @@ private:
     bool mProtectedMemorySupported = false;
     bool mIsUnifiedMemoryArchitecture = false;
     bool mStagingBufferBypassEnabled = false;
-    bool mPipelineCreationFeedbackSupported = false;
 
     fvkutils::VkFormatList mDepthStencilFormats;
     fvkutils::VkFormatList mBlittableDepthStencilFormats;

filament/backend/src/vulkan/VulkanDescriptorSetLayoutCache.cpp

@@ -15,6 +15,7 @@
  */
 
 #include "VulkanDescriptorSetLayoutCache.h"
+#include "VulkanDescriptorSetLayoutSerializer.h"
 
 #include "VulkanHandles.h"
 
@@ -100,22 +101,32 @@ uint32_t appendSamplerBindings(VkDescriptorSetLayoutBinding* toBind,
     return count;
 }
 
-uint64_t computeImmutableSamplerHash(utils::FixedCapacityVector<VkSampler> const& samplers) {
+uint64_t computeImmutableSamplerHash(utils::FixedCapacityVector<VkSampler> const& samplers, VulkanSamplerCache* cache) {
     size_t const size = samplers.size();
     if (size == 0) {
         return 0;
-    } else if (size == 1) {
-        return (uint64_t) samplers[0];
     }
-    return utils::hash::murmur3((uint32_t*) samplers.data(), samplers.size() * 2, 0);
+
+    utils::FixedCapacityVector<uint32_t> samplerHash;
+    if (size > 0) {
+        samplerHash.reserve(MAX_SAMPLER_COUNT);
+        for (auto sampler : samplers) {
+            samplerHash.push_back(cache->getKey(sampler));
+        }
+    }
+
+    if (size == 1) {
+        return (uint64_t) samplerHash[0];
+    }
+    return utils::hash::murmur3((uint32_t*) samplerHash.data(), samplerHash.size() * 2, 0);
 }
 
 } // anonymous namespace
 
 VulkanDescriptorSetLayoutCache::VulkanDescriptorSetLayoutCache(VkDevice device,
-        fvkmemory::ResourceManager* resourceManager)
+        fvkmemory::ResourceManager* resourceManager, VulkanSamplerCache* cache)
     : mDevice(device),
-      mResourceManager(resourceManager) {}
+      mResourceManager(resourceManager), mSamplerCache(cache) {}
 
 VulkanDescriptorSetLayoutCache::~VulkanDescriptorSetLayoutCache() = default;
 
@@ -131,7 +142,7 @@ VkDescriptorSetLayout VulkanDescriptorSetLayoutCache::getVkLayout(
         utils::FixedCapacityVector<VkSampler> immutableSamplers) {
     LayoutKey key = {
         .bitmask = bitmasks,
-        .immutableSamplerHash = computeImmutableSamplerHash(immutableSamplers),
+        .immutableSamplerHash = computeImmutableSamplerHash(immutableSamplers, mSamplerCache),
     };
     if (auto itr = mVkLayouts.find(key); itr != mVkLayouts.end()) {
         return itr->second;
@@ -153,9 +164,20 @@ VkDescriptorSetLayout VulkanDescriptorSetLayoutCache::getVkLayout(
         .bindingCount = count,
         .pBindings = toBind,
     };
+    LayoutKeyHashFn hashFunc;
+    uint32_t hashedKey = hashFunc(key);
+    utils::FixedCapacityVector<uint32_t> immutableSamplersHashes;
+    immutableSamplersHashes.reserve(MAX_SAMPLER_COUNT);
+    for (auto sampler: immutableSamplers) {
+        immutableSamplersHashes.push_back(mSamplerCache->getKey(sampler));
+    }
+    VulkanDescriptorSetLayoutSerializer setSer(dlinfo, immutableSamplersHashes, hashedKey);
+
     VkDescriptorSetLayout vklayout;
     vkCreateDescriptorSetLayout(mDevice, &dlinfo, VKALLOC, &vklayout);
     mVkLayouts[key] = vklayout;
+
+    mLayoutToKey[vklayout] = hashedKey;
     return vklayout;
 }
 

filament/backend/src/vulkan/VulkanDescriptorSetLayoutCache.h

@@ -19,6 +19,8 @@
 
 #include "VulkanHandles.h"
 
+#include "VulkanSamplerCache.h"
+
 #include "vulkan/memory/ResourcePointer.h"
 
 #include <backend/DriverEnums.h>
@@ -35,7 +37,8 @@ namespace filament::backend {
 
 class VulkanDescriptorSetLayoutCache {
 public:
-    VulkanDescriptorSetLayoutCache(VkDevice device, fvkmemory::ResourceManager* resourceManager);
+    VulkanDescriptorSetLayoutCache(VkDevice device, fvkmemory::ResourceManager* resourceManager,
+            VulkanSamplerCache* cache);
     ~VulkanDescriptorSetLayoutCache();
 
     void terminate() noexcept;
@@ -48,10 +51,19 @@ public:
     VkDescriptorSetLayout getVkLayout(VulkanDescriptorSetLayout::Bitmask const& bitmasks,
             fvkutils::SamplerBitmask externalSamplers,
             utils::FixedCapacityVector<VkSampler> immutableSamplers = {});
+    uint32_t getKey(VkDescriptorSetLayout layout) {
+        uint32_t key = 0;
+        auto iter = mLayoutToKey.find(layout);
+        if (iter != mLayoutToKey.end()) {
+            key = iter->second;
+        }
+        return key;
+    }
 
 private:
     VkDevice mDevice;
     fvkmemory::ResourceManager* mResourceManager;
+    VulkanSamplerCache* mSamplerCache;
 
     struct LayoutKey {
         // this describes the layout using bitset.
@@ -69,6 +81,7 @@ private:
     };
 
     tsl::robin_map<LayoutKey, VkDescriptorSetLayout, LayoutKeyHashFn, LayoutKeyEqual> mVkLayouts;
+    std::map<VkDescriptorSetLayout, uint32_t> mLayoutToKey;
 };
 
 } // namespace filament::backend

filament/backend/src/vulkan/VulkanDescriptorSetLayoutSerializer.cpp

@@ -0,0 +1,87 @@
+#include "VulkanDescriptorSetLayoutSerializer.h"
+#include <fstream>
+#include <sstream>
+#include <utils/Hash.h>
+
+namespace filament::backend {
+VulkanDescriptorSetLayoutSerializer::VulkanDescriptorSetLayoutSerializer(
+        const VkDescriptorSetLayoutCreateInfo& info,
+        utils::FixedCapacityVector<uint32_t> immutableSamplers, uint32_t hash) {
+    std::stringstream filename;
+    filename << "descriptor_set_layout_" << hash << ".json";
+
+    std::ofstream file(filename.str());
+    if (file.is_open()) {
+        std::stringstream buffer;
+        buffer << "[" << std::endl;
+        for (uint32_t i = 0; i < info.bindingCount; ++i) {
+            buffer << "{" << std::endl;
+            buffer << "\"binding\":" << info.pBindings[i].binding << "," << std::endl;
+            buffer << "\"descriptor_type\":" << info.pBindings[i].descriptorType << "," << std::endl;
+            buffer << "\"stage_flags\":" << info.pBindings[i].stageFlags;
+            if (immutableSamplers.size()) {
+                buffer << "," << std::endl;
+                buffer << "\"immutable_samplers\":[" << std::endl;
+                for (uint32_t i = 0; i < immutableSamplers.size(); ++i) {
+                    buffer << immutableSamplers[i];
+                    if (i < immutableSamplers.size() - 1) {
+                        buffer << ",";
+                    }
+                }
+                buffer << "]";
+            }
+            buffer << std::endl;
+            buffer << "}";
+            if (i < info.bindingCount - 1) {
+                buffer << ",";
+            }
+            buffer << std::endl;
+        }
+        buffer << "]" << std::endl;
+
+        file << buffer.str();
+    }
+    file.close();
+}
+
+VulkanPipelineLayoutSerializer::VulkanPipelineLayoutSerializer(
+        const VkPipelineLayoutCreateInfo& info, VulkanDescriptorSetLayoutCache* cache,
+        uint32_t hash) {
+    std::stringstream filename;
+    filename << "pipeline_layout_" << hash << ".json";
+    std::ofstream file(filename.str());
+    if (file.is_open()) {
+        std::stringstream buffer;
+        buffer << "{" << std::endl;
+        buffer << "\"flags\":" << info.flags << "," << std::endl;
+
+        buffer << "\"layouts\":[" << std::endl;
+        for (uint32_t i = 0; i < info.setLayoutCount; ++i) {
+            buffer << cache->getKey(info.pSetLayouts[i]);
+            if (i < info.setLayoutCount - 1) {
+                buffer << ",";
+            }
+        }
+        buffer << "]," << std::endl;
+
+        buffer << "\"push_constants\":[" << std::endl;
+        for (uint32_t i = 0; i < info.pushConstantRangeCount; ++i) {
+            buffer << "{" << std::endl;
+            buffer << "\"stage_flags\":" << info.pPushConstantRanges[i].stageFlags << "," << std::endl;
+            buffer << "\"offset\":" << info.pPushConstantRanges[i].offset << ","
+                   << std::endl;
+            buffer << "\"size\":" << info.pPushConstantRanges[i].size
+                   << std::endl;
+            buffer << "}" << std::endl;
+            if (i < info.pushConstantRangeCount - 1) {
+                buffer << ",";
+            }
+        }
+        buffer << "]" << std::endl;
+
+        buffer << "}" << std::endl;
+        file << buffer.str();
+    }
+    file.close();
+ }
+} // namespace filament::backend

filament/backend/src/vulkan/VulkanDescriptorSetLayoutSerializer.h

@@ -0,0 +1,53 @@
+/*
+ * Copyright (C) 2022 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef TNT_FILAMENT_BACKEND_VULKANDESCRIPTORSETLAYOUTSERIALIZER_H
+#define TNT_FILAMENT_BACKEND_VULKANDESCRIPTORSETLAYOUTSERIALIZER_H
+
+#include "VulkanPipelineLayoutCache.h"
+#include "VulkanYcbcrConversionCache.h"
+#include <backend/DriverEnums.h>
+
+#include <bluevk/BlueVK.h>
+#include <iostream>
+#include <sstream>
+#include <vector>
+
+namespace filament::backend {
+
+class VulkanDescriptorSetLayoutSerializer {
+public:
+    VulkanDescriptorSetLayoutSerializer(VulkanDescriptorSetLayoutSerializer const&) = delete;
+    VulkanDescriptorSetLayoutSerializer& operator=(
+            VulkanDescriptorSetLayoutSerializer const&) = delete;
+
+    VulkanDescriptorSetLayoutSerializer(const VkDescriptorSetLayoutCreateInfo& info,
+            utils::FixedCapacityVector<uint32_t> immutableSamplers,
+            uint32_t hash);
+};
+
+class VulkanPipelineLayoutSerializer {
+public:
+    VulkanPipelineLayoutSerializer(VulkanPipelineLayoutSerializer const&) = delete;
+    VulkanPipelineLayoutSerializer& operator=(VulkanPipelineLayoutSerializer const&) = delete;
+
+    VulkanPipelineLayoutSerializer(const VkPipelineLayoutCreateInfo& layout,
+            VulkanDescriptorSetLayoutCache* cache, uint32_t hash);
+};
+
+} // namespace filament::backend
+
+#endif // TNT_FILAMENT_BACKEND_VULKANDESCRIPTORSETLAYOUTSERIALIZER_H

filament/backend/src/vulkan/VulkanDriver.cpp

@@ -230,16 +230,16 @@ VulkanDriver::VulkanDriver(VulkanPlatform* platform, VulkanContext& context,
       mCommands(mPlatform->getDevice(), mPlatform->getGraphicsQueue(),
               mPlatform->getGraphicsQueueFamilyIndex(), mPlatform->getProtectedGraphicsQueue(),
               mPlatform->getProtectedGraphicsQueueFamilyIndex(), mContext, &mSemaphoreManager),
-      mPipelineLayoutCache(mPlatform->getDevice()),
-      mPipelineCache(mPlatform->getDevice(), mContext),
+      mPipelineCache(mPlatform->getDevice()),
       mStagePool(mAllocator, &mResourceManager, &mCommands, &mContext.getPhysicalDeviceLimits()),
       mBufferCache(mContext, mResourceManager, mAllocator),
       mFramebufferCache(mPlatform->getDevice()),
       mYcbcrConversionCache(mPlatform->getDevice()),
-      mSamplerCache(mPlatform->getDevice()),
+      mSamplerCache(mPlatform->getDevice(), &mYcbcrConversionCache),
       mBlitter(mPlatform->getPhysicalDevice(), &mCommands),
       mReadPixels(mPlatform->getDevice()),
-      mDescriptorSetLayoutCache(mPlatform->getDevice(), &mResourceManager),
+      mDescriptorSetLayoutCache(mPlatform->getDevice(), &mResourceManager, &mSamplerCache),
+      mPipelineLayoutCache(mPlatform->getDevice(), &mDescriptorSetLayoutCache),
       mDescriptorSetCache(mPlatform->getDevice(), &mResourceManager),
       mQueryManager(mPlatform->getDevice()),
       mExternalImageManager(platform, &mSamplerCache, &mYcbcrConversionCache, &mDescriptorSetCache,
@@ -1666,7 +1666,7 @@ void VulkanDriver::beginRenderPass(Handle<HwRenderTarget> rth, const RenderPassP
     rpkey.subpassMask = uint8_t(params.subpassMask);
 
     VkRenderPass renderPass = mFramebufferCache.getRenderPass(rpkey);
-    mPipelineCache.bindRenderPass(renderPass, 0);
+    mPipelineCache.bindRenderPass(renderPass, mFramebufferCache.getRenderPassKey(renderPass), 0);
 
     // Create the VkFramebuffer or fetch it from cache.
     VulkanFboCache::FboKey fbkey = rt->getFboKey();
@@ -1785,6 +1785,7 @@ void VulkanDriver::nextSubpass(int) {
             VK_SUBPASS_CONTENTS_INLINE);
 
     mPipelineCache.bindRenderPass(mCurrentRenderPass.renderPass,
+            mFramebufferCache.getRenderPassKey(mCurrentRenderPass.renderPass),
             ++mCurrentRenderPass.currentSubpass);
 
     if (mCurrentRenderPass.params.subpassMask & 0x1) {
@@ -2160,7 +2161,8 @@ void VulkanDriver::bindPipelineImpl(PipelineState const& pipelineState,
     mPipelineState.pipelineLayout = pipelineLayout;
     mPipelineState.descriptorSetMask = descriptorSetMask;
 
-    mPipelineCache.bindLayout(pipelineLayout);
+    auto key = mPipelineLayoutCache.getKey(pipelineLayout);
+    mPipelineCache.bindLayout(pipelineLayout, key);
     mPipelineCache.bindPipeline(mCurrentRenderPass.commandBuffer);
 }
 

filament/backend/src/vulkan/VulkanDriver.h

@@ -148,7 +148,6 @@ private:
 
     VulkanSemaphoreManager mSemaphoreManager;
     VulkanCommands mCommands;
-    VulkanPipelineLayoutCache mPipelineLayoutCache;
     VulkanPipelineCache mPipelineCache;
     VulkanStagePool mStagePool;
     VulkanBufferCache mBufferCache;
@@ -158,6 +157,7 @@ private:
     VulkanBlitter mBlitter;
     VulkanReadPixels mReadPixels;
     VulkanDescriptorSetLayoutCache mDescriptorSetLayoutCache;
+    VulkanPipelineLayoutCache mPipelineLayoutCache;
     VulkanDescriptorSetCache mDescriptorSetCache;
     VulkanQueryManager mQueryManager;
     VulkanExternalImageManager mExternalImageManager;

filament/backend/src/vulkan/VulkanFboCache.cpp

@@ -15,6 +15,7 @@
  */
 
 #include "VulkanFboCache.h"
+#include "VulkanRenderPassStateSerializer.h"
 
 #include "VulkanConstants.h"
 #include "vulkan/utils/Image.h"
@@ -69,6 +70,15 @@ VulkanFboCache::~VulkanFboCache() {
             << "Please explicitly call terminate() while the VkDevice is still alive.";
 }
 
