src/libANGLE/renderer/vulkan/ProgramExecutableVk.cpp

//
// Copyright 2020 The ANGLE Project Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
//
// ProgramExecutableVk.cpp: Collects the information and interfaces common to both ProgramVks and
// ProgramPipelineVks in order to execute/draw with either.

#include "libANGLE/renderer/vulkan/ProgramExecutableVk.h"

#include "common/string_utils.h"
#include "libANGLE/renderer/glslang_wrapper_utils.h"
#include "libANGLE/renderer/vulkan/BufferVk.h"
#include "libANGLE/renderer/vulkan/DisplayVk.h"
#include "libANGLE/renderer/vulkan/FramebufferVk.h"
#include "libANGLE/renderer/vulkan/GlslangWrapperVk.h"
#include "libANGLE/renderer/vulkan/ProgramPipelineVk.h"
#include "libANGLE/renderer/vulkan/ProgramVk.h"
#include "libANGLE/renderer/vulkan/TextureVk.h"
#include "libANGLE/renderer/vulkan/TransformFeedbackVk.h"
#include "libANGLE/renderer/vulkan/vk_helpers.h"
#include "libANGLE/renderer/vulkan/vk_utils.h"

namespace rx
{
namespace
{
uint8_t GetGraphicsProgramIndex(ProgramTransformOptions transformOptions)
{
    return gl::bitCast<uint8_t, ProgramTransformOptions>(transformOptions);
}

void LoadShaderInterfaceVariableXfbInfo(gl::BinaryInputStream *stream,
                                        ShaderInterfaceVariableXfbInfo *xfb)
{
    xfb->buffer        = stream->readInt<uint32_t>();
    xfb->offset        = stream->readInt<uint32_t>();
    xfb->stride        = stream->readInt<uint32_t>();
    xfb->arraySize     = stream->readInt<uint32_t>();
    xfb->columnCount   = stream->readInt<uint32_t>();
    xfb->rowCount      = stream->readInt<uint32_t>();
    xfb->arrayIndex    = stream->readInt<uint32_t>();
    xfb->componentType = stream->readInt<uint32_t>();
    xfb->arrayElements.resize(stream->readInt<size_t>());
    for (ShaderInterfaceVariableXfbInfo &arrayElement : xfb->arrayElements)
    {
        LoadShaderInterfaceVariableXfbInfo(stream, &arrayElement);
    }
}

void SaveShaderInterfaceVariableXfbInfo(const ShaderInterfaceVariableXfbInfo &xfb,
                                        gl::BinaryOutputStream *stream)
{
    stream->writeInt(xfb.buffer);
    stream->writeInt(xfb.offset);
    stream->writeInt(xfb.stride);
    stream->writeInt(xfb.arraySize);
    stream->writeInt(xfb.columnCount);
    stream->writeInt(xfb.rowCount);
    stream->writeInt(xfb.arrayIndex);
    stream->writeInt(xfb.componentType);
    stream->writeInt(xfb.arrayElements.size());
    for (const ShaderInterfaceVariableXfbInfo &arrayElement : xfb.arrayElements)
    {
        SaveShaderInterfaceVariableXfbInfo(arrayElement, stream);
    }
}

bool ValidateTransformedSpirV(const ContextVk *contextVk,
                              const gl::ShaderBitSet &linkedShaderStages,
                              const ShaderInterfaceVariableInfoMap &variableInfoMap,
                              const gl::ShaderMap<angle::spirv::Blob> &spirvBlobs)
{
    gl::ShaderType lastPreFragmentStage = gl::GetLastPreFragmentStage(linkedShaderStages);

    for (gl::ShaderType shaderType : linkedShaderStages)
    {
        GlslangSpirvOptions options;
        options.shaderType                = shaderType;
        options.negativeViewportSupported = false;
        options.removeDebugInfo           = true;
        options.isLastPreFragmentStage    = shaderType == lastPreFragmentStage;
        options.isTransformFeedbackStage  = shaderType == lastPreFragmentStage;
        options.useSpirvVaryingPrecisionFixer =
            contextVk->getFeatures().varyingsRequireMatchingPrecisionInSpirv.enabled;

        angle::spirv::Blob transformed;
        if (GlslangWrapperVk::TransformSpirV(options, variableInfoMap, spirvBlobs[shaderType],
                                             &transformed) != angle::Result::Continue)
        {
            return false;
        }
    }
    return true;
}

uint32_t GetInterfaceBlockArraySize(const std::vector<gl::InterfaceBlock> &blocks,
                                    uint32_t bufferIndex)
{
    const gl::InterfaceBlock &block = blocks[bufferIndex];

    if (!block.isArray)
    {
        return 1;
    }

    ASSERT(block.arrayElement == 0);

    // Search consecutively until all array indices of this block are visited.
    uint32_t arraySize;
    for (arraySize = 1; bufferIndex + arraySize < blocks.size(); ++arraySize)
    {
        const gl::InterfaceBlock &nextBlock = blocks[bufferIndex + arraySize];

        if (nextBlock.arrayElement != arraySize)
        {
            break;
        }

        // It's unexpected for an array to start at a non-zero array size, so we can always rely on
        // the sequential `arrayElement`s to belong to the same block.
        ASSERT(nextBlock.name == block.name);
        ASSERT(nextBlock.isArray);
    }

    return arraySize;
}

void SetupDefaultPipelineState(const ContextVk *contextVk,
                               const gl::ProgramExecutable &glExecutable,
                               gl::PrimitiveMode mode,
                               vk::GraphicsPipelineDesc *graphicsPipelineDescOut)
{
    graphicsPipelineDescOut->initDefaults(contextVk, vk::GraphicsPipelineSubset::Complete);
    graphicsPipelineDescOut->setTopology(mode);
    graphicsPipelineDescOut->setRenderPassSampleCount(1);
    graphicsPipelineDescOut->setRenderPassFramebufferFetchMode(glExecutable.usesFramebufferFetch());