+uint32_t VulkanFboCache::getRenderPassKey(VkRenderPass pass) noexcept {
+    uint32_t key = 0;
+    auto iter = mRenderPassToKey.find(pass);
+    if (iter != mRenderPassToKey.end()) {
+        key = iter->second;
+    }
+    return key;
+}
+
 VkFramebuffer VulkanFboCache::getFramebuffer(FboKey const& config) noexcept {
     FboMap::iterator iter = mFramebufferCache.find(config);
     if (UTILS_LIKELY(iter != mFramebufferCache.end() && iter->second.handle != VK_NULL_HANDLE)) {
@@ -117,6 +127,7 @@ VkFramebuffer VulkanFboCache::getFramebuffer(FboKey const& config) noexcept {
     VkResult error = vkCreateFramebuffer(mDevice, &info, VKALLOC, &framebuffer);
     FILAMENT_CHECK_POSTCONDITION(error == VK_SUCCESS) << "Unable to create framebuffer."
                                                      << " error=" << static_cast<int32_t>(error);
+
     mFramebufferCache[config] = {framebuffer, mCurrentTime};
     return framebuffer;
 }
@@ -326,6 +337,12 @@ VkRenderPass VulkanFboCache::getRenderPass(RenderPassKey const& config) noexcept
     VkResult error = vkCreateRenderPass(mDevice, &renderPassInfo, VKALLOC, &renderPass);
     FILAMENT_CHECK_POSTCONDITION(error == VK_SUCCESS) << "Unable to create render pass."
                                                       << " error=" << error;
+
+    RenderPassHash hashFunc;
+    uint32_t key = hashFunc(config);
+    VulkanRenderPassStateSerializer passSer(renderPassInfo, key); 
+    mRenderPassToKey[renderPass] = key;
+
     mRenderPassCache[config] = {renderPass, mCurrentTime};
 
 #if FVK_ENABLED(FVK_DEBUG_FBO_CACHE)

filament/backend/src/vulkan/VulkanFboCache.h

@@ -24,6 +24,7 @@
 #include <backend/TargetBufferInfo.h>
 
 #include <tsl/robin_map.h>
+#include <map>
 
 namespace filament::backend {
 
@@ -97,6 +98,8 @@ public:
     explicit VulkanFboCache(VkDevice device);
     ~VulkanFboCache();
 
+    uint32_t getRenderPassKey(VkRenderPass pass) noexcept;
+
     // Retrieves or creates a VkFramebuffer handle.
     VkFramebuffer getFramebuffer(FboKey const& config) noexcept;
 
@@ -118,6 +121,7 @@ private:
     FboMap mFramebufferCache;
 
     tsl::robin_map<RenderPassKey, RenderPassVal, RenderPassHash, RenderPassEq> mRenderPassCache;
+    std::map<VkRenderPass, uint32_t> mRenderPassToKey;
     tsl::robin_map<VkRenderPass, uint32_t> mRenderPassRefCount;
     uint32_t mCurrentTime = 0;
 };

filament/backend/src/vulkan/VulkanHandles.cpp

@@ -33,6 +33,7 @@
 #include <utils/compiler.h> // UTILS_FALLTHROUGH
 #include <utils/Panic.h>    // ASSERT_POSTCONDITION
 #include <utils/CString.h>
+#include <fstream>
 
 using namespace bluevk;
 
@@ -179,6 +180,23 @@ VulkanAttachment createSwapchainAttachment(const fvkmemory::resource_ptr<VulkanT
     };
 }
 
+static inline uint64_t serializeShaderModule(utils::CString progName, VkShaderModuleCreateInfo info,
+        VkShaderStageFlagBits stage) {
+    std::stringstream filename;
+    filename << progName.c_str() << "_" << stage << "_";
+    constexpr uint32_t wordReadCount = 8;
+    uint64_t key = utils::hash::murmur3(info.pCode, wordReadCount, 0);
+    key |= uint64_t(utils::hash::murmur3(info.pCode + (wordReadCount / 2), wordReadCount, 0)) << 32;
+    filename << key << ".bin";
+
+    std::ofstream file(filename.str());
+    if (file.is_open()) {
+        file.write(reinterpret_cast<const char*>(info.pCode), info.codeSize);
+    }
+    file.close();
+    return key;
+}
+
 } // anonymous namespace
 
 VulkanDescriptorSetLayout::VulkanDescriptorSetLayout(DescriptorSetLayout&& layout,
@@ -280,6 +298,12 @@ VulkanProgram::VulkanProgram(VkDevice device, Program const& builder) noexcept
             .codeSize = dataSize,
             .pCode = data,
         };
+        if (i == 0) {
+            mVertexShaderHash = serializeShaderModule(name, moduleInfo, VK_SHADER_STAGE_VERTEX_BIT);
+        } else if (i == 1) {
+            mFragmentShaderHash =
+                    serializeShaderModule(name, moduleInfo, VK_SHADER_STAGE_FRAGMENT_BIT);
+        }
         VkResult result = vkCreateShaderModule(mDevice, &moduleInfo, VKALLOC, &module);
         FILAMENT_CHECK_POSTCONDITION(result == VK_SUCCESS)
                 << "Unable to create shader module."

filament/backend/src/vulkan/VulkanHandles.h

@@ -65,6 +65,8 @@ struct VulkanDescriptorSetLayout : public HwDescriptorSetLayout, fvkmemory::Reso
 
     using DescriptorSetLayoutArray = std::array<VkDescriptorSetLayout,
             VulkanDescriptorSetLayout::UNIQUE_DESCRIPTOR_SET_COUNT>;
+    using DescriptorSetLayoutHashArray = std::array<uint64_t,
+            VulkanDescriptorSetLayout::UNIQUE_DESCRIPTOR_SET_COUNT>;
 
     // The bitmask representation of a set layout.
     struct Bitmask {
@@ -280,6 +282,10 @@ struct VulkanProgram : public HwProgram, fvkmemory::Resource {
         mInfo->pushConstantDescription.write(cmdbuf, layout, stage, index, value);
     }
 
+    inline uint64_t getVertexShaderHash() { return mVertexShaderHash;
+    }
+    inline uint64_t getFragmentShaderHash() { return mFragmentShaderHash; }
+
     // TODO: handle compute shaders.
     // The expected order of shaders - from frontend to backend - is vertex, fragment, compute.
     static constexpr uint8_t const MAX_SHADER_MODULES = 2;
@@ -295,6 +301,8 @@ private:
     };
 
     PipelineInfo* mInfo;
+    uint64_t mVertexShaderHash;
+    uint64_t mFragmentShaderHash;
     VkDevice mDevice = VK_NULL_HANDLE;
 };
 

filament/backend/src/vulkan/VulkanPipelineCache.cpp

@@ -15,6 +15,7 @@
  */
 
 #include "VulkanPipelineCache.h"
+#include "VulkanPipelineStateSerializer.h"
 
 #include <utils/Log.h>
 #include <utils/Panic.h>
@@ -34,49 +35,17 @@ using namespace bluevk;
 
 namespace filament::backend {
 
-namespace {
-
-#if FVK_ENABLED(FVK_DEBUG_SHADER_MODULE)
-void printPipelineFeedbackInfo(VkPipelineCreationFeedbackCreateInfo const& feedbackInfo) {
-    VkPipelineCreationFeedback const& pipelineInfo = *feedbackInfo.pPipelineCreationFeedback;
-    if (!(pipelineInfo.flags & VK_PIPELINE_CREATION_FEEDBACK_VALID_BIT)) {
-        return;
-    }
-
-    bool const isCacheHit =
-            (pipelineInfo.flags & VK_PIPELINE_CREATION_FEEDBACK_APPLICATION_PIPELINE_CACHE_HIT_BIT);
-    FVK_LOGD << "Pipeline build stats - Cache hit: " << isCacheHit
-             << ", Time: " << pipelineInfo.duration / 1000000.0 << "ms";
-
-    for (uint32_t i = 0; i < feedbackInfo.pipelineStageCreationFeedbackCount; ++i) {
-        VkPipelineCreationFeedback const& stageInfo = feedbackInfo.pPipelineStageCreationFeedbacks[i];
-        if (!(stageInfo.flags & VK_PIPELINE_CREATION_FEEDBACK_VALID_BIT)) {
-            continue;
-        }
-
-        bool const isVertexShader = (i == 0);
-        bool const isCacheHit = (stageInfo.flags &
-                                 VK_PIPELINE_CREATION_FEEDBACK_APPLICATION_PIPELINE_CACHE_HIT_BIT);
-        FVK_LOGD << (isVertexShader ? "Vertex" : "Fragment")
-                 << " shader build stats - Cache hit: " << isCacheHit
-                 << ", Time: " << stageInfo.duration / 1000000.0 << "ms";
-    }
-}
-#endif
-
-} // namespace
-
-VulkanPipelineCache::VulkanPipelineCache(VkDevice device, VulkanContext const& context)
-        : mDevice(device),
-          mContext(context) {
+VulkanPipelineCache::VulkanPipelineCache(VkDevice device)
+    : mDevice(device) {
     VkPipelineCacheCreateInfo createInfo = {
         .sType = VK_STRUCTURE_TYPE_PIPELINE_CACHE_CREATE_INFO,
     };
     bluevk::vkCreatePipelineCache(mDevice, &createInfo, VKALLOC, &mPipelineCache);
 }
 
-void VulkanPipelineCache::bindLayout(VkPipelineLayout layout) noexcept {
+void VulkanPipelineCache::bindLayout(VkPipelineLayout layout, uint32_t key) noexcept {
     mPipelineRequirements.layout = layout;
+    mPipelineLayoutKey = key;
 }
 
 VulkanPipelineCache::PipelineCacheEntry* VulkanPipelineCache::getOrCreatePipeline() noexcept {
@@ -110,6 +79,7 @@ void VulkanPipelineCache::bindPipeline(VulkanCommandBuffer* commands) {
 VulkanPipelineCache::PipelineCacheEntry* VulkanPipelineCache::createPipeline() noexcept {
     assert_invariant(mPipelineRequirements.shaders[0] && "Vertex shader is not bound.");
     assert_invariant(mPipelineRequirements.layout && "No pipeline layout specified");
+    VulkanPipelineStateSerializer serializer(mProgramName);
 
     VkPipelineShaderStageCreateInfo shaderStages[SHADER_MODULE_COUNT];
     shaderStages[0] = {
@@ -122,6 +92,12 @@ VulkanPipelineCache::PipelineCacheEntry* VulkanPipelineCache::createPipeline() n
     shaderStages[1].stage = VK_SHADER_STAGE_FRAGMENT_BIT;
     shaderStages[1].module = mPipelineRequirements.shaders[1];
 
+    VulkanPipelineStateSerializer::ShaderStageInfo vtx = { shaderStages[0].stage, mVertexShaderHash,
+        shaderStages[0].pName };
+    VulkanPipelineStateSerializer::ShaderStageInfo frg = { shaderStages[1].stage, mFragmentShaderHash,
+        shaderStages[1].pName };
+    serializer << vtx << frg;
+
     bool const hasFragmentShader = shaderStages[1].module != VK_NULL_HANDLE;
 
     VkPipelineColorBlendAttachmentState colorBlendAttachments[MRT::MAX_SUPPORTED_RENDER_TARGET_COUNT];
@@ -140,9 +116,11 @@ VulkanPipelineCache::PipelineCacheEntry* VulkanPipelineCache::createPipeline() n
     for (uint32_t i = 0; i < VERTEX_ATTRIBUTE_COUNT; i++) {
         if (mPipelineRequirements.vertexAttributes[i].format > 0) {
             vertexAttributes[numVertexAttribs++] = mPipelineRequirements.vertexAttributes[i];
+            serializer << mPipelineRequirements.vertexAttributes[i];
         }
         if (mPipelineRequirements.vertexBuffers[i].stride > 0) {
             vertexBuffers[numVertexBuffers++] = mPipelineRequirements.vertexBuffers[i];
+            serializer << mPipelineRequirements.vertexBuffers[i];
         }
     }
     VkPipelineVertexInputStateCreateInfo vertexInputState = {
@@ -156,11 +134,15 @@ VulkanPipelineCache::PipelineCacheEntry* VulkanPipelineCache::createPipeline() n
         .sType = VK_STRUCTURE_TYPE_PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO,
         .topology = (VkPrimitiveTopology) mPipelineRequirements.topology,
     };
+    serializer << inputAssemblyState;
+
     VkPipelineViewportStateCreateInfo viewportState = {
         .sType = VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_STATE_CREATE_INFO,
         .viewportCount = 1,
         .scissorCount = 1,
     };
+    serializer << viewportState;
+
     VkDynamicState dynamicStateEnables[] = {
         VK_DYNAMIC_STATE_VIEWPORT,
         VK_DYNAMIC_STATE_SCISSOR,
@@ -170,6 +152,8 @@ VulkanPipelineCache::PipelineCacheEntry* VulkanPipelineCache::createPipeline() n
         .dynamicStateCount = 2,
         .pDynamicStates = dynamicStateEnables,
     };
+    serializer << dynamicState;
+
     auto const& raster = mPipelineRequirements.rasterState;
     VkPipelineRasterizationStateCreateInfo vkRaster = {
         .sType = VK_STRUCTURE_TYPE_PIPELINE_RASTERIZATION_STATE_CREATE_INFO,
@@ -183,6 +167,8 @@ VulkanPipelineCache::PipelineCacheEntry* VulkanPipelineCache::createPipeline() n
         .depthBiasSlopeFactor = raster.depthBiasSlopeFactor,
         .lineWidth = 1.0f,
     };
+    serializer << vkRaster;
+
     VkPipelineMultisampleStateCreateInfo vkMs = {
         .sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO,
         .rasterizationSamples = (VkSampleCountFlagBits) raster.rasterizationSamples,
@@ -191,6 +177,8 @@ VulkanPipelineCache::PipelineCacheEntry* VulkanPipelineCache::createPipeline() n
         .alphaToCoverageEnable = raster.alphaToCoverageEnable,
         .alphaToOneEnable = VK_FALSE,
     };
+    serializer << vkMs;
+
     VkPipelineDepthStencilStateCreateInfo vkDs = {
         .sType = VK_STRUCTURE_TYPE_PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO,
         .depthTestEnable = VK_TRUE,
@@ -210,6 +198,12 @@ VulkanPipelineCache::PipelineCacheEntry* VulkanPipelineCache::createPipeline() n
         .writeMask = 0u,
         .reference = 0u,
     };
+    serializer << vkDs;
+
+    //layout
+    serializer.setPipelineLayoutKey(mPipelineLayoutKey);
+    // renderPass
+    serializer.setRenderPassKey(mRenderPassKey);
 
     VkGraphicsPipelineCreateInfo pipelineCreateInfo = {
         .sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO,
@@ -248,52 +242,36 @@ VulkanPipelineCache::PipelineCacheEntry* VulkanPipelineCache::createPipeline() n
             colorBlendAttachments[i] = colorBlendAttachments[0];
         }
     }
+    serializer << colorBlendState;
 
-#if FVK_ENABLED(FVK_DEBUG_SHADER_MODULE)
-    FVK_LOGD << "vkCreateGraphicsPipelines with shaders = (" << shaderStages[0].module << ", "
-             << shaderStages[1].module << ")";
-
-    VkPipelineCreationFeedback stageFeedbacks[SHADER_MODULE_COUNT] = {};
-    VkPipelineCreationFeedback pipelineFeedback = {};
-    VkPipelineCreationFeedbackCreateInfo feedbackInfo = {
-        .sType = VK_STRUCTURE_TYPE_PIPELINE_CREATION_FEEDBACK_CREATE_INFO_EXT,
-        .pNext = nullptr,
-        .pPipelineCreationFeedback = &pipelineFeedback,
-        .pipelineStageCreationFeedbackCount = hasFragmentShader ? SHADER_MODULE_COUNT : 1,
-        .pPipelineStageCreationFeedbacks = stageFeedbacks,
-    };
-
-    if (mContext.pipelineCreationFeedbackSupported()) {
-        feedbackInfo.pNext = pipelineCreateInfo.pNext;
-        pipelineCreateInfo.pNext = &feedbackInfo;
-    }
-#endif
+    #if FVK_ENABLED(FVK_DEBUG_SHADER_MODULE)
+        FVK_LOGD << "vkCreateGraphicsPipelines with shaders = ("
+                 << shaderStages[0].module << ", " << shaderStages[1].module << ")";
+    #endif
     PipelineCacheEntry cacheEntry = {
         .lastUsed = mCurrentTime,
     };
     VkResult error = vkCreateGraphicsPipelines(mDevice, mPipelineCache, 1, &pipelineCreateInfo,
             VKALLOC, &cacheEntry.handle);
-
-#if FVK_ENABLED(FVK_DEBUG_SHADER_MODULE)
-    FVK_LOGD << "vkCreateGraphicsPipelines with shaders = (" << shaderStages[0].module << ", "
-             << shaderStages[1].module << ")";
-
-    if (mContext.pipelineCreationFeedbackSupported()) {
-        printPipelineFeedbackInfo(feedbackInfo);
-    }
-#endif
-
     assert_invariant(error == VK_SUCCESS);
     if (error != VK_SUCCESS) {
         FVK_LOGE << "vkCreateGraphicsPipelines error " << error;
         return nullptr;
     }
+
+    PipelineHashFn hashFunc;
+
+    serializer.setID(hashFunc(mPipelineRequirements));
     return &mPipelines.emplace(mPipelineRequirements, cacheEntry).first.value();
 }
 
 void VulkanPipelineCache::bindProgram(fvkmemory::resource_ptr<VulkanProgram> program) noexcept {
     mPipelineRequirements.shaders[0] = program->getVertexShader();
     mPipelineRequirements.shaders[1] = program->getFragmentShader();
+    mProgramName = program->name;
+    mVertexShaderHash = program->getVertexShaderHash();
+    mFragmentShaderHash= program->getFragmentShaderHash();
+    
 
     // If this is a debug build, validate the current shader.
 #if FVK_ENABLED(FVK_DEBUG_SHADER_MODULE)
@@ -308,9 +286,10 @@ void VulkanPipelineCache::bindRasterState(RasterState const& rasterState) noexce
     mPipelineRequirements.rasterState = rasterState;
 }
 
-void VulkanPipelineCache::bindRenderPass(VkRenderPass renderPass, int subpassIndex) noexcept {
+void VulkanPipelineCache::bindRenderPass(VkRenderPass renderPass, uint32_t passKey, int subpassIndex) noexcept {
     mPipelineRequirements.renderPass = renderPass;
     mPipelineRequirements.subpassIndex = subpassIndex;
+    mRenderPassKey = passKey;
 }
 
 void VulkanPipelineCache::bindPrimitiveTopology(VkPrimitiveTopology topology) noexcept {

filament/backend/src/vulkan/VulkanPipelineCache.h

@@ -86,9 +86,9 @@ public:
 
     static_assert(sizeof(RasterState) == 16, "RasterState must not have implicit padding.");
 
-    VulkanPipelineCache(VkDevice device, VulkanContext const& context);
+    VulkanPipelineCache(VkDevice device);
 
-    void bindLayout(VkPipelineLayout layout) noexcept;
+    void bindLayout(VkPipelineLayout layout, uint32_t key) noexcept;
 
     // Creates a new pipeline if necessary and binds it using vkCmdBindPipeline.
     void bindPipeline(VulkanCommandBuffer* commands);
@@ -96,7 +96,7 @@ public:
     // Each of the following methods are fast and do not make Vulkan calls.
     void bindProgram(fvkmemory::resource_ptr<VulkanProgram> program) noexcept;
     void bindRasterState(RasterState const& rasterState) noexcept;
-    void bindRenderPass(VkRenderPass renderPass, int subpassIndex) noexcept;
+    void bindRenderPass(VkRenderPass renderPass, uint32_t passKey, int subpassIndex) noexcept;
     void bindPrimitiveTopology(VkPrimitiveTopology topology) noexcept;
     void bindVertexArray(VkVertexInputAttributeDescription const* attribDesc,
             VkVertexInputBindingDescription const* bufferDesc, uint8_t count);
@@ -207,11 +207,14 @@ private:
 
     // Current requirements for the pipeline layout, pipeline, and descriptor sets.
     PipelineKey mPipelineRequirements = {};
+    utils::CString mProgramName;
+    uint64_t mVertexShaderHash;
+    uint64_t mFragmentShaderHash;
+    uint32_t mPipelineLayoutKey;
+    uint32_t mRenderPassKey;
 
     // Current bindings for the pipeline and descriptor sets.
     PipelineKey mBoundPipeline = {};
-
-    [[maybe_unused]] VulkanContext const& mContext;
 };
 
 } // namespace filament::backend

filament/backend/src/vulkan/VulkanPipelineLayoutCache.cpp

@@ -15,6 +15,7 @@
  */
 
 #include "VulkanPipelineLayoutCache.h"
+#include "VulkanDescriptorSetLayoutSerializer.h"
 
 namespace filament::backend {
 
@@ -23,7 +24,11 @@ VkPipelineLayout VulkanPipelineLayoutCache::getLayout(
         fvkmemory::resource_ptr<VulkanProgram> program) {
     PipelineLayoutKey key = {};
     uint8_t descSetLayoutCount = 0;
-    key.descSetLayouts = descriptorSetLayouts;
+    DescriptorSetLayoutHashArray hashArray;
+    for (uint32_t i = 0; i < descriptorSetLayouts.size(); ++i) {
+        hashArray[i] = mDescriptorSetCache->getKey(descriptorSetLayouts[i]);
+    }
+    key.descSetLayouts = hashArray;
     for (auto layoutHandle: descriptorSetLayouts) {
         if (layoutHandle == VK_NULL_HANDLE) {
             break;
@@ -62,11 +67,14 @@ VkPipelineLayout VulkanPipelineLayoutCache::getLayout(
         .sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO,
         .pNext = nullptr,
         .setLayoutCount = (uint32_t) descSetLayoutCount,
-        .pSetLayouts = key.descSetLayouts.data(),
+        .pSetLayouts = descriptorSetLayouts.data(),
         .pushConstantRangeCount = pushConstantRangeCount,
         .pPushConstantRanges = pushConstantRanges,
     };
 
+    PipelineLayoutKeyHashFn hashFunc;
+    uint32_t hashedKey = hashFunc(key);
+    VulkanPipelineLayoutSerializer pipelineSer(info, mDescriptorSetCache, hashedKey);
     VkPipelineLayout layout;
     vkCreatePipelineLayout(mDevice, &info, VKALLOC, &layout);
 
@@ -74,6 +82,8 @@ VkPipelineLayout VulkanPipelineLayoutCache::getLayout(
         .handle = layout,
         .lastUsed = mTimestamp++,
     };
+
+    mPipelineToKey[layout] = hashedKey;
     return layout;
 }
 

filament/backend/src/vulkan/VulkanPipelineLayoutCache.h

@@ -18,6 +18,7 @@
 #define TNT_FILAMENT_BACKEND_VULKANPIPELINELAYOUTCACHE_H
 
 #include "VulkanHandles.h"
+#include "VulkanDescriptorSetLayoutCache.h"
 
 #include <bluevk/BlueVK.h>
 
@@ -30,10 +31,11 @@ namespace filament::backend {
 class VulkanPipelineLayoutCache {
 public:
     using DescriptorSetLayoutArray = VulkanDescriptorSetLayout::DescriptorSetLayoutArray;
+    using DescriptorSetLayoutHashArray = VulkanDescriptorSetLayout::DescriptorSetLayoutHashArray;
     
-    VulkanPipelineLayoutCache(VkDevice device)
+    VulkanPipelineLayoutCache(VkDevice device, VulkanDescriptorSetLayoutCache* cache)
         : mDevice(device),
-          mTimestamp(0) {}
+          mTimestamp(0), mDescriptorSetCache(cache) {}
 
     void terminate() noexcept;
 
@@ -45,7 +47,7 @@ public:
     };
 
     struct PipelineLayoutKey {
-        DescriptorSetLayoutArray descSetLayouts = {};                                     // 8 * 4
+        DescriptorSetLayoutHashArray descSetLayouts = {};              // 8 * 4
         PushConstantKey pushConstant[Program::SHADER_TYPE_COUNT] = {};                    // 2 * 3
         uint16_t padding = 0;
     };
@@ -58,6 +60,14 @@ public:
     // are described in the program.
     VkPipelineLayout getLayout(DescriptorSetLayoutArray const& descriptorSetLayouts,
             fvkmemory::resource_ptr<VulkanProgram> program);
+    uint32_t getKey(VkPipelineLayout layout) {
+        uint32_t key = 0;
+        auto iter = mPipelineToKey.find(layout);
+        if (iter != mPipelineToKey.end()) {
+            key = iter->second;
+        }
+        return key;
+    }
 
 private:
     using Timestamp = uint64_t;
@@ -79,6 +89,8 @@ private:
     VkDevice mDevice;
     Timestamp mTimestamp;
     PipelineLayoutMap mPipelineLayouts;
+    std::map<VkPipelineLayout, uint32_t> mPipelineToKey;
+    VulkanDescriptorSetLayoutCache* mDescriptorSetCache;
 };
 
 } // filament::backend

filament/backend/src/vulkan/VulkanPipelineStateSerializer.cpp

@@ -0,0 +1,251 @@
+/*
+ * 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 <fstream>
+
+#include "VulkanPipelineStateSerializer.h"
+#include <utils/Log.h>
+#include <utils/Panic.h>
+
+#include "VulkanConstants.h"
+#include "VulkanHandles.h"
+
+#if defined(__clang__)
+// Vulkan functions often immediately dereference pointers, so it's fine to pass in a pointer
+// to a stack-allocated variable.
+#pragma clang diagnostic push
+#pragma clang diagnostic ignored "-Wreturn-stack-address"
+#endif
+
+using namespace bluevk;
+
+namespace filament::backend {
+
+VulkanPipelineStateSerializer::VulkanPipelineStateSerializer(const utils::CString& name) { 
+    mBuffer << "{" << std::endl;
+    mBuffer << "\"program_name\":" << "\"" << name.c_str() << "\"" << "," << std::endl;
+}
+VulkanPipelineStateSerializer::~VulkanPipelineStateSerializer() {
+    //Shader Modules
+    mBuffer << "\"shader_modules\":[" << std::endl;
+    for (uint32_t i = 0; i < mShaderStages.size(); ++i) {
+        mBuffer << "{" << std::endl;
+        mBuffer << "\"stage\":" << mShaderStages[i].stage << "," << std::endl;
+        mBuffer << "\"key\":" << mShaderStages[i].module  << "," << std::endl;
+        mBuffer << "\"name\":" << "\"" << mShaderStages[i].mName << "\"" << std::endl;
+        mBuffer << "}";
+        if (i < mShaderStages.size() - 1) {
+            mBuffer << ",";
+        }
+        mBuffer <<  std::endl;
+    }
+    mBuffer << "]," << std::endl;
+
+    //Vertex Attribute
+    mBuffer << "\"vertex_attributes\":[" << std::endl;
+    for (uint32_t i = 0; i < mVtxAttribs.size(); ++i) {
+        mBuffer << "{" << std::endl;
+        mBuffer << "\"location\":" << mVtxAttribs[i].location << "," << std::endl;
+        mBuffer << "\"binding\":" <<  mVtxAttribs[i].binding << "," << std::endl;
+        mBuffer << "\"format\":" << mVtxAttribs[i].format << "," << std::endl;
+        mBuffer << "\"offset\":" << mVtxAttribs[i].offset << std::endl;
+        mBuffer << "}";
+        if (i < mVtxAttribs.size() - 1) {
+            mBuffer << ",";
+        }
+        mBuffer << std::endl;
+    }
+    mBuffer << "]," << std::endl;
+
+    // Vertex Bindings
+    mBuffer << "\"vertex_bindings\":[" << std::endl;
+    for (uint32_t i = 0; i < mVtxBindings.size(); ++i) {
+        mBuffer << "{" << std::endl;
+        mBuffer << "\"binding\":" << mVtxBindings[i].binding << "," << std::endl;
+        mBuffer << "\"stride\":" << mVtxBindings[i].stride << "," << std::endl;
+        mBuffer << "\"input_rate\":" << mVtxBindings[i].inputRate << std::endl;
+        mBuffer << "}";
+        if (i < mVtxBindings.size() - 1) {
+            mBuffer << ",";
+        }
+        mBuffer << std::endl;
+    }
+    mBuffer << "]" << std::endl;
+
+    mBuffer << "}" << std::endl;
+
+    std::ofstream file(mFileName);
+    if (file.is_open()) {
+        file << mBuffer.str();
+    }
+    file.close();
+}
+
+void VulkanPipelineStateSerializer::setPipelineLayoutKey(uint32_t pipelineLayoutKey) {
+    mBuffer << "\"layout\":" << pipelineLayoutKey << "," << std::endl;
+}
+
+void VulkanPipelineStateSerializer::setRenderPassKey(uint32_t renderPassKey) {
+    mBuffer << "\"render_pass\":" << renderPassKey << "," << std::endl;
+}
+
+void VulkanPipelineStateSerializer::setID(uint32_t key) {
+    std::stringstream temp;
+    temp << "pipeline_" << key << ".json";
+    temp >> mFileName;
+}
+
+VulkanPipelineStateSerializer& VulkanPipelineStateSerializer::operator<<(
+        const VkPipelineInputAssemblyStateCreateInfo& inputAsm) {
+    mBuffer << "\"topology\":" << inputAsm.topology << "," << std::endl;
+    return *this;
+}
+
+VulkanPipelineStateSerializer& VulkanPipelineStateSerializer::operator<<(
+        const VkPipelineViewportStateCreateInfo& viewport) {
+    mBuffer << "\"viewport_count\":" << viewport.viewportCount << "," << std::endl;
+    mBuffer << "\"scissor_count\":" << viewport.scissorCount << "," << std::endl;
+    return *this;
+}
+
+VulkanPipelineStateSerializer& VulkanPipelineStateSerializer::operator<<(
+        const VkPipelineRasterizationStateCreateInfo& rasterState) {
+    mBuffer << "\"raster_states\":{" << std::endl;
+    mBuffer << "\"depth_clamp_enable\":" << rasterState.depthClampEnable << "," << std::endl;
+    mBuffer << "\"polygon_mode\":" << rasterState.polygonMode << "," << std::endl;
+    mBuffer << "\"cull_mode\":" << rasterState.cullMode << "," << std::endl;
+    mBuffer << "\"front_face\":" << rasterState.frontFace << "," << std::endl;
+    mBuffer << "\"depth_bias_enable\":" << rasterState.depthBiasEnable << "," << std::endl;
+    mBuffer << "\"depth_bias_constant_factor\":" << rasterState.depthBiasConstantFactor << ","
+            << std::endl;
+    mBuffer << "\"depth_bias_clamp\":" << rasterState.depthBiasClamp << "," << std::endl;
+    mBuffer << "\"depth_bias_slope_factor\":" << rasterState.depthBiasSlopeFactor << "," << std::endl;
+    mBuffer << "\"line_width\":" << rasterState.lineWidth << std::endl;
+    mBuffer << "}," << std::endl;
+    return *this;
+}
+
+VulkanPipelineStateSerializer& VulkanPipelineStateSerializer::operator<<(
+        const VkPipelineDepthStencilStateCreateInfo& depthStencil) {
+    mBuffer << "\"depth_stencil_states\":{" << std::endl;
+    mBuffer << "\"depth_test_enable\":" << depthStencil.depthTestEnable << "," << std::endl;
+    mBuffer << "\"depth_write_enable\":" << depthStencil.depthWriteEnable << "," << std::endl;
+    mBuffer << "\"depth_compare_op\":" << depthStencil.depthCompareOp << "," << std::endl;
+    mBuffer << "\"depth_bounds_test_enable\":" << depthStencil.depthBoundsTestEnable << ","
+            << std::endl;
+    mBuffer << "\"min_depth_bounds\":" << depthStencil.minDepthBounds << "," << std::endl;
+    mBuffer << "\"max_depth_bounds\":" << depthStencil.maxDepthBounds << "," << std::endl;
+    mBuffer << "\"front\":{" << std::endl;
+    mBuffer << "\"fail_op\":" << depthStencil.front.failOp << "," << std::endl;
+    mBuffer << "\"pass_op\":" << depthStencil.front.passOp << "," << std::endl;
+    mBuffer << "\"depth_fail_op\":" << depthStencil.front.depthFailOp << "," << std::endl;
+    mBuffer << "\"compare_op\":" << depthStencil.front.compareOp << "," << std::endl;
+    mBuffer << "\"compare_mask\":" << depthStencil.front.compareMask << "," << std::endl;
+    mBuffer << "\"write_mask\":" << depthStencil.front.writeMask << "," << std::endl;
+    mBuffer << "\"reference\":" << depthStencil.front.reference << std::endl;
+    mBuffer << "}," << std::endl;
+    mBuffer << "\"back\":{" << std::endl;
+    mBuffer << "\"fail_op\":" << depthStencil.back.failOp << "," << std::endl;
+    mBuffer << "\"pass_op\":" << depthStencil.back.passOp << "," << std::endl;
+    mBuffer << "\"depth_fail_op\":" << depthStencil.back.depthFailOp << "," << std::endl;
+    mBuffer << "\"compare_op\":" << depthStencil.back.compareOp << "," << std::endl;
+    mBuffer << "\"compare_mask\":" << depthStencil.back.compareMask << "," << std::endl;
+    mBuffer << "\"write_mask\":" << depthStencil.back.writeMask << "," << std::endl;
+    mBuffer << "\"reference\":" << depthStencil.back.reference << std::endl;
+    mBuffer << "}" << std::endl;
+    mBuffer << "}," << std::endl;
+    return *this;
+}
+
+VulkanPipelineStateSerializer& VulkanPipelineStateSerializer::operator<<(
+        const VkPipelineMultisampleStateCreateInfo& multiSample) {
+    mBuffer << "\"multisample_states\":{" << std::endl;
+    mBuffer << "\"rasterization_samples\":" << multiSample.rasterizationSamples << "," << std::endl;
+    mBuffer << "\"sample_shading_enable\":" << multiSample.sampleShadingEnable << "," << std::endl;
+    mBuffer << "\"min_sample_shading\":" << multiSample.minSampleShading << "," << std::endl;
+    mBuffer << "\"alpha_to_coverage_enable\":" << multiSample.alphaToCoverageEnable << ","
+            << std::endl;
+    mBuffer << "\"alpha_to_one_enable\":" << multiSample.alphaToOneEnable << std::endl;
+    mBuffer << "}," << std::endl;
+    return *this;
+}
+
+VulkanPipelineStateSerializer& VulkanPipelineStateSerializer::operator<<(
+        const VkPipelineDynamicStateCreateInfo& dyStates) {
+    mBuffer << "\"dynamic_states\":[" << std::endl;
+    for (uint32_t i = 0; i < dyStates.dynamicStateCount; ++i) {
+        mBuffer << dyStates.pDynamicStates[i];
+        if (i < dyStates.dynamicStateCount - 1) {
+            mBuffer << ",";
+        }
+        mBuffer << std::endl;
+    }
+    mBuffer << "]," << std::endl;
+    return *this;
+}
+
+VulkanPipelineStateSerializer& VulkanPipelineStateSerializer::operator<<(
+        const VkPipelineColorBlendStateCreateInfo& blendStateInfo) {
+    mBuffer << "\"color_blend_state_attachments\":[" << std::endl;
+    for (uint32_t i = 0; i < blendStateInfo.attachmentCount; ++i) {
+        mBuffer << "{" << std::endl;
+        mBuffer << "\"blend_enable\":" << blendStateInfo.pAttachments[i].blendEnable << ","
+                << std::endl;
+        mBuffer << "\"src_color_blend_factor\":" << blendStateInfo.pAttachments[i].srcColorBlendFactor
+                << ","
+                << std::endl;
+        mBuffer << "\"dst_color_blend_factor\":" << blendStateInfo.pAttachments[i].dstColorBlendFactor
+                << ","
+                << std::endl;
+        mBuffer << "\"color_blend_op\":" << blendStateInfo.pAttachments[i].colorBlendOp << ","
+                << std::endl;
+        mBuffer << "\"src_alpha_blend_factor\":" << blendStateInfo.pAttachments[i].srcAlphaBlendFactor
+                << ","
+                << std::endl;
+        mBuffer << "\"dst_alpha_blend_factor\":" << blendStateInfo.pAttachments[i].dstAlphaBlendFactor
+                << ","
+                << std::endl;
+        mBuffer << "\"alpha_blend_op\":" << blendStateInfo.pAttachments[i].alphaBlendOp << ","
+                << std::endl;
+        mBuffer << "\"color_write_mask\":" << blendStateInfo.pAttachments[i].colorWriteMask
+                << std::endl;
+        mBuffer << "}";
+        if (i < blendStateInfo.attachmentCount - 1) {
+            mBuffer << ",";
+        }
+        mBuffer << std::endl;
+    }
+    mBuffer << "]," << std::endl;
+    return *this;
+}
+
+VulkanPipelineStateSerializer& VulkanPipelineStateSerializer::operator<<(
+        const ShaderStageInfo& shaderStages) {
+    mShaderStages.push_back(shaderStages);
+    return *this;
+}
+
+VulkanPipelineStateSerializer& VulkanPipelineStateSerializer::operator<<(
+    const VkVertexInputAttributeDescription& attrib) {
+    mVtxAttribs.push_back(attrib);
+    return *this;
+}
+VulkanPipelineStateSerializer& VulkanPipelineStateSerializer::operator<<(
+    const VkVertexInputBindingDescription& binding) {
+    mVtxBindings.push_back(binding);
+    return *this;
+}
+}