    const std::vector<sh::ShaderVariable> &outputVariables   = glExecutable.getOutputVariables();
    const std::vector<gl::VariableLocation> &outputLocations = glExecutable.getOutputLocations();

    for (const gl::VariableLocation &outputLocation : outputLocations)
    {
        if (outputLocation.arrayIndex == 0 && outputLocation.used() && !outputLocation.ignored)
        {
            const sh::ShaderVariable &outputVar = outputVariables[outputLocation.index];

            if (angle::BeginsWith(outputVar.name, "gl_") && outputVar.name != "gl_FragColor")
            {
                continue;
            }

            uint32_t location = 0;
            if (outputVar.location != -1)
            {
                location = outputVar.location;
            }

            GLenum type            = gl::VariableComponentType(outputVar.type);
            angle::FormatID format = angle::FormatID::R8G8B8A8_UNORM;
            if (type == GL_INT)
            {
                format = angle::FormatID::R8G8B8A8_SINT;
            }
            else if (type == GL_UNSIGNED_INT)
            {
                format = angle::FormatID::R8G8B8A8_UINT;
            }

            graphicsPipelineDescOut->setRenderPassColorAttachmentFormat(location, format);
        }
    }

    for (const sh::ShaderVariable &outputVar : outputVariables)
    {
        if (outputVar.name == "gl_FragColor" || outputVar.name == "gl_FragData")
        {
            size_t arraySize = outputVar.isArray() ? outputVar.getOutermostArraySize() : 1;
            for (size_t arrayIndex = 0; arrayIndex < arraySize; ++arrayIndex)
            {
                graphicsPipelineDescOut->setRenderPassColorAttachmentFormat(
                    arrayIndex, angle::FormatID::R8G8B8A8_UNORM);
            }
        }
    }
}

void GetPipelineCacheData(ContextVk *contextVk,
                          const vk::PipelineCache &pipelineCache,
                          angle::MemoryBuffer *cacheDataOut)
{
    ASSERT(pipelineCache.valid() || contextVk->getState().isGLES1() ||
           !contextVk->getFeatures().warmUpPipelineCacheAtLink.enabled);
    if (!pipelineCache.valid() ||
        !contextVk->getFeatures().hasEffectivePipelineCacheSerialization.enabled)
    {
        return;
    }

    // Extract the pipeline data.  If failed, or empty, it's simply not stored on disk.
    size_t pipelineCacheSize = 0;
    VkResult result =
        pipelineCache.getCacheData(contextVk->getDevice(), &pipelineCacheSize, nullptr);
    if (result != VK_SUCCESS || pipelineCacheSize == 0)
    {
        return;
    }

    std::vector<uint8_t> pipelineCacheData(pipelineCacheSize);
    result = pipelineCache.getCacheData(contextVk->getDevice(), &pipelineCacheSize,
                                        pipelineCacheData.data());
    if (result != VK_SUCCESS && result != VK_INCOMPLETE)
    {
        return;
    }

    // Compress it.
    if (!egl::CompressBlobCacheData(pipelineCacheData.size(), pipelineCacheData.data(),
                                    cacheDataOut))
    {
        cacheDataOut->clear();
    }
}
}  // namespace

DefaultUniformBlock::DefaultUniformBlock() = default;

DefaultUniformBlock::~DefaultUniformBlock() = default;

// ShaderInfo implementation.
ShaderInfo::ShaderInfo() {}

ShaderInfo::~ShaderInfo() = default;

angle::Result ShaderInfo::initShaders(ContextVk *contextVk,
                                      const gl::ShaderBitSet &linkedShaderStages,
                                      const gl::ShaderMap<const angle::spirv::Blob *> &spirvBlobs,
                                      const ShaderInterfaceVariableInfoMap &variableInfoMap)
{
    clear();

    for (gl::ShaderType shaderType : gl::AllShaderTypes())
    {
        if (spirvBlobs[shaderType] != nullptr)
        {
            mSpirvBlobs[shaderType] = *spirvBlobs[shaderType];
        }
    }

    // Assert that SPIR-V transformation is correct, even if the test never issues a draw call.
    // Don't validate GLES1 programs because they are always created right before a draw, so they
    // will naturally be validated.  This improves GLES1 test run times.
    if (!contextVk->getState().isGLES1())
    {
        ASSERT(
            ValidateTransformedSpirV(contextVk, linkedShaderStages, variableInfoMap, mSpirvBlobs));
    }

    mIsInitialized = true;
    return angle::Result::Continue;
}

void ShaderInfo::initShaderFromProgram(gl::ShaderType shaderType,
                                       const ShaderInfo &programShaderInfo)
{
    mSpirvBlobs[shaderType] = programShaderInfo.mSpirvBlobs[shaderType];
    mIsInitialized          = true;
}

void ShaderInfo::clear()
{
    for (angle::spirv::Blob &spirvBlob : mSpirvBlobs)
    {
        spirvBlob.clear();
    }
    mIsInitialized = false;
}

void ShaderInfo::load(gl::BinaryInputStream *stream)
{
    clear();

    // Read in shader codes for all shader types
    for (gl::ShaderType shaderType : gl::AllShaderTypes())
    {
        angle::spirv::Blob *spirvBlob = &mSpirvBlobs[shaderType];

        // Read the SPIR-V
        stream->readIntVector<uint32_t>(spirvBlob);
    }

    mIsInitialized = true;
}

void ShaderInfo::save(gl::BinaryOutputStream *stream)
{
    ASSERT(valid());

    // Write out shader codes for all shader types
    for (gl::ShaderType shaderType : gl::AllShaderTypes())
    {
        const angle::spirv::Blob &spirvBlob = mSpirvBlobs[shaderType];

        // Write the SPIR-V
        stream->writeIntVector(spirvBlob);
    }
}

// ProgramInfo implementation.
ProgramInfo::ProgramInfo() {}

ProgramInfo::~ProgramInfo() = default;

angle::Result ProgramInfo::initProgram(ContextVk *contextVk,
                                       gl::ShaderType shaderType,
                                       bool isLastPreFragmentStage,
                                       bool isTransformFeedbackProgram,
                                       const ShaderInfo &shaderInfo,
                                       ProgramTransformOptions optionBits,
                                       const ShaderInterfaceVariableInfoMap &variableInfoMap)
{
    const gl::ShaderMap<angle::spirv::Blob> &originalSpirvBlobs = shaderInfo.getSpirvBlobs();
    const angle::spirv::Blob &originalSpirvBlob                 = originalSpirvBlobs[shaderType];
    gl::ShaderMap<angle::spirv::Blob> transformedSpirvBlobs;
    angle::spirv::Blob &transformedSpirvBlob = transformedSpirvBlobs[shaderType];

    GlslangSpirvOptions options;
    options.shaderType               = shaderType;
    options.removeDebugInfo          = !contextVk->getFeatures().retainSPIRVDebugInfo.enabled;
    options.isLastPreFragmentStage   = isLastPreFragmentStage;
    options.isTransformFeedbackStage = isLastPreFragmentStage && isTransformFeedbackProgram &&
                                       !optionBits.removeTransformFeedbackEmulation;
    options.isTransformFeedbackEmulated = contextVk->getFeatures().emulateTransformFeedback.enabled;
    options.negativeViewportSupported   = contextVk->getFeatures().supportsNegativeViewport.enabled;
    options.isMultisampledFramebufferFetch =
        optionBits.multiSampleFramebufferFetch && shaderType == gl::ShaderType::Fragment;

    // Don't validate SPIR-V generated for GLES1 shaders when validation layers are enabled.  The
    // layers already validate SPIR-V, and since GLES1 shaders are controlled by ANGLE, they don't
    // typically require debugging at the SPIR-V level.  This improves GLES1 conformance test run
    // time.
    options.validate =
        !(contextVk->getState().isGLES1() && contextVk->getRenderer()->getEnableValidationLayers());

    options.useSpirvVaryingPrecisionFixer =
        contextVk->getFeatures().varyingsRequireMatchingPrecisionInSpirv.enabled;

    ANGLE_TRY(GlslangWrapperVk::TransformSpirV(options, variableInfoMap, originalSpirvBlob,
                                               &transformedSpirvBlob));
    ANGLE_TRY(vk::InitShaderModule(contextVk, &mShaders[shaderType].get(),
                                   transformedSpirvBlob.data(),
                                   transformedSpirvBlob.size() * sizeof(uint32_t)));

    mProgramHelper.setShader(shaderType, &mShaders[shaderType]);

    return angle::Result::Continue;
}

void ProgramInfo::release(ContextVk *contextVk)
{
    mProgramHelper.release(contextVk);

    for (vk::RefCounted<vk::ShaderModule> &shader : mShaders)
    {
        shader.get().destroy(contextVk->getDevice());
    }
}

ProgramExecutableVk::ProgramExecutableVk()
    : mNumDefaultUniformDescriptors(0),
      mImmutableSamplersMaxDescriptorCount(1),
      mUniformBufferDescriptorType(VK_DESCRIPTOR_TYPE_MAX_ENUM),
      mDynamicUniformDescriptorOffsets{}
{
    for (std::shared_ptr<DefaultUniformBlock> &defaultBlock : mDefaultUniformBlocks)
    {
        defaultBlock = std::make_shared<DefaultUniformBlock>();
    }
}

ProgramExecutableVk::~ProgramExecutableVk()
{
    ASSERT(!mPipelineCache.valid());
}

void ProgramExecutableVk::resetLayout(ContextVk *contextVk)
{
    for (auto &descriptorSetLayout : mDescriptorSetLayouts)
    {
        descriptorSetLayout.reset();
    }
    mImmutableSamplersMaxDescriptorCount = 1;
    mImmutableSamplerIndexMap.clear();
    mPipelineLayout.reset();

    mDescriptorSets.fill(VK_NULL_HANDLE);
    mNumDefaultUniformDescriptors = 0;

    for (vk::RefCountedDescriptorPoolBinding &binding : mDescriptorPoolBindings)
    {
        binding.reset();
    }

    for (vk::DescriptorPoolPointer &pool : mDescriptorPools)
    {
        pool.reset();
    }

    // Initialize with an invalid BufferSerial
    mCurrentDefaultUniformBufferSerial = vk::BufferSerial();

    for (ProgramInfo &programInfo : mGraphicsProgramInfos)
    {
        programInfo.release(contextVk);
    }
    mComputeProgramInfo.release(contextVk);

    for (CompleteGraphicsPipelineCache &pipelines : mCompleteGraphicsPipelines)
    {
        pipelines.release(contextVk);
    }
    for (ShadersGraphicsPipelineCache &pipelines : mShadersGraphicsPipelines)
    {
        pipelines.release(contextVk);
    }
    for (vk::PipelineHelper &pipeline : mComputePipelines)
    {
        pipeline.release(contextVk);
    }

    contextVk->onProgramExecutableReset(this);
}

void ProgramExecutableVk::reset(ContextVk *contextVk)
{
    resetLayout(contextVk);

    if (mPipelineCache.valid())
    {
        mPipelineCache.destroy(contextVk->getDevice());
    }
}

angle::Result ProgramExecutableVk::initializePipelineCache(
    ContextVk *contextVk,
    const std::vector<uint8_t> &compressedPipelineData)
{
    ASSERT(!mPipelineCache.valid());

    angle::MemoryBuffer uncompressedData;
    if (!egl::DecompressBlobCacheData(compressedPipelineData.data(), compressedPipelineData.size(),
                                      &uncompressedData))
    {
        return angle::Result::Stop;
    }