filament/backend/src/vulkan/VulkanPipelineStateSerializer.h

@@ -0,0 +1,79 @@
+/*
+ * Copyright (C) 2022 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef TNT_FILAMENT_BACKEND_VULKANPIPELINESTATESERIALIZER_H
+#define TNT_FILAMENT_BACKEND_VULKANPIPELINESTATESERIALIZER_H
+
+#include "VulkanCommands.h"
+
+#include <backend/DriverEnums.h>
+#include <backend/TargetBufferInfo.h>
+
+#include <bluevk/BlueVK.h>
+#include <iostream>
+#include <type_traits>
+#include <sstream>
+#include <vector>
+
+namespace filament::backend {
+
+// VulkanPipelineStateSerializer stores to a file the pipeline states.
+class VulkanPipelineStateSerializer {
+public:
+    VulkanPipelineStateSerializer(VulkanPipelineStateSerializer const&) = delete;
+    VulkanPipelineStateSerializer& operator=(VulkanPipelineStateSerializer const&) = delete;
+
+    VulkanPipelineStateSerializer(const utils::CString& name);
+    ~VulkanPipelineStateSerializer();
+
+    void setID(uint32_t key);
+    void setPipelineLayoutKey(uint32_t pipelineLayoutKey);
+    void setRenderPassKey(uint32_t renderPassKey); 
+    VulkanPipelineStateSerializer& operator<<(
+            const VkPipelineInputAssemblyStateCreateInfo& inputAsm);
+    VulkanPipelineStateSerializer& operator<<(const VkPipelineViewportStateCreateInfo& viewport);
+    VulkanPipelineStateSerializer& operator<<(const VkPipelineDynamicStateCreateInfo& dyStates);
+    VulkanPipelineStateSerializer& operator<<(
+            const VkPipelineRasterizationStateCreateInfo& rasterState);
+    VulkanPipelineStateSerializer& operator<<(
+            const VkPipelineDepthStencilStateCreateInfo& depthStencil);
+    VulkanPipelineStateSerializer& operator<<(const VkPipelineMultisampleStateCreateInfo& multiSample);
+    VulkanPipelineStateSerializer& operator<<(
+            const VkPipelineColorBlendStateCreateInfo& blendStateInfo);
+
+    struct ShaderStageInfo {
+        VkShaderStageFlagBits stage;
+        uint64_t module;
+        std::string mName;
+    };
+    VulkanPipelineStateSerializer& operator<<(const ShaderStageInfo& shaderStages);
+    VulkanPipelineStateSerializer& operator<<(
+            const VkVertexInputAttributeDescription& attrib);
+    VulkanPipelineStateSerializer& operator<<(const VkVertexInputBindingDescription& binding);
+
+private:
+    std::string mFileName;
+    std::stringstream mBuffer;
+
+    //temp data
+    utils::CString mProgramName;
+    std::vector<ShaderStageInfo> mShaderStages;
+    std::vector<VkVertexInputAttributeDescription> mVtxAttribs;
+    std::vector<VkVertexInputBindingDescription> mVtxBindings;
+};
+} // namespace filament::backend
+
+#endif //TNT_FILAMENT_BACKEND_VULKANPIPELINESTATESERIALIZER_H

filament/backend/src/vulkan/VulkanRenderPassStateSerializer.cpp

@@ -0,0 +1,157 @@
+/*
+ * 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 <fstream>
+
+#include "VulkanRenderPassStateSerializer.h"
+#include <utils/Log.h>
+#include <utils/Panic.h>
+
+#include "VulkanConstants.h"
+#include "VulkanHandles.h"
+
+#if defined(__clang__)
+// Vulkan functions often immediately dereference pointers, so it's fine to pass in a pointer
+// to a stack-allocated variable.
+#pragma clang diagnostic push
+#pragma clang diagnostic ignored "-Wreturn-stack-address"
+#endif
+
+using namespace bluevk;
+
+namespace filament::backend {
+
+VulkanRenderPassStateSerializer::VulkanRenderPassStateSerializer(
+        const VkRenderPassCreateInfo& info, uint32_t key) {
+    std::stringstream filename;
+    filename << "render_pass_" << key << ".json";
+
+    std::ofstream file(filename.str());
+    if (file.is_open()) {
+        std::stringstream buffer;
+        buffer << "{" << std::endl;
+        buffer << "\"attachments\":[" << std::endl;
+        for (uint32_t i = 0; i < info.attachmentCount; ++i) {
+            buffer << "{" << std::endl;
+            buffer << "\"flags\":" << info.pAttachments[i].flags << "," << std::endl;
+            buffer << "\"format\":" << info.pAttachments[i].format << "," << std::endl;
+            buffer << "\"samples\":" << info.pAttachments[i].samples << "," << std::endl;
+            buffer << "\"load_op\":" << info.pAttachments[i].loadOp << "," << std::endl;
+            buffer << "\"store_op\":" << info.pAttachments[i].storeOp << "," << std::endl;
+            buffer << "\"stencil_load_op\":" << info.pAttachments[i].stencilLoadOp << "," << std::endl;
+            buffer << "\"stencil_store_op\":" << info.pAttachments[i].stencilStoreOp << "," << std::endl;
+            buffer << "\"initial_layout\":" << info.pAttachments[i].initialLayout << "," << std::endl;
+            buffer << "\"final_layout\":" << info.pAttachments[i].finalLayout << std::endl;
+            buffer << "}";
+            if (i < info.attachmentCount - 1) {
+                buffer << ",";
+            }
+            buffer << std::endl;
+        }
+        buffer << "]," << std::endl;
+
+        buffer << "\"subpasses\":[" << std::endl;
+        for (uint32_t i = 0; i < info.subpassCount; ++i) {
+            buffer << "{" << std::endl;
+            buffer << "\"flags\":" << info.pSubpasses[i].flags << "," << std::endl;
+            buffer << "\"pipeline_bind_point\":" << info.pSubpasses[i].pipelineBindPoint;
+            
+            //------------depth stencil-------------------
+            if (info.pSubpasses[i].pDepthStencilAttachment) {
+                buffer << "," << std::endl;
+                buffer << "\"depth_stencil_attachment\":{" << std::endl;
+                buffer << "\"attachement\":"
+                       << info.pSubpasses[i].pDepthStencilAttachment->attachment << ","
+                       << std::endl;
+                buffer << "\"layout\":" << info.pSubpasses[i].pDepthStencilAttachment->layout
+                       << std::endl;
+                buffer << "}";
+            }
+
+            //------------Resolve-------------------
+            if (info.pSubpasses[i].pResolveAttachments) {
+                buffer << "," << std::endl;
+                buffer << "\"resolve_attachments\":[" << std::endl;
+                for (uint32_t j = 0; j < info.pSubpasses[i].colorAttachmentCount; ++j) {
+                    buffer << "{" << std::endl;
+                    buffer << "\"attachement\":"
+                           << info.pSubpasses[i].pResolveAttachments[j].attachment << ","
+                           << std::endl;
+                    buffer << "\"layout\":" << info.pSubpasses[i].pResolveAttachments[j].layout
+                           << std::endl;
+                    buffer << "}" << std::endl;
+                    if (j < info.pSubpasses[i].colorAttachmentCount - 1) {
+                        buffer << ",";
+                    }
+                    buffer << std::endl;
+                }
+                buffer << "]";
+            }
+
+            //------------Color-------------------
+            if (info.pSubpasses[i].pColorAttachments) {
+                buffer << "," << std::endl;
+                buffer << "\"color_attachments\":[" << std::endl;
+                for (uint32_t j = 0; j < info.pSubpasses[i].colorAttachmentCount; ++j) {
+                    buffer << "{" << std::endl;
+                    buffer << "\"attachement\":"
+                           << info.pSubpasses[i].pColorAttachments[j].attachment << ","
+                           << std::endl;
+                    buffer << "\"layout\":" << info.pSubpasses[i].pColorAttachments[j].layout
+                           << std::endl;
+                    buffer << "}" << std::endl;
+                    if (j < info.pSubpasses[i].colorAttachmentCount - 1) {
+                        buffer << ",";
+                    }
+                    buffer << std::endl;
+                }
+                buffer << "]";
+            }
+
+            //------------Input-------------------
+            if (info.pSubpasses[i].pInputAttachments) {
+                buffer << "," << std::endl;
+                buffer << "\"input_attachments\":[" << std::endl;
+                for (uint32_t j = 0; j < info.pSubpasses[i].inputAttachmentCount; ++j) {
+                    buffer << "{" << std::endl;
+                    buffer << "\"attachement\":"
+                           << info.pSubpasses[i].pInputAttachments[j].attachment << ","
+                           << std::endl;
+                    buffer << "\"layout\":" << info.pSubpasses[i].pInputAttachments[j].layout
+                           << std::endl;
+                    buffer << "}" << std::endl;
+                    if (j < info.pSubpasses[i].inputAttachmentCount - 1) {
+                        buffer << ",";
+                    }
+                    buffer << std::endl;
+                }
+                buffer << "]" << std::endl;
+            }
+
+            buffer << "}";
+            if (i < info.subpassCount - 1) {
+                buffer << ",";
+            }
+            buffer << std::endl;
+        }
+        buffer << "]" << std::endl;
+
+        buffer << "}";
+        file << buffer.str();
+    }
+    file.close();
+}
+}

filament/backend/src/vulkan/VulkanRenderPassStateSerializer.h

@@ -0,0 +1,42 @@
+/*
+ * Copyright (C) 2022 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef TNT_FILAMENT_BACKEND_VULKANRENDERPASSSTATESERIALIZER_H
+#define TNT_FILAMENT_BACKEND_VULKANRENDERPASSSTATESERIALIZER_H
+
+#include "VulkanCommands.h"
+
+#include <backend/DriverEnums.h>
+#include <backend/TargetBufferInfo.h>
+
+#include <bluevk/BlueVK.h>
+#include <iostream>
+#include <type_traits>
+#include <sstream>
+#include <vector>
+
+namespace filament::backend {
+
+class VulkanRenderPassStateSerializer {
+public:
+    VulkanRenderPassStateSerializer(VulkanRenderPassStateSerializer const&) = delete;
+    VulkanRenderPassStateSerializer& operator=(VulkanRenderPassStateSerializer const&) = delete;
+
+    VulkanRenderPassStateSerializer(const VkRenderPassCreateInfo& info, uint32_t key);
+};
+} // namespace filament::backend
+
+#endif //TNT_FILAMENT_BACKEND_VULKANRENDERPASSSTATESERIALIZER_H

filament/backend/src/vulkan/VulkanSamplerCache.cpp

@@ -21,12 +21,14 @@
 
 #include <utils/Panic.h>
 
+#include "VulkanSamplerStateSerializer.h"
+
 using namespace bluevk;
 
 namespace filament::backend {
 
-VulkanSamplerCache::VulkanSamplerCache(VkDevice device)
-    : mDevice(device) {}
+VulkanSamplerCache::VulkanSamplerCache(VkDevice device, VulkanYcbcrConversionCache* conversionCache)
+    : mDevice(device), mConversionCache(conversionCache) {}
 
 VkSampler VulkanSamplerCache::getSampler(Params params) {
     auto iter = mCache.find(params);
@@ -57,11 +59,16 @@ VkSampler VulkanSamplerCache::getSampler(Params params) {
         .borderColor = VK_BORDER_COLOR_INT_OPAQUE_BLACK,
         .unnormalizedCoordinates = VK_FALSE,
     };
+    VulkanSamplerStateSerializer samplerSer(params,
+            params.conversion != VK_NULL_HANDLE ? mConversionCache->getKey(params.conversion) : 0);
+
     VkSampler sampler;
     VkResult result = vkCreateSampler(mDevice, &samplerInfo, VKALLOC, &sampler);
     FILAMENT_CHECK_POSTCONDITION(result == VK_SUCCESS) << "Unable to create sampler."
                                                        << " error=" << static_cast<int32_t>(result);
     mCache.insert({ params, sampler });
+    SamplerHashFn hashFn;
+    mSamplerToKey[sampler] = hashFn(params);
     return sampler;
 }
 

filament/backend/src/vulkan/VulkanSamplerCache.h

@@ -17,9 +17,11 @@
 #ifndef TNT_FILAMENT_BACKEND_VULKANSAMPLERCACHE_H
 #define TNT_FILAMENT_BACKEND_VULKANSAMPLERCACHE_H
 
+#include "VulkanYcbcrConversionCache.h"
 #include <backend/DriverEnums.h>
 
 #include <utils/Hash.h>
+#include <map>
 
 #include <bluevk/BlueVK.h>
 #include <tsl/robin_map.h>
@@ -37,11 +39,20 @@ public:
 
     static_assert(sizeof(Params) == 16);
 