    VkPipelineCacheCreateInfo pipelineCacheCreateInfo = {};
    pipelineCacheCreateInfo.sType           = VK_STRUCTURE_TYPE_PIPELINE_CACHE_CREATE_INFO;
    pipelineCacheCreateInfo.initialDataSize = uncompressedData.size();
    pipelineCacheCreateInfo.pInitialData    = uncompressedData.data();

    if (contextVk->getFeatures().supportsPipelineCreationCacheControl.enabled)
    {
        pipelineCacheCreateInfo.flags |= VK_PIPELINE_CACHE_CREATE_EXTERNALLY_SYNCHRONIZED_BIT_EXT;
    }

    ANGLE_VK_TRY(contextVk, mPipelineCache.init(contextVk->getDevice(), pipelineCacheCreateInfo));

    // Merge the pipeline cache into RendererVk's.
    if (contextVk->getFeatures().mergeProgramPipelineCachesToGlobalCache.enabled)
    {
        ANGLE_TRY(contextVk->getRenderer()->mergeIntoPipelineCache(mPipelineCache));
    }

    return angle::Result::Continue;
}

std::unique_ptr<rx::LinkEvent> ProgramExecutableVk::load(ContextVk *contextVk,
                                                         const gl::ProgramExecutable &glExecutable,
                                                         bool isSeparable,
                                                         gl::BinaryInputStream *stream)
{
    gl::ShaderMap<ShaderInterfaceVariableInfoMap::VariableTypeToInfoMap> data;
    gl::ShaderMap<ShaderInterfaceVariableInfoMap::NameToTypeAndIndexMap> nameToTypeAndIndexMap;
    gl::ShaderMap<ShaderInterfaceVariableInfoMap::VariableTypeToIndexMap> indexedResourceMap;

    for (gl::ShaderType shaderType : gl::AllShaderTypes())
    {
        size_t nameCount = stream->readInt<size_t>();
        for (size_t nameIndex = 0; nameIndex < nameCount; ++nameIndex)
        {
            const std::string variableName  = stream->readString();
            ShaderVariableType variableType = stream->readEnum<ShaderVariableType>();
            uint32_t index                  = stream->readInt<uint32_t>();
            nameToTypeAndIndexMap[shaderType][variableName] = {variableType, index};
        }

        for (ShaderVariableType variableType : angle::AllEnums<ShaderVariableType>())
        {
            size_t infoArraySize = stream->readInt<size_t>();
            for (size_t infoIndex = 0; infoIndex < infoArraySize; ++infoIndex)
            {
                ShaderInterfaceVariableInfo info;

                info.descriptorSet = stream->readInt<uint32_t>();
                info.binding       = stream->readInt<uint32_t>();
                info.location      = stream->readInt<uint32_t>();
                info.component     = stream->readInt<uint32_t>();
                info.index         = stream->readInt<uint32_t>();
                // PackedEnumBitSet uses uint8_t
                info.activeStages = gl::ShaderBitSet(stream->readInt<uint8_t>());
                LoadShaderInterfaceVariableXfbInfo(stream, &info.xfb);
                info.fieldXfb.resize(stream->readInt<size_t>());
                for (ShaderInterfaceVariableXfbInfo &xfb : info.fieldXfb)
                {
                    LoadShaderInterfaceVariableXfbInfo(stream, &xfb);
                }
                info.useRelaxedPrecision     = stream->readBool();
                info.varyingIsInput          = stream->readBool();
                info.varyingIsOutput         = stream->readBool();
                info.attributeComponentCount = stream->readInt<uint8_t>();
                info.attributeLocationCount  = stream->readInt<uint8_t>();
                info.isDuplicate             = stream->readBool();

                data[shaderType][variableType].push_back(info);
            }

            uint32_t resourceMapSize = stream->readInt<uint32_t>();
            for (uint32_t resourceIndex = 0; resourceIndex < resourceMapSize; ++resourceIndex)
            {
                uint32_t variableIndex = stream->readInt<uint32_t>();
                indexedResourceMap[shaderType][variableType][resourceIndex] = variableIndex;
            }
        }
    }

    mVariableInfoMap.load(data, nameToTypeAndIndexMap, indexedResourceMap);

    mOriginalShaderInfo.load(stream);

    // Deserializes the uniformLayout data of mDefaultUniformBlocks
    for (gl::ShaderType shaderType : gl::AllShaderTypes())
    {
        const size_t uniformCount = stream->readInt<size_t>();
        for (unsigned int uniformIndex = 0; uniformIndex < uniformCount; ++uniformIndex)
        {
            sh::BlockMemberInfo blockInfo;
            gl::LoadBlockMemberInfo(stream, &blockInfo);
            mDefaultUniformBlocks[shaderType]->uniformLayout.push_back(blockInfo);
        }
    }

    gl::ShaderMap<size_t> requiredBufferSize;
    requiredBufferSize.fill(0);
    // Deserializes required uniform block memory sizes
    for (gl::ShaderType shaderType : gl::AllShaderTypes())
    {
        requiredBufferSize[shaderType] = stream->readInt<size_t>();
    }

    if (!isSeparable)
    {
        size_t compressedPipelineDataSize = 0;
        stream->readInt<size_t>(&compressedPipelineDataSize);

        std::vector<uint8_t> compressedPipelineData(compressedPipelineDataSize);
        if (compressedPipelineDataSize > 0)
        {
            stream->readBytes(compressedPipelineData.data(), compressedPipelineDataSize);

            // Initialize the pipeline cache based on cached data.
            angle::Result status = initializePipelineCache(contextVk, compressedPipelineData);
            if (status != angle::Result::Continue)
            {
                return std::make_unique<LinkEventDone>(status);
            }
        }
    }

    // Initialize and resize the mDefaultUniformBlocks' memory
    angle::Result status = resizeUniformBlockMemory(contextVk, glExecutable, requiredBufferSize);
    if (status != angle::Result::Continue)
    {
        return std::make_unique<LinkEventDone>(status);
    }

    status = createPipelineLayout(contextVk, glExecutable, nullptr);
    return std::make_unique<LinkEventDone>(status);
}

void ProgramExecutableVk::save(ContextVk *contextVk,
                               bool isSeparable,
                               gl::BinaryOutputStream *stream)
{
    const gl::ShaderMap<ShaderInterfaceVariableInfoMap::VariableTypeToInfoMap> &data =
        mVariableInfoMap.getData();
    const gl::ShaderMap<ShaderInterfaceVariableInfoMap::NameToTypeAndIndexMap>
        &nameToTypeAndIndexMap = mVariableInfoMap.getNameToTypeAndIndexMap();
    const gl::ShaderMap<ShaderInterfaceVariableInfoMap::VariableTypeToIndexMap>
        &indexedResourceMap = mVariableInfoMap.getIndexedResourceMap();

    for (gl::ShaderType shaderType : gl::AllShaderTypes())
    {
        stream->writeInt(nameToTypeAndIndexMap[shaderType].size());
        for (const auto &iter : nameToTypeAndIndexMap[shaderType])
        {
            const std::string &name          = iter.first;
            const TypeAndIndex &typeAndIndex = iter.second;
            stream->writeString(name);
            stream->writeEnum(typeAndIndex.variableType);
            stream->writeInt(typeAndIndex.index);
        }

        for (ShaderVariableType variableType : angle::AllEnums<ShaderVariableType>())
        {
            const ShaderInterfaceVariableInfoMap::VariableInfoArray &infoArray =
                data[shaderType][variableType];

            stream->writeInt(infoArray.size());
            for (const ShaderInterfaceVariableInfo &info : infoArray)
            {
                stream->writeInt(info.descriptorSet);
                stream->writeInt(info.binding);
                stream->writeInt(info.location);
                stream->writeInt(info.component);
                stream->writeInt(info.index);
                // PackedEnumBitSet uses uint8_t
                stream->writeInt(info.activeStages.bits());
                SaveShaderInterfaceVariableXfbInfo(info.xfb, stream);
                stream->writeInt(info.fieldXfb.size());
                for (const ShaderInterfaceVariableXfbInfo &xfb : info.fieldXfb)
                {
                    SaveShaderInterfaceVariableXfbInfo(xfb, stream);
                }
                stream->writeBool(info.useRelaxedPrecision);
                stream->writeBool(info.varyingIsInput);
                stream->writeBool(info.varyingIsOutput);
                stream->writeInt(info.attributeComponentCount);
                stream->writeInt(info.attributeLocationCount);
                stream->writeBool(info.isDuplicate);
            }

            const ShaderInterfaceVariableInfoMap::ResourceIndexMap &resourceIndexMap =
                indexedResourceMap[shaderType][variableType];
            stream->writeInt(static_cast<uint32_t>(resourceIndexMap.size()));
            for (uint32_t resourceIndex = 0; resourceIndex < resourceIndexMap.size();
                 ++resourceIndex)
            {
                stream->writeInt(resourceIndexMap[resourceIndex]);
            }
        }
    }

    mOriginalShaderInfo.save(stream);

    // Serializes the uniformLayout data of mDefaultUniformBlocks
    for (gl::ShaderType shaderType : gl::AllShaderTypes())
    {
        const size_t uniformCount = mDefaultUniformBlocks[shaderType]->uniformLayout.size();
        stream->writeInt(uniformCount);
        for (unsigned int uniformIndex = 0; uniformIndex < uniformCount; ++uniformIndex)
        {
            sh::BlockMemberInfo &blockInfo =
                mDefaultUniformBlocks[shaderType]->uniformLayout[uniformIndex];
            gl::WriteBlockMemberInfo(stream, blockInfo);
        }
    }

    // Serializes required uniform block memory sizes
    for (gl::ShaderType shaderType : gl::AllShaderTypes())
    {
        stream->writeInt(mDefaultUniformBlocks[shaderType]->uniformData.size());
    }

    // Compress and save mPipelineCache.  Separable programs don't warm up the cache, while program
    // pipelines do.  However, currently ANGLE doesn't sync program pipelines to cache.  ANGLE could
    // potentially use VK_EXT_graphics_pipeline_library to create separate pipelines for
    // pre-rasterization and fragment subsets, but currently those subsets are bundled together.
    if (!isSeparable)
    {
        angle::MemoryBuffer cacheData;
        GetPipelineCacheData(contextVk, mPipelineCache, &cacheData);

        stream->writeInt(cacheData.size());
        if (cacheData.size() > 0)
        {
            stream->writeBytes(cacheData.data(), cacheData.size());
        }
    }
}

void ProgramExecutableVk::clearVariableInfoMap()
{
    mVariableInfoMap.clear();
}

angle::Result ProgramExecutableVk::warmUpPipelineCache(ContextVk *contextVk,
                                                       const gl::ProgramExecutable &glExecutable)
{
    // The cache warm up is skipped for GLES1 for two reasons:
    //
    // - Since GLES1 shaders are limited, the individual programs don't necessarily add new
    //   pipelines, but rather it's draw time state that controls that.  Since the programs are
    //   generated at draw time, it's just as well to let the pipelines be created using the
    //   renderer's shared cache.
    // - Individual GLES1 tests are long, and this adds a considerable overhead to those tests
    if (contextVk->getState().isGLES1())
    {
        return angle::Result::Continue;
    }

    if (!contextVk->getFeatures().warmUpPipelineCacheAtLink.enabled)
    {
        return angle::Result::Continue;
    }