-    explicit VulkanSamplerCache(VkDevice device);
+    explicit VulkanSamplerCache(VkDevice device, VulkanYcbcrConversionCache* conversionCache);
     VkSampler getSampler(Params params);
+    uint32_t getKey(VkSampler sampler) {
+        uint32_t key = 0;
+        auto iter = mSamplerToKey.find(sampler);
+        if (iter != mSamplerToKey.end()) {
+            key = iter->second;
+        }
+        return key;
+    }
     void terminate() noexcept;
 private:
     VkDevice mDevice;
+    VulkanYcbcrConversionCache* mConversionCache;
 
     struct SamplerEqualTo {
         bool operator()(Params lhs, Params rhs) const noexcept {
@@ -51,6 +62,7 @@ private:
     };
     using SamplerHashFn = utils::hash::MurmurHashFn<Params>;
     tsl::robin_map<Params, VkSampler, SamplerHashFn, SamplerEqualTo> mCache;
+    std::map<VkSampler, uint32_t> mSamplerToKey;
 };
 
 } // namespace filament::backend

filament/backend/src/vulkan/VulkanSamplerStateSerializer.cpp

@@ -0,0 +1,64 @@
+#include "VulkanSamplerStateSerializer.h"
+#include <fstream>
+#include <sstream>
+#include <utils/Hash.h>
+
+namespace filament::backend {
+VulkanYcbcrConversionSerializer::VulkanYcbcrConversionSerializer(
+        const VulkanYcbcrConversionCache::Params& params) {
+    using ConversionHashFn = utils::hash::MurmurHashFn<VulkanYcbcrConversionCache::Params>;
+    ConversionHashFn hashFn;
+    uint32_t key = hashFn(params);
+    std::stringstream filename;
+    filename << "ycbcr_conv_" << key << ".json";
+
+    std::ofstream file(filename.str());
+    if (file.is_open()) {
+        std::stringstream buffer;
+        auto const& chroma = params.conversion;
+        buffer << "{" << std::endl;
+        buffer << "\"format\":" << params.format << "," << std::endl;
+        buffer << "\"external_format\":" << params.externalFormat << "," << std::endl;
+        buffer << "\"ycbcr_model\":" << uint8_t(chroma.ycbcrModel) << "," << std::endl;
+        buffer << "\"ycbcr_range\":" << uint8_t(chroma.ycbcrRange) << "," << std::endl;
+        buffer << "\"swizzle_array\":[" << uint8_t(chroma.r) << "," << uint8_t(chroma.g) << ","
+               << uint8_t(chroma.b) << "," << uint8_t(chroma.a) << "]," << std::endl;
+        buffer << "\"x_chroma_offset\":" << uint8_t(chroma.xChromaOffset) << "," << std::endl;
+        buffer << "\"y_chroma_offset\":" << uint8_t(chroma.yChromaOffset) << "," << std::endl;
+        buffer << "\"chroma_filter\":" << uint8_t(chroma.chromaFilter) << std::endl;
+        buffer << "}" << std::endl;
+
+        file << buffer.str();
+    }
+    file.close();
+}
+
+VulkanSamplerStateSerializer::VulkanSamplerStateSerializer(const VulkanSamplerCache::Params& params, uint32_t ycbcr_conv_key) {
+    using ConversionHashFn = utils::hash::MurmurHashFn<VulkanSamplerCache::Params>;
+    ConversionHashFn hashFn;
+    uint32_t key = hashFn(params);
+    std::stringstream filename;
+    filename << "sampler_" << key << ".json";
+
+    std::ofstream file(filename.str());
+    if (file.is_open()) {
+        std::stringstream buffer;
+        auto const& samplerParams = params.sampler;
+        buffer << "{" << std::endl;
+        buffer << "\"filter_mag\":" << uint32_t(samplerParams.filterMag) << "," << std::endl;
+        buffer << "\"filter_min\":" << uint32_t(samplerParams.filterMin) << "," << std::endl;
+        buffer << "\"wrap_s\":" << uint32_t(samplerParams.wrapS) << "," << std::endl;
+        buffer << "\"wrap_t\":" << uint32_t(samplerParams.wrapT) << "," << std::endl;
+        buffer << "\"wrap_r\":" << uint32_t(samplerParams.wrapR) << "," << std::endl;
+        buffer << "\"anisotropy_log2\":" << uint32_t(samplerParams.anisotropyLog2) << "," << std::endl;
+        buffer << "\"compare_mode\":" << uint32_t(samplerParams.compareMode) << "," << std::endl;
+        buffer << "\"compare_func\":" << uint32_t(samplerParams.compareFunc) << "," << std::endl;
+        buffer << "\"ycbcr_conv_key\":" << ycbcr_conv_key << std::endl;
+        buffer << "}" << std::endl;
+
+        file << buffer.str();
+    }
+    file.close();
+}
+
+} // namespace filament::backend

filament/backend/src/vulkan/VulkanSamplerStateSerializer.h

@@ -0,0 +1,48 @@
+/*
+ * Copyright (C) 2022 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef TNT_FILAMENT_BACKEND_VULKANSAMPLERSTATESERIALIZER_H
+#define TNT_FILAMENT_BACKEND_VULKANSAMPLERSTATESERIALIZER_H
+
+#include <backend/DriverEnums.h>
+#include "VulkanYcbcrConversionCache.h"
+#include "VulkanSamplerCache.h"
+
+#include <bluevk/BlueVK.h>
+#include <vector>
+#include <iostream>
+#include <sstream>
+
+namespace filament::backend {
+
+class VulkanYcbcrConversionSerializer {
+public:
+    VulkanYcbcrConversionSerializer(VulkanYcbcrConversionSerializer const&) = delete;
+    VulkanYcbcrConversionSerializer& operator=(VulkanYcbcrConversionSerializer const&) = delete;
+
+    VulkanYcbcrConversionSerializer(const VulkanYcbcrConversionCache::Params& params);
+};
+        // VulkanPipelineStateSerializer stores to a file the pipeline states.
+class VulkanSamplerStateSerializer {
+public:
+    VulkanSamplerStateSerializer(VulkanSamplerStateSerializer const&) = delete;
+    VulkanSamplerStateSerializer& operator=(VulkanSamplerStateSerializer const&) = delete;
+
+    VulkanSamplerStateSerializer(const VulkanSamplerCache::Params& params, uint32_t ycbcr_conv_key);
+};
+}
+
+#endif //TNT_FILAMENT_BACKEND_VULKANSAMPLERSTATESERIALIZER_H

filament/backend/src/vulkan/VulkanYcbcrConversionCache.cpp

@@ -15,6 +15,7 @@
  */
 
 #include "VulkanYcbcrConversionCache.h"
+#include "VulkanSamplerStateSerializer.h"
 
 #include "vulkan/VulkanConstants.h"
 #include "vulkan/utils/Conversion.h"
@@ -53,6 +54,7 @@ VkSamplerYcbcrConversion VulkanYcbcrConversionCache::getConversion(
         .yChromaOffset = fvkutils::getChromaLocation(chroma.yChromaOffset),
         .chromaFilter = fvkutils::getFilter(chroma.chromaFilter),
     };
+    VulkanYcbcrConversionSerializer ycbcr_ser(params);
 
     // We could put this in the platform class, but that seems like a bit of an overkill
 #if defined(__ANDROID__)
@@ -73,6 +75,8 @@ VkSamplerYcbcrConversion VulkanYcbcrConversionCache::getConversion(
                                                        << " error=" << static_cast<int32_t>(result);
 
     mCache.insert({ params, conversion });
+    ConversionHashFn hashFn;
+    mConversionToKey[conversion] = hashFn(params);
     return conversion;
 }
 

filament/backend/src/vulkan/VulkanYcbcrConversionCache.h

@@ -25,6 +25,7 @@
 
 #include <bluevk/BlueVK.h>
 #include <tsl/robin_map.h>
+#include <map>
 
 namespace filament::backend {
 
@@ -40,6 +41,14 @@ public:
 
     explicit VulkanYcbcrConversionCache(VkDevice device);
     VkSamplerYcbcrConversion getConversion(Params params);
+    uint32_t getKey(VkSamplerYcbcrConversion conv) { 
+        uint32_t key = 0;
+        auto iter = mConversionToKey.find(conv);
+        if (iter != mConversionToKey.end()) {
+            key = iter->second;
+        }
+        return key;
+    }
     void terminate() noexcept;
 
 private:
@@ -54,6 +63,7 @@ private:
     };
     using ConversionHashFn = utils::hash::MurmurHashFn<Params>;
     tsl::robin_map<Params, VkSamplerYcbcrConversion, ConversionHashFn, ConversionEqualTo> mCache;
+    std::map<VkSamplerYcbcrConversion, uint32_t> mConversionToKey;
 };
 
 } // namespace filament::backend

filament/backend/src/vulkan/platform/VulkanPlatform.cpp

@@ -228,10 +228,6 @@ ExtensionSet getDeviceExtensions(VkPhysicalDevice device) {
         VK_KHR_PORTABILITY_SUBSET_EXTENSION_NAME,
 #endif
         VK_KHR_MULTIVIEW_EXTENSION_NAME,
-
-#if FVK_ENABLED(FVK_DEBUG_SHADER_MODULE)
-        VK_EXT_PIPELINE_CREATION_FEEDBACK_EXTENSION_NAME,
-#endif
     };
     ExtensionSet exts;
     // Identify supported physical device extensions
@@ -869,8 +865,6 @@ Driver* VulkanPlatform::createDriver(void* sharedContext,
     if (!mImpl->mSharedContext) {
         context.mDebugUtilsSupported = setContains(instExts, VK_EXT_DEBUG_UTILS_EXTENSION_NAME);
         context.mDebugMarkersSupported = setContains(deviceExts, VK_EXT_DEBUG_MARKER_EXTENSION_NAME);
-        context.mPipelineCreationFeedbackSupported =
-                setContains(deviceExts, VK_EXT_PIPELINE_CREATION_FEEDBACK_EXTENSION_NAME);
     } else {
         VulkanSharedContext const* scontext = (VulkanSharedContext const*) sharedContext;
         context.mDebugUtilsSupported = scontext->debugUtilsSupported;
@@ -879,16 +873,13 @@ Driver* VulkanPlatform::createDriver(void* sharedContext,
 
     // Check the availability of lazily allocated memory
     context.mLazilyAllocatedMemorySupported = false;
-    // RenderDoc doesn't support lazy allocated memory
-    if constexpr (!FVK_RENDERDOC_CAPTURE_MODE) {
-        for (uint32_t i = 0, typeCount = context.mMemoryProperties.memoryTypeCount; i < typeCount;
-                ++i) {
-            VkMemoryType const type = context.mMemoryProperties.memoryTypes[i];
-            if (type.propertyFlags & VK_MEMORY_PROPERTY_LAZILY_ALLOCATED_BIT) {
-                context.mLazilyAllocatedMemorySupported = true;
-                assert_invariant(type.propertyFlags & VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT);
-                break;
-            }
+    for (uint32_t i = 0, typeCount = context.mMemoryProperties.memoryTypeCount; i < typeCount;
+         ++i) {
+        VkMemoryType const type = context.mMemoryProperties.memoryTypes[i];
+        if (type.propertyFlags & VK_MEMORY_PROPERTY_LAZILY_ALLOCATED_BIT) {
+            context.mLazilyAllocatedMemorySupported = true;
+            assert_invariant(type.propertyFlags & VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT);
+            break;
         }
     }
 
@@ -1031,8 +1022,4 @@ VkExternalFenceHandleTypeFlagBits VulkanPlatform::getFenceExportFlags() const no
     return static_cast<VkExternalFenceHandleTypeFlagBits>(0);
 }
 
-bool VulkanPlatform::isTransientAttachmentSupported() const noexcept {
-    return mImpl->mContext.isLazilyAllocatedMemorySupported();
-}
-
 } // namespace filament::backend

filament/backend/src/vulkan/platform/VulkanPlatformAndroid.cpp

@@ -158,47 +158,7 @@ std::pair<TextureFormat, TextureUsage> getFilamentFormatAndUsage(const AHardware
     };
 }
 
-uint32_t selectMemoryTypeForExternalImage(VkPhysicalDevice physicalDevice, VkDevice device,
-        VkImage image, uint32_t types, VkFlags requiredMemoryFlags) {
-    VkPhysicalDeviceMemoryProperties memoryProperties;
-    vkGetPhysicalDeviceMemoryProperties(physicalDevice, &memoryProperties);
-    uint32_t const memoryTypeIndex =
-            VulkanContext::selectMemoryType(memoryProperties, types, requiredMemoryFlags);
-
-    if constexpr (FVK_RENDERDOC_CAPTURE_MODE) {
-        // RenderDoc will replay external resources as non-external.
-        // Adjust properties that will trip it up when replaying, even though these are not valid.
-        // Update memory type index if necessary so that we can replay the capture.
-
-        VkMemoryRequirements imageMemoryRequirements;
-        vkGetImageMemoryRequirements(device, image, &imageMemoryRequirements);
-
-        uint32_t const imageMemoryTypeBits = imageMemoryRequirements.memoryTypeBits;
-        bool const isMemoryTypeSupported = ((1 << memoryTypeIndex) & imageMemoryTypeBits) != 0;
-        if (isMemoryTypeSupported) {
-            return memoryTypeIndex;
-        }
-
-        // Current memory type will not be replayable by RenderDoc.
-        // Attempt to change the memory type index
-        VkMemoryPropertyFlags const kRenderDocFallBackReqs = 0;
-        uint32_t commonMemoryTypeBits = types & imageMemoryTypeBits;
-        uint32_t commonTypeIndex = VulkanContext::selectMemoryType(memoryProperties,
-                commonMemoryTypeBits, kRenderDocFallBackReqs);
-        if (commonMemoryTypeBits && commonTypeIndex != VK_MAX_MEMORY_TYPES) {
-            return commonTypeIndex;
-        }
-        uint32_t imageTypeIndex = VulkanContext::selectMemoryType(memoryProperties,
-                imageMemoryTypeBits, kRenderDocFallBackReqs);
-        if (imageTypeIndex != VK_MAX_MEMORY_TYPES) {
-            return imageTypeIndex;
-        }
-    }
-
-    return memoryTypeIndex;
-}
-
-} // namespace
+}// namespace
 
 VulkanPlatformAndroid::ExternalImageVulkanAndroid::~ExternalImageVulkanAndroid() {
     if (__builtin_available(android 26, *)) {
@@ -394,6 +354,8 @@ VulkanPlatform::ImageData VulkanPlatformAndroid::createVkImageFromExternal(
             .image = image,
             .buffer = VK_NULL_HANDLE,
         };
+        VkPhysicalDeviceMemoryProperties memoryProperties;
+        vkGetPhysicalDeviceMemoryProperties(physicalDevice, &memoryProperties);
         VkMemoryPropertyFlags requiredMemoryFlags =
                 !isExternal && any(metadata.filamentUsage & TextureUsage::UPLOADABLE)
                         ? VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT
@@ -403,8 +365,8 @@ VulkanPlatform::ImageData VulkanPlatformAndroid::createVkImageFromExternal(
             requiredMemoryFlags |= VK_MEMORY_PROPERTY_PROTECTED_BIT;
         }
 
-        uint32_t const memoryTypeIndex = selectMemoryTypeForExternalImage(physicalDevice, device,
-                image, metadata.memoryTypeBits, requiredMemoryFlags);
+        uint32_t const memoryTypeIndex = VulkanContext::selectMemoryType(memoryProperties,
+                metadata.memoryTypeBits, requiredMemoryFlags);
 
         VkMemoryAllocateInfo const allocInfo = {
             .sType = VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO,

filament/backend/src/webgpu/WebGPUBufferBase.cpp

@@ -63,7 +63,7 @@ WebGPUBufferBase::WebGPUBufferBase(wgpu::Device const& device, const wgpu::Buffe
 // WebGPU requires that the size of the data copied from the staging buffer to the GPU buffer is a
 // multiple of 4. This function handles cases where the buffer descriptor's size is not a multiple
 // of 4 by padding with zeros.
-void WebGPUBufferBase::updateGPUBuffer(BufferDescriptor&& bufferDescriptor,
+void WebGPUBufferBase::updateGPUBuffer(BufferDescriptor const& bufferDescriptor,
         const uint32_t byteOffset, wgpu::Device const& device,
         WebGPUQueueManager* const webGPUQueueManager) {
     FILAMENT_CHECK_PRECONDITION(bufferDescriptor.buffer)
@@ -89,49 +89,24 @@ void WebGPUBufferBase::updateGPUBuffer(BufferDescriptor&& bufferDescriptor,
     wgpu::BufferDescriptor descriptor{
         .label = "Filament WebGPU Staging Buffer",
         .usage = wgpu::BufferUsage::MapWrite | wgpu::BufferUsage::CopySrc,
-        .size = stagingBufferSize};
+        .size = stagingBufferSize,
+        .mappedAtCreation = true };
     wgpu::Buffer stagingBuffer = device.CreateBuffer(&descriptor);
 