    // No synchronization necessary when accessing the program executable's cache as there is no
    // access to it from other threads at this point.
    vk::PipelineCacheAccess pipelineCache;
    pipelineCache.init(&mPipelineCache, nullptr);

    // Create a set of pipelines.  Ideally, that would be the entire set of possible pipelines so
    // there would be none created at draw time.  This is gated on the removal of some
    // specialization constants and adoption of VK_EXT_graphics_pipeline_library.
    const bool isCompute = glExecutable.hasLinkedShaderStage(gl::ShaderType::Compute);
    if (isCompute)
    {
        // There is no state associated with compute programs, so only one pipeline needs creation
        // to warm up the cache.
        vk::PipelineHelper *pipeline = nullptr;
        ANGLE_TRY(getOrCreateComputePipeline(contextVk, &pipelineCache, PipelineSource::WarmUp,
                                             glExecutable, &pipeline));

        // Merge the cache with RendererVk's
        if (contextVk->getFeatures().mergeProgramPipelineCachesToGlobalCache.enabled)
        {
            ANGLE_TRY(contextVk->getRenderer()->mergeIntoPipelineCache(mPipelineCache));
        }

        return angle::Result::Continue;
    }

    const vk::GraphicsPipelineDesc *descPtr = nullptr;
    vk::PipelineHelper *pipeline            = nullptr;
    vk::GraphicsPipelineDesc graphicsPipelineDesc;

    // It is only at drawcall time that we will have complete information required to build the
    // graphics pipeline descriptor. Use the most "commonly seen" state values and create the
    // pipeline.
    gl::PrimitiveMode mode = (glExecutable.hasLinkedShaderStage(gl::ShaderType::TessControl) ||
                              glExecutable.hasLinkedShaderStage(gl::ShaderType::TessEvaluation))
                                 ? gl::PrimitiveMode::Patches
                                 : gl::PrimitiveMode::TriangleStrip;
    SetupDefaultPipelineState(contextVk, glExecutable, mode, &graphicsPipelineDesc);

    // Variations that definitely matter:
    //
    // - PreRotation: It's a boolean specialization constant
    // - Depth correction: It's a SPIR-V transformation
    //
    // There are a number of states that are not currently dynamic (and may never be, such as sample
    // shading), but pre-creating shaders for them is impractical.  Most such state is likely unused
    // by most applications, but variations can be added here for certain apps that are known to
    // benefit from it.
    ProgramTransformOptions transformOptions = {};

    angle::FixedVector<bool, 2> surfaceRotationVariations = {false};
    if (contextVk->getFeatures().enablePreRotateSurfaces.enabled &&
        !contextVk->getFeatures().preferDriverUniformOverSpecConst.enabled)
    {
        surfaceRotationVariations.push_back(true);
    }

    // Only build the shaders subset of the pipeline if VK_EXT_graphics_pipeline_library is
    // supported, especially since the vertex input and fragment output state set up here is
    // completely bogus.
    vk::GraphicsPipelineSubset subset =
        contextVk->getFeatures().supportsGraphicsPipelineLibrary.enabled
            ? vk::GraphicsPipelineSubset::Shaders
            : vk::GraphicsPipelineSubset::Complete;

    for (bool rotation : surfaceRotationVariations)
    {
        transformOptions.surfaceRotation = rotation;

        ANGLE_TRY(createGraphicsPipelineImpl(contextVk, transformOptions, subset, &pipelineCache,
                                             PipelineSource::WarmUp, graphicsPipelineDesc,
                                             glExecutable, &descPtr, &pipeline));
    }

    // Merge the cache with RendererVk's
    if (contextVk->getFeatures().mergeProgramPipelineCachesToGlobalCache.enabled)
    {
        ANGLE_TRY(contextVk->getRenderer()->mergeIntoPipelineCache(mPipelineCache));
    }

    return angle::Result::Continue;
}

void ProgramExecutableVk::addInterfaceBlockDescriptorSetDesc(
    const std::vector<gl::InterfaceBlock> &blocks,
    gl::ShaderType shaderType,
    ShaderVariableType variableType,
    VkDescriptorType descType,
    vk::DescriptorSetLayoutDesc *descOut)
{
    for (uint32_t bufferIndex = 0, arraySize = 0; bufferIndex < blocks.size();
         bufferIndex += arraySize)
    {
        gl::InterfaceBlock block = blocks[bufferIndex];
        arraySize                = GetInterfaceBlockArraySize(blocks, bufferIndex);

        if (!block.isActive(shaderType))
        {
            continue;
        }

        const ShaderInterfaceVariableInfo &info =
            mVariableInfoMap.getIndexedVariableInfo(shaderType, variableType, bufferIndex);
        if (info.isDuplicate)
        {
            continue;
        }