-    struct UserData final {
-        uint32_t byteOffset;
-        size_t stagingBufferSize;
-        size_t remainder;
-        BufferDescriptor srcBufferDescriptor;
-        wgpu::Buffer stagingBuffer;
-        WebGPUQueueManager* const webGPUQueueManager;
-        wgpu::Buffer dstBuffer;
-    };
-    auto userData = std::make_unique<UserData>(UserData{
-        .byteOffset = byteOffset,
-        .stagingBufferSize = stagingBufferSize,
-        .remainder = remainder,
-        .srcBufferDescriptor = std::move(bufferDescriptor),
-        .stagingBuffer = stagingBuffer,
-        .webGPUQueueManager = webGPUQueueManager,
-        .dstBuffer = mBuffer});
-    stagingBuffer.MapAsync(
-            wgpu::MapMode::Write, 0, stagingBufferSize, wgpu::CallbackMode::AllowProcessEvents,
-            [](wgpu::MapAsyncStatus status, const char* message, UserData* userdata) {
-                std::unique_ptr<UserData> data(static_cast<UserData*>(userdata));
-                if (UTILS_LIKELY(status == wgpu::MapAsyncStatus::Success)) {
-                    void* mappedRange = data->stagingBuffer.GetMappedRange();
-                    memcpy(mappedRange, data->srcBufferDescriptor.buffer,
-                            data->srcBufferDescriptor.size);
-                    if (data->remainder != 0) {
-                        uint8_t* paddingStart =
-                                static_cast<uint8_t*>(mappedRange) + data->srcBufferDescriptor.size;
-                        memset(paddingStart, 0,
-                                FILAMENT_WEBGPU_BUFFER_SIZE_MODULUS - data->remainder);
-                    }
-                    data->stagingBuffer.Unmap();
-                    data->webGPUQueueManager->getCommandEncoder().CopyBufferToBuffer(
-                            data->stagingBuffer, 0, data->dstBuffer, data->byteOffset,
-                            data->stagingBufferSize);
-                } else {
-                    FWGPU_LOGE << "Failed to map staging buffer for readPixels: " << message;
-                }
-            },
-            userData.release());
+    void* mappedRange = stagingBuffer.GetMappedRange();
+    memcpy(mappedRange, bufferDescriptor.buffer, bufferDescriptor.size);
+
+    // Make sure the padded memory is set to 0 to have deterministic behaviors
+    if (remainder != 0) {
+        uint8_t* paddingStart = static_cast<uint8_t*>(mappedRange) + bufferDescriptor.size;
+        memset(paddingStart, 0, FILAMENT_WEBGPU_BUFFER_SIZE_MODULUS - remainder);
+    }
+
+    stagingBuffer.Unmap();
+
+    // Copy the staging buffer contents to the destination buffer.
+    webGPUQueueManager->getCommandEncoder().CopyBufferToBuffer(stagingBuffer, 0, mBuffer,
+            byteOffset, stagingBufferSize);
 }
 
 } // namespace filament::backend

filament/backend/src/webgpu/WebGPUBufferBase.h

@@ -39,7 +39,7 @@ public:
      * happen after draw commands encoded in the encoder. Submitting any commands up to this point
      * ensures the calls happen in the expected sequence.
      */
-    void updateGPUBuffer(BufferDescriptor&&, uint32_t byteOffset, wgpu::Device const& device,
+    void updateGPUBuffer(BufferDescriptor const&, uint32_t byteOffset, wgpu::Device const& device,
             WebGPUQueueManager* const webGPUQueueManager);
 
     [[nodiscard]] wgpu::Buffer const& getBuffer() const { return mBuffer; }

filament/backend/src/webgpu/WebGPUDriver.cpp

@@ -856,7 +856,7 @@ void WebGPUDriver::updateIndexBuffer(Handle<HwIndexBuffer> indexBufferHandle,
     // draw calls are made.
     flush();
     handleCast<WebGPUIndexBuffer>(indexBufferHandle)
-            ->updateGPUBuffer(std::move(bufferDescriptor), byteOffset, mDevice, &mQueueManager);
+            ->updateGPUBuffer(bufferDescriptor, byteOffset, mDevice, &mQueueManager);
     scheduleDestroy(std::move(bufferDescriptor));
 }
 
@@ -867,14 +867,14 @@ void WebGPUDriver::updateBufferObject(Handle<HwBufferObject> bufferObjectHandle,
     // draw calls are made.
     flush();
     handleCast<WebGPUBufferObject>(bufferObjectHandle)
-            ->updateGPUBuffer(std::move(bufferDescriptor), byteOffset, mDevice, &mQueueManager);
+            ->updateGPUBuffer(bufferDescriptor, byteOffset, mDevice, &mQueueManager);
     scheduleDestroy(std::move(bufferDescriptor));
 }
 
 void WebGPUDriver::updateBufferObjectUnsynchronized(Handle<HwBufferObject> bufferObjectHandle,
         BufferDescriptor&& bufferDescriptor, const uint32_t byteOffset) {
     handleCast<WebGPUBufferObject>(bufferObjectHandle)
-            ->updateGPUBuffer(std::move(bufferDescriptor), byteOffset, mDevice, &mQueueManager);
+            ->updateGPUBuffer(bufferDescriptor, byteOffset, mDevice, &mQueueManager);
     scheduleDestroy(std::move(bufferDescriptor));
 }
 

filament/backend/src/webgpu/WebGPUQueueManager.cpp

@@ -21,7 +21,6 @@
 #include <chrono>
 #include <cstdint>
 #include <thread>
-#include <iostream>
 
 namespace filament::backend {
 
@@ -59,7 +58,6 @@ WebGPUQueueManager::WebGPUQueueManager(wgpu::Device const& device)
 WebGPUQueueManager::~WebGPUQueueManager() = default;
 
 wgpu::CommandEncoder WebGPUQueueManager::getCommandEncoder() {
-    // std::unique_lock<std::mutex> lock(mLock);
     if (!mCommandEncoder) {
         wgpu::CommandEncoderDescriptor commandEncoderDescriptor = {
             .label = "Filament Command Encoder",
@@ -92,9 +90,7 @@ std::shared_ptr<WebGPUSubmissionState> WebGPUQueueManager::getLatestSubmissionSt
 }
 
 void WebGPUQueueManager::submit() {
-    // std::unique_lock<std::mutex> lock(mLock);
     if (!mCommandEncoder) {
-        std::cout << "Run Yu: no mCommandEncoder found!\n";
         return;
     }
 

filament/backend/src/webgpu/WebGPUQueueManager.h

@@ -74,7 +74,6 @@ private:
     wgpu::Queue mQueue;
     wgpu::CommandEncoder mCommandEncoder;
     std::shared_ptr<WebGPUSubmissionState> mLatestSubmissionState;
-    std::mutex mLock;
 };
 
 } // namespace filament::backend

filament/src/PostProcessManager.cpp

@@ -258,10 +258,6 @@ void PostProcessManager::bindPerRenderableDescriptorSet(DriverApi& driver) const
             { { 0, 0 }, driver });
 }
 
-UboManager* PostProcessManager::getUboManager() const noexcept {
-    return mEngine.getUboManager();
-}
-
 UTILS_NOINLINE
 void PostProcessManager::registerPostProcessMaterial(std::string_view const name,
         StaticMaterialInfo const& info) {
@@ -517,7 +513,7 @@ UTILS_NOINLINE
 void PostProcessManager::commitAndRenderFullScreenQuad(DriverApi& driver,
         FrameGraphResources::RenderPassInfo const& out, FMaterialInstance const* mi,
         PostProcessVariant const variant) const noexcept {
-    mi->commit(driver, getUboManager());
+    mi->commit(driver);
     mi->use(driver);
     FMaterial const* const ma = mi->getMaterial();
     PipelineState const pipeline = getPipelineState(ma, variant);
@@ -649,7 +645,7 @@ PostProcessManager::StructurePassOutput PostProcessManager::structure(FrameGraph
                     auto th = driver.createTextureView(in, level, 1);
                     mi->setParameter("depth", th, SamplerParams{
                         .filterMin = SamplerMinFilter::NEAREST_MIPMAP_NEAREST });
-                    mi->commit(driver, getUboManager());
+                    mi->commit(driver);
                     mi->use(driver);
                     renderFullScreenQuad(out, pipeline, driver);
                     DescriptorSet::unbind(driver, DescriptorSetBindingPoints::PER_MATERIAL);
@@ -1083,7 +1079,7 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::screenSpaceAmbientOcclusion(
                 mi->setParameter("ssctRayCount",
                         float2{ options.ssct.rayCount, 1.0f / float(options.ssct.rayCount) });
 
-                mi->commit(driver, getUboManager());
+                mi->commit(driver);
                 mi->use(driver);
 
                 auto pipeline = getPipelineState(ma);
@@ -1204,7 +1200,7 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::bilateralBlurPass(FrameGraph
                 mi->setParameter("sampleCount", kGaussianCount);
                 mi->setParameter("farPlaneOverEdgeDistance", -zf / config.bilateralThreshold);
 
-                mi->commit(driver, getUboManager());
+                mi->commit(driver);
                 mi->use(driver);
 
                 auto pipeline = getPipelineState(ma);
@@ -1899,7 +1895,7 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::dof(FrameGraph& fg,
                             .filterMin = SamplerMinFilter::NEAREST_MIPMAP_NEAREST });
                     mi->setParameter("weightScale", 0.5f / float(1u << level));   // FIXME: halfres?
                     mi->setParameter("texelSize", float2{ 1.0f / w, 1.0f / h });
-                    mi->commit(driver, getUboManager());
+                    mi->commit(driver);
                     mi->use(driver);
 
                     renderFullScreenQuad(out, pipeline, driver);
@@ -2429,7 +2425,7 @@ PostProcessManager::BloomPassOutput PostProcessManager::bloom(FrameGraph& fg,
                     mi->setParameter("source", hwOutView, SamplerParams{
                             .filterMag = SamplerMagFilter::LINEAR,
                             .filterMin = SamplerMinFilter::LINEAR_MIPMAP_NEAREST});
-                    mi->commit(driver, getUboManager());
+                    mi->commit(driver);
                     mi->use(driver);
                     renderFullScreenQuad(hwDstRT, pipeline, driver);
                     DescriptorSet::unbind(driver, DescriptorSetBindingPoints::PER_MATERIAL);
@@ -2537,7 +2533,7 @@ void PostProcessManager::colorGradingPrepareSubpass(DriverApi& driver,
     FMaterialInstance* const mi =
             configureColorGradingMaterial(material, colorGrading, colorGradingConfig,
                     vignetteOptions, width, height);
-    mi->commit(driver, getUboManager());
+    mi->commit(driver);
 }
 
 void PostProcessManager::colorGradingSubpass(DriverApi& driver,
@@ -2570,7 +2566,7 @@ void PostProcessManager::customResolvePrepareSubpass(DriverApi& driver, CustomRe
     auto [mi, fixedIndex] = mMaterialInstanceManager.getFixedMaterialInstance(ma);
     mFixedMaterialInstanceIndex.customResolve = fixedIndex;
     mi->setParameter("direction", op == CustomResolveOp::COMPRESS ? 1.0f : -1.0f),
-    mi->commit(driver, getUboManager());
+    mi->commit(driver);
     material.getMaterial(mEngine);
 }
 
@@ -2623,7 +2619,7 @@ void PostProcessManager::clearAncillaryBuffersPrepare(DriverApi& driver) noexcep
     auto ma = material.getMaterial(mEngine, PostProcessVariant::OPAQUE);
     auto [mi, fixedIndex] = mMaterialInstanceManager.getFixedMaterialInstance(ma);
     mFixedMaterialInstanceIndex.clearDepth = fixedIndex;
-    mi->commit(driver, getUboManager());
+    mi->commit(driver);
     material.getMaterial(mEngine);
 }
 
@@ -3115,7 +3111,7 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::taa(FrameGraph& fg,
                         mat4f{ historyProjection * inverse(current.projection) } *
                         normalizedToClip);
 
-                mi->commit(driver, getUboManager());
+                mi->commit(driver);
                 mi->use(driver);
 
                 if (colorGradingConfig.asSubpass) {
@@ -3200,7 +3196,7 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::rcas(
                 mi->setParameter("resolution", float4{
                         outputDesc.width, outputDesc.height,
                         1.0f / outputDesc.width, 1.0f / outputDesc.height });
-                mi->commit(driver, getUboManager());
+                mi->commit(driver);
                 mi->use(driver);
 
                 auto pipeline = getPipelineState(material.getMaterial(mEngine), variant);
@@ -3291,7 +3287,7 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::upscaleBilinear(FrameGraph&
                         float(vp.width)  / inputDesc.width,
                         float(vp.height) / inputDesc.height
                 });
-                mi->commit(driver, getUboManager());
+                mi->commit(driver);
                 mi->use(driver);
 
                 auto out = resources.getRenderPassInfo();
@@ -3378,7 +3374,7 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::upscaleSGSR1(FrameGraph& fg,
                         float(inputDesc.height)
                 });
 
-                mi->commit(driver, getUboManager());
+                mi->commit(driver);
                 mi->use(driver);
 
                 auto const out = resources.getRenderPassInfo();
@@ -3473,7 +3469,7 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::upscaleFSR1(FrameGraph& fg,
                     mi->setParameter("resolution",
                             float4{ outputDesc.width, outputDesc.height,
                                     1.0f / outputDesc.width, 1.0f / outputDesc.height });
-                    mi->commit(driver, getUboManager());
+                    mi->commit(driver);
                     mi->use(driver);
                 }
 
@@ -3500,7 +3496,7 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::upscaleFSR1(FrameGraph& fg,
                             float(vp.width)  / inputDesc.width,
                             float(vp.height) / inputDesc.height
                     });
-                    mi->commit(driver, getUboManager());
+                    mi->commit(driver);
                     mi->use(driver);
                 }
 
@@ -3586,7 +3582,7 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::blit(FrameGraph& fg, bool co
                 if (layer) {
                     mi->setParameter("layerIndex", layer);
                 }
-                mi->commit(driver, getUboManager());
+                mi->commit(driver);
                 mi->use(driver);
 
                 auto pipeline = getPipelineState(ma);
@@ -3844,7 +3840,7 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::vsmMipmapPass(FrameGraph& fg
                 });
                 mi->setParameter("layer", uint32_t(layer));
                 mi->setParameter("uvscale", 1.0f / float(dim));
-                mi->commit(driver, getUboManager());
+                mi->commit(driver);
                 mi->use(driver);
 
                 renderFullScreenQuadWithScissor(out, pipeline, scissor, driver);
@@ -3946,7 +3942,7 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::debugCombineArrayTexture(Fra
                         float(vp.width) / inputDesc.width,
                         float(vp.height) / inputDesc.height
                     });
-                mi->commit(driver, getUboManager());
+                mi->commit(driver);
                 mi->use(driver);
 
                 auto pipeline = getPipelineState(ma);
@@ -3963,7 +3959,7 @@ FrameGraphId<FrameGraphTexture> PostProcessManager::debugCombineArrayTexture(Fra
                 // Render all layers of the texture to the screen side-by-side.
                 for (uint32_t i = 0; i < inputTextureDesc.depth; ++i) {
                     mi->setParameter("layerIndex", i);
-                    mi->commit(driver, getUboManager());
+                    mi->commit(driver);
                     renderFullScreenQuad(out, pipeline, driver);
                     DescriptorSet::unbind(driver, DescriptorSetBindingPoints::PER_MATERIAL);
                     // From the second draw, don't clear the targetbuffer.

filament/src/PostProcessManager.h

@@ -65,7 +65,6 @@ class FMaterialInstance;
 class FrameGraph;
 class RenderPass;
 class RenderPassBuilder;
-class UboManager;
 struct CameraInfo;
 
 class PostProcessManager {
@@ -423,8 +422,6 @@ private:
         return getMaterialInstance(ma);
     }
 