        VkShaderStageFlags activeStages = gl_vk::GetShaderStageFlags(info.activeStages);

        descOut->update(info.binding, descType, arraySize, activeStages, nullptr);
    }
}

void ProgramExecutableVk::addAtomicCounterBufferDescriptorSetDesc(
    const std::vector<gl::AtomicCounterBuffer> &atomicCounterBuffers,
    gl::ShaderType shaderType,
    vk::DescriptorSetLayoutDesc *descOut)
{
    if (atomicCounterBuffers.empty() || !mVariableInfoMap.hasAtomicCounterInfo(shaderType))
    {
        return;
    }

    const ShaderInterfaceVariableInfo &info = mVariableInfoMap.getAtomicCounterInfo(shaderType);
    if (info.isDuplicate || !info.activeStages[shaderType])
    {
        return;
    }

    VkShaderStageFlags activeStages = gl_vk::GetShaderStageFlags(info.activeStages);

    // A single storage buffer array is used for all stages for simplicity.
    descOut->update(info.binding, vk::kStorageBufferDescriptorType,
                    gl::IMPLEMENTATION_MAX_ATOMIC_COUNTER_BUFFER_BINDINGS, activeStages, nullptr);
}

void ProgramExecutableVk::addImageDescriptorSetDesc(const gl::ProgramExecutable &executable,
                                                    vk::DescriptorSetLayoutDesc *descOut)
{
    const std::vector<gl::ImageBinding> &imageBindings = executable.getImageBindings();
    const std::vector<gl::LinkedUniform> &uniforms     = executable.getUniforms();

    for (uint32_t imageIndex = 0; imageIndex < imageBindings.size(); ++imageIndex)
    {
        uint32_t uniformIndex = executable.getUniformIndexFromImageIndex(imageIndex);
        const gl::LinkedUniform &imageUniform = uniforms[uniformIndex];

        // 2D arrays are split into multiple 1D arrays when generating LinkedUniforms. Since they
        // are flattened into one array, ignore the nonzero elements and expand the array to the
        // total array size.
        if (imageUniform.outerArrayOffset > 0)
        {
            ASSERT(gl::SamplerNameContainsNonZeroArrayElement(imageUniform.name));
            continue;
        }

        ASSERT(!gl::SamplerNameContainsNonZeroArrayElement(imageUniform.name));

        // The front-end always binds array image units sequentially.
        const gl::ImageBinding &imageBinding = imageBindings[imageIndex];
        uint32_t arraySize = static_cast<uint32_t>(imageBinding.boundImageUnits.size());
        for (unsigned int outerArraySize : imageUniform.outerArraySizes)
        {
            arraySize *= outerArraySize;
        }

        for (gl::ShaderType shaderType : executable.getLinkedShaderStages())
        {
            if (!imageUniform.isActive(shaderType))
            {
                continue;
            }

            const ShaderInterfaceVariableInfo &info = mVariableInfoMap.getIndexedVariableInfo(
                shaderType, ShaderVariableType::Image, imageIndex);
            if (info.isDuplicate)
            {
                continue;
            }

            VkShaderStageFlags activeStages = gl_vk::GetShaderStageFlags(info.activeStages);

            const VkDescriptorType descType = imageBinding.textureType == gl::TextureType::Buffer
                                                  ? VK_DESCRIPTOR_TYPE_STORAGE_TEXEL_BUFFER
                                                  : VK_DESCRIPTOR_TYPE_STORAGE_IMAGE;
            descOut->update(info.binding, descType, arraySize, activeStages, nullptr);
        }
    }
}

void ProgramExecutableVk::addInputAttachmentDescriptorSetDesc(
    const gl::ProgramExecutable &executable,
    gl::ShaderType shaderType,
    vk::DescriptorSetLayoutDesc *descOut)
{
    if (shaderType != gl::ShaderType::Fragment)
    {
        return;
    }

    if (!executable.usesFramebufferFetch())
    {
        return;
    }

    const ShaderInterfaceVariableInfo &baseInfo =
        mVariableInfoMap.getFramebufferFetchInfo(shaderType);
    if (baseInfo.isDuplicate)
    {
        return;
    }

    const std::vector<gl::LinkedUniform> &uniforms = executable.getUniforms();
    const uint32_t baseUniformIndex                = executable.getFragmentInoutRange().low();
    const gl::LinkedUniform &baseInputAttachment   = uniforms.at(baseUniformIndex);
    uint32_t baseBinding = baseInfo.binding - baseInputAttachment.location;

    for (uint32_t colorIndex = 0; colorIndex < gl::IMPLEMENTATION_MAX_DRAW_BUFFERS; ++colorIndex)
    {
        descOut->update(baseBinding, VK_DESCRIPTOR_TYPE_INPUT_ATTACHMENT, 1,
                        VK_SHADER_STAGE_FRAGMENT_BIT, nullptr);
        baseBinding++;
    }
}

angle::Result ProgramExecutableVk::addTextureDescriptorSetDesc(
    ContextVk *contextVk,
    const gl::ProgramExecutable &executable,
    const gl::ActiveTextureArray<TextureVk *> *activeTextures,
    vk::DescriptorSetLayoutDesc *descOut)
{
    const std::vector<gl::SamplerBinding> &samplerBindings = executable.getSamplerBindings();
    const std::vector<gl::LinkedUniform> &uniforms         = executable.getUniforms();

    for (uint32_t textureIndex = 0; textureIndex < samplerBindings.size(); ++textureIndex)
    {
        uint32_t uniformIndex = executable.getUniformIndexFromSamplerIndex(textureIndex);
        const gl::LinkedUniform &samplerUniform = uniforms[uniformIndex];

        // 2D arrays are split into multiple 1D arrays when generating LinkedUniforms. Since they
        // are flattened into one array, ignore the nonzero elements and expand the array to the
        // total array size.
        if (samplerUniform.outerArrayOffset > 0)
        {
            ASSERT(gl::SamplerNameContainsNonZeroArrayElement(samplerUniform.name));
            continue;
        }