-    UboManager* getUboManager() const noexcept;
-
     backend::RenderPrimitiveHandle mFullScreenQuadRph;
     backend::VertexBufferInfoHandle mFullScreenQuadVbih;
     backend::DescriptorSetLayoutHandle mPerRenderableDslh;

filament/src/RenderPass.cpp

@@ -1012,8 +1012,6 @@ void RenderPass::Executor::execute(FEngine const& engine, DriverApi& driver,
                     uint32_t const index = (first->key & CUSTOM_INDEX_MASK) >> CUSTOM_INDEX_SHIFT;
                     assert_invariant(index < mCustomCommands.size());
                     pCustomCommands[index]();
-                    currentPipeline = {};
-                    currentPrimitiveHandle ={};
                     continue;
                 }
 
@@ -1071,34 +1069,26 @@ void RenderPass::Executor::execute(FEngine const& engine, DriverApi& driver,
                     pipeline.pipelineLayout.setLayout[+DescriptorSetBindingPoints::PER_MATERIAL] =
                             ma->getDescriptorSetLayout(info.materialVariant).getHandle();
 
-                    // If we have a ColorPassDescriptorSet we use it to bind the per-view
-                    // descriptor-set (ideally only if it changed).
-                    // If we don't, it means the descriptor-set is already bound and the layout we
-                    // got from the material above should match. This is the case for situations
-                    // where we have a known per-view descriptor-set layout,
-                    // e.g.: postfx, shadow-maps, ssr and structure passes.
-                    if (mColorPassDescriptorSet) {
-                        if (UTILS_UNLIKELY(ma->getMaterialDomain() == MaterialDomain::POST_PROCESS)) {
-                            // It is possible to get a post-process material here (even though it's
-                            // not technically a public API yet, it is used by the IBLPrefilterLibrary.
-                            // Ideally we would have a more formal compute API). In this case, we need
-                            // to set the post-process descriptor-set.
-                            engine.getPostProcessManager().bindPostProcessDescriptorSet(driver);
-                        } else {
+
+                    if (UTILS_UNLIKELY(ma->getMaterialDomain() == MaterialDomain::POST_PROCESS)) {
+                        // It is possible to get a post-process material here (even though it's
+                        // not technically a public API yet, it is used by the IBLPrefilterLibrary.
+                        // Ideally we would have a more formal compute API). In this case, we need
+                        // to set the post-process descriptor-set.
+                        engine.getPostProcessManager().bindPostProcessDescriptorSet(driver);
+                    } else {
+                        // If we have a ColorPassDescriptorSet we use it to bind the per-view
+                        // descriptor-set (ideally only if it changed). If we don't, it means
+                        // the descriptor-set is already bound and the layout we got from the
+                        // material above should match. This is the case for situations where we
+                        // have a known per-view descriptor-set layout, e.g.: shadow-maps, ssr and
+                        // structure passes.
+                        if (mColorPassDescriptorSet) {
                             // We have a ColorPassDescriptorSet, we need to go through it for binding
                             // the per-view descriptor-set because its layout can change based on the
                             // material.
                             mColorPassDescriptorSet->bind(driver, ma->getPerViewLayoutIndex());
                         }
-                    } else {
-                        // if we're here it means the per-view descriptor set is constant and
-                        // already set. This will be the case for postfx, ssr, structure and
-                        // shadow passes. All these passes use a static descriptor set layout
-                        // (albeit potentially different for each pass). In particular the
-                        // per-view UBO must be compatible for all material domains.
-                        // This is the case by construction for postfx, ssr. However, shadows
-                        // and structure have their own UBO, but it's content is (must be)
-                        // compatible with POST_PROCESS and COMPUTE materials.
                     }
 
                     // Each MaterialInstance has its own descriptor set. This binds it.

filament/src/details/BufferAllocator.cpp

@@ -142,11 +142,7 @@ void BufferAllocator::retire(AllocationId id) {
     InternalSlotNode* targetNode = getNodeById(id);
     assert_invariant(targetNode != nullptr);
 
-    Slot& slot = targetNode->slot;
-    slot.isAllocated = false;
-    if (slot.gpuUseCount == 0) {
-        mHasPendingFrees = true;
-    }
+    targetNode->slot.isAllocated = false;
 }
 
 void BufferAllocator::acquireGpu(AllocationId id) {
@@ -161,18 +157,10 @@ void BufferAllocator::releaseGpu(AllocationId id) {
     assert_invariant(targetNode != nullptr);
     assert_invariant(targetNode->slot.gpuUseCount > 0);
 
-    Slot& slot = targetNode->slot;
-    slot.gpuUseCount--;
-    if (slot.gpuUseCount == 0 && !slot.isAllocated) {
-        mHasPendingFrees = true;
-    }
+    targetNode->slot.gpuUseCount--;
 }
 
 void BufferAllocator::releaseFreeSlots() {
-    if (!mHasPendingFrees) {
-        return;
-    }
-
     auto curr = mSlotPool.begin();
     while (curr != mSlotPool.end()) {
         if (!curr->slot.isFree()) {
@@ -211,7 +199,6 @@ void BufferAllocator::releaseFreeSlots() {
 
         curr = next;
     }
-    mHasPendingFrees = false;
 }
 
 BufferAllocator::allocation_size_t BufferAllocator::getTotalSize() const noexcept {

filament/src/details/BufferAllocator.h

@@ -113,9 +113,8 @@ private:
 
     [[nodiscard]] InternalSlotNode* getNodeById(AllocationId id) const noexcept;
 
-    bool mHasPendingFrees = false;
     allocation_size_t mTotalSize;
-    const allocation_size_t mSlotSize; // Size of a single slot in bytes
+    const allocation_size_t mSlotSize; // Size of a single slot in bytes.
     std::list<InternalSlotNode> mSlotPool; // All slots, including both allocated and freed
     std::multimap</*slot size*/allocation_size_t, InternalSlotNode*> mFreeList;
     std::unordered_map</*slot offset*/allocation_size_t, InternalSlotNode*> mOffsetMap;

filament/src/details/Engine.cpp

@@ -468,26 +468,11 @@ void FEngine::init() {
         }
         mDefaultMaterial = downcast(defaultMaterialBuilder.build(*this));
     }
-
     // We must commit the default material instance here. It may not be used in a scene, but its
     // descriptor set may still be used for shared variants.
-    //
-    // Note that this material instance is instantiated before the creation of UboManager, so at
-    // this point `isUboBatchingEnabled` is `false`, and it will fall back to individual UBO
-    // automatically.
-    mDefaultMaterial->getDefaultInstance()->commit(driverApi, mUboManager);
+    mDefaultMaterial->getDefaultInstance()->commit(driverApi);
 
     if (UTILS_UNLIKELY(getSupportedFeatureLevel() >= FeatureLevel::FEATURE_LEVEL_1)) {
-        // UBO batching is not supported in feature level 0
-        if (features.material.enable_material_instance_uniform_batching) {
-            // Ubo size of each material instance is at least 16 bytes.
-            constexpr BufferAllocator::allocation_size_t minSlotSize = 16;
-            auto const uboOffsetAlignment = static_cast<BufferAllocator::allocation_size_t>(
-                    driverApi.getUniformBufferOffsetAlignment());
-            BufferAllocator::allocation_size_t slotSize = std::max(minSlotSize, uboOffsetAlignment);
-            mUboManager = new UboManager(getDriverApi(), slotSize, mConfig.sharedUboInitialSizeInBytes);
-        }
-
         mDefaultColorGrading = downcast(ColorGrading::Builder().build(*this));
 
         constexpr float3 dummyPositions[1] = {};
@@ -668,12 +653,6 @@ void FEngine::shutdown() {
 
     driver.destroyRenderTarget(std::move(mDefaultRenderTarget));
 
-    if (isUboBatchingEnabled()) {
-        mUboManager->terminate(driver);
-        delete mUboManager;
-        mUboManager = nullptr;
-    }
-
     /*
      * Shutdown the backend...
      */
@@ -714,31 +693,18 @@ void FEngine::prepare() {
     // UBOs that are visible only. It's not such a big issue because the actual upload() is
     // skipped if the UBO hasn't changed. Still we could have a lot of these.
     DriverApi& driver = getDriverApi();
-    const bool useUboBatching = isUboBatchingEnabled();
 
-    if (useUboBatching) {
-        assert_invariant(mUboManager != nullptr);
-
-        mUboManager->beginFrame(driver, mMaterialInstances);
-    }
-
-    UboManager* uboManager = mUboManager;
     for (auto& materialInstanceList: mMaterialInstances) {
-        materialInstanceList.second.forEach([&driver, uboManager](FMaterialInstance* item) {
+        materialInstanceList.second.forEach([&driver](FMaterialInstance* item) {
             // post-process materials instances must be commited explicitly because their
             // parameters are typically not set at this point in time.
             if (item->getMaterial()->getMaterialDomain() == MaterialDomain::SURFACE) {
                 item->commitStreamUniformAssociations(driver);
-                item->commit(driver, uboManager);
+                item->commit(driver);
             }
         });
     }
 
-    if (useUboBatching) {
-        assert_invariant(mUboManager != nullptr);
-        getUboManager()->finishBeginFrame(getDriverApi());
-    }
-
     mMaterials.forEach([](FMaterial* material) {
 #if FILAMENT_ENABLE_MATDBG // NOLINT(*-include-cleaner)
         material->checkProgramEdits();
@@ -755,13 +721,6 @@ void FEngine::gc() {
     mCameraManager.gc(*this, em);
 }
 
-void FEngine::submitFrame() {
-    if (isUboBatchingEnabled()) {
-        DriverApi& driver = getDriverApi();
-        getUboManager()->endFrame(driver, getMaterialInstanceResourceList());
-    }
-}
-
 void FEngine::flush() {
     // flush the command buffer
     flushCommandBuffer(mCommandBufferQueue);
@@ -992,10 +951,9 @@ FMaterialInstance* FEngine::createMaterialInstance(const FMaterial* material,
     return p;
 }
 
-FMaterialInstance* FEngine::createMaterialInstance(const FMaterial* material, const char* name,
-        UboBatchingMode batchingMode) noexcept {
-    FMaterialInstance* p =
-            mHeapAllocator.make<FMaterialInstance>(*this, material, name, batchingMode);
+FMaterialInstance* FEngine::createMaterialInstance(const FMaterial* material,
+                                                   const char* name) noexcept {
+    FMaterialInstance* p = mHeapAllocator.make<FMaterialInstance>(*this, material, name);
     if (UTILS_LIKELY(p)) {
         auto pos = mMaterialInstances.emplace(material, "MaterialInstance");
         pos.first->second.insert(p);
@@ -1285,11 +1243,6 @@ UTILS_NOINLINE
 bool FEngine::destroy(const FMaterialInstance* p) {
     if (p == nullptr) return true;
 
-    if (p->isUsingUboBatching()) {
-        assert_invariant(isUboBatchingEnabled());
-        mUboManager->retireSlot(p->getAllocationId());
-    }
-
     // Check that the material instance we're destroying is not in use in the RenderableManager
     // To do this, we currently need to inspect all render primitives in the RenderableManager
     EntityManager const& em = mEntityManager;

filament/src/details/Engine.h

@@ -21,17 +21,16 @@
 
 #include "Allocators.h"
 #include "DFG.h"
+#include "PostProcessManager.h"
+#include "ResourceList.h"
 #include "HwDescriptorSetLayoutFactory.h"
 #include "HwVertexBufferInfoFactory.h"
 #include "MaterialCache.h"
-#include "PostProcessManager.h"
-#include "ResourceList.h"
-#include "UboManager.h"
 
 #include "components/CameraManager.h"
 #include "components/LightManager.h"
-#include "components/RenderableManager.h"
 #include "components/TransformManager.h"
+#include "components/RenderableManager.h"
 
 #include "ds/DescriptorSetLayout.h"
 
@@ -264,14 +263,6 @@ public:
         return mBackend;
     }
 
-    bool isUboBatchingEnabled() const noexcept {
-        return mUboManager != nullptr;
-    }
-
-    UboManager* getUboManager() noexcept {
-        return mUboManager;
-    }
-
     Platform* getPlatform() const noexcept {
         return mPlatform;
     }
@@ -345,21 +336,11 @@ public:
 
     FRenderer* createRenderer() noexcept;
 
-    // Defines whether a material instance should use UBO batching or not.
-    enum class UboBatchingMode {
-        // For default, it follows the engine settings.
-        // If UBO batching is enabled on the engine and the material domain is not SURFACE, it
-        // turns on the UBO batching. Otherwise, it turns off the UBO batching.
-        DEFAULT,
-        NO_UBO_BATCHING,
-        UBO_BATCHING
-    };
-
     FMaterialInstance* createMaterialInstance(const FMaterial* material,
             const FMaterialInstance* other, const char* name) noexcept;
 
-    FMaterialInstance* createMaterialInstance(const FMaterial* material, const char* name,
-            UboBatchingMode batchingMode) noexcept;
+    FMaterialInstance* createMaterialInstance(const FMaterial* material,
+                                              const char* name) noexcept;
 
     FScene* createScene() noexcept;
     FView* createView() noexcept;
@@ -465,7 +446,6 @@ public:
 
     void prepare();
     void gc();
-    void submitFrame();
 
     using ShaderContent = utils::FixedCapacityVector<uint8_t>;
 
@@ -670,7 +650,6 @@ private:
 
     uint32_t mFlushCounter = 0;
 
-    UboManager* mUboManager = nullptr;
     RootArenaScope::Arena mPerRenderPassArena;
     HeapAllocatorArena mHeapAllocator;
 

filament/src/details/Material.cpp

@@ -73,7 +73,6 @@ namespace filament {
 using namespace backend;
 using namespace filaflat;
 using namespace utils;
-using UboBatchingMode = FEngine::UboBatchingMode;
 
 struct Material::BuilderDetails {
     const void* mPayload = nullptr;
@@ -218,17 +217,16 @@ void FMaterial::terminate(FEngine& engine) {
 
 filament::DescriptorSetLayout const& FMaterial::getPerViewDescriptorSetLayout(
         Variant const variant, bool const useVsmDescriptorSetLayout) const noexcept {
-    if (mDefinition.materialDomain == MaterialDomain::SURFACE) {
-        // `variant` is only sensical for MaterialDomain::SURFACE
-        if (Variant::isValidDepthVariant(variant)) {
-            return mEngine.getPerViewDescriptorSetLayoutDepthVariant();
-        }
-        if (Variant::isSSRVariant(variant)) {
-            return mEngine.getPerViewDescriptorSetLayoutSsrVariant();
-        }
+    if (Variant::isValidDepthVariant(variant)) {
+        assert_invariant(mDefinition.materialDomain == MaterialDomain::SURFACE);
+        return mEngine.getPerViewDescriptorSetLayoutDepthVariant();
+    }
+    if (Variant::isSSRVariant(variant)) {
+        assert_invariant(mDefinition.materialDomain == MaterialDomain::SURFACE);
+        return mEngine.getPerViewDescriptorSetLayoutSsrVariant();
     }
-    // mDefinition.perViewDescriptorSetLayout{Vsm} is already resolved for MaterialDomain
     if (useVsmDescriptorSetLayout) {
+        assert_invariant(mDefinition.materialDomain == MaterialDomain::SURFACE);
         return mDefinition.perViewDescriptorSetLayoutVsm;
     }
     return mDefinition.perViewDescriptorSetLayout;
@@ -282,14 +280,14 @@ FMaterialInstance* FMaterial::createInstance(const char* name) const noexcept {
         return FMaterialInstance::duplicate(mDefaultMaterialInstance, name);
     } else {
         // but if we don't, just create an instance with all the default parameters
-        return mEngine.createMaterialInstance(this, name, UboBatchingMode::DEFAULT);
+        return mEngine.createMaterialInstance(this, name);
     }
 }
 
 FMaterialInstance* FMaterial::getDefaultInstance() noexcept {
     if (UTILS_UNLIKELY(!mDefaultMaterialInstance)) {
-        mDefaultMaterialInstance =
-                mEngine.createMaterialInstance(this, mDefinition.name.c_str(), UboBatchingMode::DEFAULT);
+        mDefaultMaterialInstance = mEngine.createMaterialInstance(this,
+                mDefinition.name.c_str());
         mDefaultMaterialInstance->setDefaultInstance(true);
     }
     return mDefaultMaterialInstance;

filament/src/details/Material.h

@@ -108,7 +108,7 @@ public:
             backend::CallbackHandler* handler,
             utils::Invocable<void(Material*)>&& callback) noexcept;
 
-    // Creates an instance of this material, specifying the batching mode.
+    // Create an instance of this material
     FMaterialInstance* createInstance(const char* name) const noexcept;
 
     bool hasParameter(const char* name) const noexcept;

filament/src/details/MaterialInstance.cpp

@@ -58,19 +58,15 @@ using namespace utils;
 namespace filament {
 
 using namespace backend;
-using UboBatchingMode = FEngine::UboBatchingMode;
 
-bool shouldEnableBatching(FEngine& engine, UboBatchingMode batchingMode, MaterialDomain domain) {
-    if (batchingMode != UboBatchingMode::DEFAULT) {
-        return batchingMode == UboBatchingMode::UBO_BATCHING;
-    }
-
-    return engine.isUboBatchingEnabled() && domain == MaterialDomain::SURFACE;
+FMaterialInstance::FMaterialInstance(FEngine& engine, FMaterial const* material,
+        const char* name) noexcept
+        : FMaterialInstance(engine, material, name,
+                  engine.features.material.enable_material_instance_uniform_batching) {
 }
 
-FMaterialInstance::FMaterialInstance(FEngine& engine, FMaterial const* material, const char* name,
-        UboBatchingMode batchingMode) noexcept
-        : mMaterial(material),
+FMaterialInstance::FMaterialInstance(FEngine& engine, FMaterial const* material,
+                      const char* name, bool useUboBatching) noexcept : mMaterial(material),
           mDescriptorSet("MaterialInstance", material->getDescriptorSetLayout()),
           mCulling(CullingMode::BACK),
           mShadowCulling(CullingMode::BACK),
@@ -80,12 +76,10 @@ FMaterialInstance::FMaterialInstance(FEngine& engine, FMaterial const* material,
           mHasScissor(false),
           mIsDoubleSided(false),
           mIsDefaultInstance(false),
-          mUseUboBatching(
-                  shouldEnableBatching(engine, batchingMode, material->getMaterialDomain())),
+          mUseUboBatching(useUboBatching),
           mTransparencyMode(TransparencyMode::DEFAULT),
           mName(name ? CString(name) : material->getName()) {
-    FILAMENT_CHECK_PRECONDITION(!mUseUboBatching || engine.isUboBatchingEnabled())
-            << "UBO batching is not enabled.";
+
     FEngine::DriverApi& driver = engine.getDriverApi();
 
     // even if the material doesn't have any parameters, we allocate a small UBO because it's
@@ -159,7 +153,7 @@ FMaterialInstance::FMaterialInstance(FEngine& engine,
           mUseUboBatching(other->mUseUboBatching),
           mScissorRect(other->mScissorRect),
           mName(name ? CString(name) : other->mName) {
-    assert_invariant(!mUseUboBatching || engine.isUboBatchingEnabled());
+
     FEngine::DriverApi& driver = engine.getDriverApi();
     FMaterial const* const material = other->getMaterial();
 
@@ -199,8 +193,8 @@ FMaterialInstance::FMaterialInstance(FEngine& engine,
     }
 }
 
-FMaterialInstance* FMaterialInstance::duplicate(FMaterialInstance const* other,
-        const char* name) noexcept {
+FMaterialInstance* FMaterialInstance::duplicate(
+        FMaterialInstance const* other, const char* name) noexcept {
     FMaterial const* const material = other->getMaterial();
     FEngine& engine = material->getEngine();
     return engine.createMaterialInstance(material, other, name);
@@ -244,21 +238,15 @@ void FMaterialInstance::commitStreamUniformAssociations(FEngine::DriverApi& driv
 
 void FMaterialInstance::commit(FEngine& engine) const {
     if (UTILS_LIKELY(mMaterial->getMaterialDomain() != MaterialDomain::SURFACE)) {
-        commit(engine.getDriverApi(), engine.getUboManager());
+        commit(engine.getDriverApi());
     }
 }
 
-void FMaterialInstance::commit(FEngine::DriverApi& driver, UboManager* uboManager) const {
+void FMaterialInstance::commit(FEngine::DriverApi& driver) const {
     if (mUniforms.isDirty() || mHasStreamUniformAssociations) {
         mUniforms.clean();
         if (mUseUboBatching) {
-            assert_invariant(uboManager != nullptr);
-            if (!BufferAllocator::isValid(getAllocationId())) {
-                // The allocation hasn't happened yet, return.
-                return;
-            }
-
-            uboManager->updateSlot(driver, getAllocationId(), mUniforms.toBufferDescriptor(driver));
+            // TODO: update the content by `copyToMemoryMappedBuffer`
         }
         else {
             auto* ubHandle = std::get_if<Handle<HwBufferObject>>(&mUboData);
@@ -283,10 +271,6 @@ void FMaterialInstance::commit(FEngine::DriverApi& driver, UboManager* uboManage
     // TODO: eventually we should remove this in RELEASE builds
     fixMissingSamplers();
 
-    if (mUseUboBatching && !BufferAllocator::isValid(getAllocationId())) {
-        return;
-    }
-
     // Commit descriptors if needed (e.g. when textures are updated,or the first time)
     mDescriptorSet.commit(mMaterial->getDescriptorSetLayout(), driver);
 }
@@ -429,13 +413,6 @@ const char* FMaterialInstance::getName() const noexcept {
 // ------------------------------------------------------------------------------------------------
 
 void FMaterialInstance::use(FEngine::DriverApi& driver, Variant variant) const {
-    if (!mDescriptorSet.getHandle()) {
-        return;
-    }
-
-    if (mUseUboBatching && !BufferAllocator::isValid(getAllocationId())) {
-        return;
-    }
 
     if (UTILS_UNLIKELY(mMissingSamplerDescriptors.any())) {
         std::call_once(mMissingSamplersFlag, [this] {
@@ -462,8 +439,7 @@ void FMaterialInstance::use(FEngine::DriverApi& driver, Variant variant) const {
         return;
     }
 
-    mDescriptorSet.bind(driver, DescriptorSetBindingPoints::PER_MATERIAL,
-            { { mUboOffset }, driver });
+    mDescriptorSet.bind(driver, DescriptorSetBindingPoints::PER_MATERIAL);
 }
 
 void FMaterialInstance::assignUboAllocation(
@@ -473,10 +449,8 @@ void FMaterialInstance::assignUboAllocation(
     assert_invariant(mUseUboBatching);
 
     mUboData = id;
-    mUboOffset = offset;
     if (BufferAllocator::isValid(id)) {
-        // Use dynamic offset during binding, so the offset here is always zero.
-        mDescriptorSet.setBuffer(mMaterial->getDescriptorSetLayout(), 0, ubHandle, 0,
+        mDescriptorSet.setBuffer(mMaterial->getDescriptorSetLayout(), 0, ubHandle, offset,
                 mUniforms.getSize());
     }
 }

filament/src/details/MaterialInstance.h

@@ -56,8 +56,11 @@ class FTexture;
 
 class FMaterialInstance : public MaterialInstance {
 public:
-    FMaterialInstance(FEngine& engine, FMaterial const* material, const char* name,
-            FEngine::UboBatchingMode batchingMode) noexcept;
+    FMaterialInstance(FEngine& engine, FMaterial const* material,
+                      const char* name) noexcept;
+    // Use this constructor when you need to override the ubo batching flag for an individual MI.
+    FMaterialInstance(FEngine& engine, FMaterial const* material,
+                      const char* name, bool useUboBatching) noexcept;
     FMaterialInstance(FEngine& engine, FMaterialInstance const* other, const char* name);
     FMaterialInstance(const FMaterialInstance& rhs) = delete;
     FMaterialInstance& operator=(const FMaterialInstance& rhs) = delete;
@@ -72,7 +75,7 @@ public:
     
     void commit(FEngine& engine) const;
 
-    void commit(FEngine::DriverApi& driver, UboManager* uboManager) const;
+    void commit(FEngine::DriverApi& driver) const;
 
     void use(FEngine::DriverApi& driver, Variant variant = {}) const;
 
@@ -284,7 +287,6 @@ private:
     };
 
     std::variant<BufferAllocator::AllocationId, backend::Handle<backend::HwBufferObject>> mUboData;
-    BufferAllocator::allocation_size_t mUboOffset = 0;
     tsl::robin_map<backend::descriptor_binding_t, TextureParameter> mTextureParameters;
     mutable DescriptorSet mDescriptorSet;
     UniformBuffer mUniforms;
@@ -306,7 +308,7 @@ private:
     bool mHasScissor : 1;
     bool mIsDoubleSided : 1;
     bool mIsDefaultInstance : 1;
-    const bool mUseUboBatching : 1;
+    bool mUseUboBatching : 1;
     TransparencyMode mTransparencyMode : 2;
 
     uint64_t mMaterialSortingKey = 0;

filament/src/details/Renderer.cpp

@@ -452,8 +452,6 @@ void FRenderer::endFrame() {
     mFrameInfoManager.endFrame(driver);
     mFrameSkipper.submitFrame(driver);
 
-    engine.submitFrame();
-
     driver.endFrame(mFrameId);
 
     // gives the backend a chance to execute periodic tasks
@@ -589,8 +587,6 @@ void FRenderer::renderStandaloneView(FView const* view) {
         // happen with Renderer::beginFrame/endFrame.
         renderInternal(view, true);
 
-        engine.submitFrame();
-
         driver.endFrame(mFrameId);
 
         // engine.flush() has already been called by renderInternal(), but we need an extra one

filament/src/details/VertexBuffer.cpp

@@ -302,7 +302,7 @@ FVertexBuffer::FVertexBuffer(FEngine& engine, const Builder& builder)
     if (!mBufferObjectsEnabled) {
         // If buffer objects are not enabled at the API level, then we create them internally.
         #pragma nounroll
-        for (size_t index = 0; index < MAX_VERTEX_ATTRIBUTE_COUNT; ++index) {
+        for (size_t index = 0; index < MAX_VERTEX_BUFFER_COUNT; ++index) {
             size_t const i = mAttributes[index].buffer;
             if (i != Attribute::BUFFER_UNUSED) {
                 assert_invariant(bufferSizes[i] > 0);

filament/src/ds/PostProcessDescriptorSet.cpp

@@ -39,7 +39,7 @@ void PostProcessDescriptorSet::init(FEngine& engine) noexcept {
     // create the descriptor-set layout
     mDescriptorSetLayout = filament::DescriptorSetLayout{
             engine.getDescriptorSetLayoutFactory(),
-            engine.getDriverApi(), descriptor_sets::getDepthVariantLayout() };
+            engine.getDriverApi(), descriptor_sets::getPostProcessLayout() };
 
     // create the descriptor-set from the layout
     mDescriptorSet = DescriptorSet{ "PostProcessDescriptorSet", mDescriptorSetLayout };

libs/filabridge/include/private/filament/DescriptorSets.h

@@ -28,9 +28,9 @@
 
 namespace filament::descriptor_sets {
 
+backend::DescriptorSetLayout const& getPostProcessLayout() noexcept;
 backend::DescriptorSetLayout const& getDepthVariantLayout() noexcept;
 backend::DescriptorSetLayout const& getSsrVariantLayout() noexcept;
-
 backend::DescriptorSetLayout const& getPerRenderableLayout() noexcept;
 
 backend::DescriptorSetLayout getPerViewDescriptorSetLayout(

libs/filabridge/src/DescriptorSets.cpp

@@ -40,7 +40,12 @@ namespace filament::descriptor_sets {
 
 using namespace backend;
 
-// used to generate shadow-maps, structure and postfx passes
+// used for post-processing passes
+static constexpr std::initializer_list<DescriptorSetLayoutBinding> postProcessDescriptorSetLayoutList = {
+    { DescriptorType::UNIFORM_BUFFER, ShaderStageFlags::VERTEX | ShaderStageFlags::FRAGMENT,  +PerViewBindingPoints::FRAME_UNIFORMS },
+};
+
+// used to generate shadow-maps
 static constexpr std::initializer_list<DescriptorSetLayoutBinding> depthVariantDescriptorSetLayoutList = {
     { DescriptorType::UNIFORM_BUFFER, ShaderStageFlags::VERTEX | ShaderStageFlags::FRAGMENT,  +PerViewBindingPoints::FRAME_UNIFORMS },
 };
@@ -137,6 +142,10 @@ static const std::unordered_map<
         {{ SamplerType::SAMPLER_EXTERNAL, SamplerFormat::FLOAT }, DescriptorType::SAMPLER_EXTERNAL }
 };
 
+// used for post-processing passes
+static DescriptorSetLayout const postProcessDescriptorSetLayout{ utils::StaticString("postProcess"),
+    postProcessDescriptorSetLayoutList };
+
 // used to generate shadow-maps
 static DescriptorSetLayout const depthVariantDescriptorSetLayout{
     utils::StaticString("depthVariant"), depthVariantDescriptorSetLayoutList
@@ -154,6 +163,10 @@ static DescriptorSetLayout perRenderableDescriptorSetLayout = {
     utils::StaticString("perRenderable"), perRenderableDescriptorSetLayoutList
 };
 
+DescriptorSetLayout const& getPostProcessLayout() noexcept {
+    return postProcessDescriptorSetLayout;
+}
+
 DescriptorSetLayout const& getDepthVariantLayout() noexcept {
     return depthVariantDescriptorSetLayout;
 }
@@ -267,10 +280,10 @@ DescriptorSetLayout getPerViewDescriptorSetLayout(
             return layout;
         }
         case MaterialDomain::POST_PROCESS:
-            return depthVariantDescriptorSetLayout;
+            return postProcessDescriptorSetLayout;
         case MaterialDomain::COMPUTE:
             // TODO: what's the layout for compute?
-            return depthVariantDescriptorSetLayout;
+            return postProcessDescriptorSetLayout;
     }
 }
 

libs/filamat/src/GLSLPostProcessor.cpp

@@ -129,9 +129,9 @@ DescriptorSetLayout getPerMaterialDescriptorSet(SamplerInterfaceBlock const& sib
     DescriptorSetLayout layout;
     layout.bindings.reserve(1 + samplers.size());
 
-    layout.bindings.push_back(DescriptorSetLayoutBinding{ DescriptorType::UNIFORM_BUFFER,
-        ShaderStageFlags::VERTEX | ShaderStageFlags::FRAGMENT,
-        +PerMaterialBindingPoints::MATERIAL_PARAMS, DescriptorFlags::DYNAMIC_OFFSET, 0 });
+    layout.bindings.push_back(DescriptorSetLayoutBinding { DescriptorType::UNIFORM_BUFFER,
+            ShaderStageFlags::VERTEX | ShaderStageFlags::FRAGMENT,
+            +PerMaterialBindingPoints::MATERIAL_PARAMS, DescriptorFlags::NONE, 0 });
 
     for (auto const& sampler: samplers) {
         DescriptorSetLayoutBinding layoutBinding{

libs/filamat/src/eiff/MaterialInterfaceBlockChunk.cpp

@@ -225,7 +225,7 @@ void MaterialDescriptorSetLayoutChunk::flatten(Flattener& f) {
     f.writeUint8(uint8_t(DescriptorType::UNIFORM_BUFFER));
     f.writeUint8(uint8_t(ShaderStageFlags::VERTEX | ShaderStageFlags::FRAGMENT));
     f.writeUint8(0);
-    f.writeUint8(uint8_t(DescriptorFlags::DYNAMIC_OFFSET));
+    f.writeUint8(uint8_t(DescriptorFlags::NONE));
     f.writeUint16(0);
 
     // all the material's sampler descriptors

libs/matdbg/src/TextWriter.cpp

@@ -281,7 +281,7 @@ static bool printParametersInfo(ostream& text, const ChunkContainer& container)
         }
 
         text << "    "
-                  << setw(alignment) << fieldName.c_str_safe()
+                  << setw(alignment) << fieldName.c_str()
                   << setw(shortAlignment) << toString(UniformType(fieldType))
                   << arraySizeToString(fieldSize)
                   << setw(shortAlignment) << toString(Precision(fieldPrecision))
@@ -325,7 +325,7 @@ static bool printParametersInfo(ostream& text, const ChunkContainer& container)
         }
 
         text << "    "
-                << setw(alignment) << fieldName.c_str_safe()
+                << setw(alignment) << fieldName.c_str()
                 << setw(shortAlignment) << +fieldBinding
                 << setw(shortAlignment) << toString(SamplerType(fieldType))
                 << setw(shortAlignment) << toString(Precision(fieldPrecision))
@@ -364,7 +364,7 @@ static bool printConstantInfo(ostream& text, const ChunkContainer& container) {
         }
 
          text << "    "
-         << setw(alignment) << fieldName.c_str_safe()
+         << setw(alignment) << fieldName.c_str()
          << setw(shortAlignment) << toString(ConstantType(fieldType))
          << endl;
     }