        ASSERT(!gl::SamplerNameContainsNonZeroArrayElement(samplerUniform.name));

        // The front-end always binds array sampler units sequentially.
        const gl::SamplerBinding &samplerBinding = samplerBindings[textureIndex];
        uint32_t arraySize = static_cast<uint32_t>(samplerBinding.boundTextureUnits.size());
        for (unsigned int outerArraySize : samplerUniform.outerArraySizes)
        {
            arraySize *= outerArraySize;
        }

        for (gl::ShaderType shaderType : executable.getLinkedShaderStages())
        {
            if (!samplerUniform.isActive(shaderType))
            {
                continue;
            }

            const ShaderInterfaceVariableInfo &info = mVariableInfoMap.getIndexedVariableInfo(
                shaderType, ShaderVariableType::Texture, textureIndex);
            if (info.isDuplicate)
            {
                continue;
            }

            VkShaderStageFlags activeStages = gl_vk::GetShaderStageFlags(info.activeStages);

            // TODO: https://issuetracker.google.com/issues/158215272: how do we handle array of
            // immutable samplers?
            GLuint textureUnit = samplerBinding.boundTextureUnits[0];
            if (activeTextures &&
                ((*activeTextures)[textureUnit]->getImage().hasImmutableSampler()))
            {
                ASSERT(samplerBinding.boundTextureUnits.size() == 1);

                // In the case of samplerExternal2DY2YEXT, we need
                // samplerYcbcrConversion object with IDENTITY conversion model
                bool isSamplerExternalY2Y =
                    samplerBinding.samplerType == GL_SAMPLER_EXTERNAL_2D_Y2Y_EXT;

                // Always take the texture's sampler, that's only way to get to yuv conversion for
                // externalFormat
                const TextureVk *textureVk = (*activeTextures)[textureUnit];
                const vk::Sampler &immutableSampler =
                    textureVk->getSampler(isSamplerExternalY2Y).get();
                descOut->update(info.binding, VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER, arraySize,
                                activeStages, &immutableSampler);
                const vk::ImageHelper &image = textureVk->getImage();
                const vk::YcbcrConversionDesc ycbcrConversionDesc =
                    isSamplerExternalY2Y ? image.getY2YConversionDesc()
                                         : image.getYcbcrConversionDesc();
                mImmutableSamplerIndexMap[ycbcrConversionDesc] = textureIndex;
                // The Vulkan spec has the following note -
                // All descriptors in a binding use the same maximum
                // combinedImageSamplerDescriptorCount descriptors to allow implementations to use a
                // uniform stride for dynamic indexing of the descriptors in the binding.
                uint64_t externalFormat        = image.getExternalFormat();
                uint32_t formatDescriptorCount = 0;

                RendererVk *renderer = contextVk->getRenderer();

                if (externalFormat != 0)
                {
                    ANGLE_TRY(renderer->getFormatDescriptorCountForExternalFormat(
                        contextVk, externalFormat, &formatDescriptorCount));
                }
                else
                {
                    VkFormat vkFormat = image.getActualVkFormat();
                    ASSERT(vkFormat != 0);
                    ANGLE_TRY(renderer->getFormatDescriptorCountForVkFormat(
                        contextVk, vkFormat, &formatDescriptorCount));
                }

                ASSERT(formatDescriptorCount > 0);
                mImmutableSamplersMaxDescriptorCount =
                    std::max(mImmutableSamplersMaxDescriptorCount, formatDescriptorCount);
            }
            else
            {
                const VkDescriptorType descType =
                    samplerBinding.textureType == gl::TextureType::Buffer
                        ? VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER
                        : VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER;
                descOut->update(info.binding, descType, arraySize, activeStages, nullptr);
            }
        }
    }

    return angle::Result::Continue;
}

ProgramTransformOptions ProgramExecutableVk::getTransformOptions(
    ContextVk *contextVk,
    const vk::GraphicsPipelineDesc &desc,
    const gl::ProgramExecutable &glExecutable)
{
    ProgramTransformOptions transformOptions = {};

    transformOptions.surfaceRotation = desc.getSurfaceRotation();
    transformOptions.removeTransformFeedbackEmulation =
        contextVk->getFeatures().emulateTransformFeedback.enabled &&
        !contextVk->getState().isTransformFeedbackActiveUnpaused();
    FramebufferVk *drawFrameBuffer = vk::GetImpl(contextVk->getState().getDrawFramebuffer());
    const bool hasFramebufferFetch = glExecutable.usesFramebufferFetch();
    const bool isMultisampled      = drawFrameBuffer->getSamples() > 1;
    transformOptions.multiSampleFramebufferFetch = hasFramebufferFetch && isMultisampled;

    return transformOptions;
}

angle::Result ProgramExecutableVk::initGraphicsShaderPrograms(
    ContextVk *contextVk,
    ProgramTransformOptions transformOptions,
    const gl::ProgramExecutable &glExecutable,
    vk::ShaderProgramHelper **shaderProgramOut)
{
    ASSERT(glExecutable.hasLinkedShaderStage(gl::ShaderType::Vertex));

    const uint8_t programIndex                = GetGraphicsProgramIndex(transformOptions);
    ProgramInfo &programInfo                  = mGraphicsProgramInfos[programIndex];
    const gl::ShaderBitSet linkedShaderStages = glExecutable.getLinkedShaderStages();
    gl::ShaderType lastPreFragmentStage       = gl::GetLastPreFragmentStage(linkedShaderStages);

    const bool isTransformFeedbackProgram =
        !glExecutable.getLinkedTransformFeedbackVaryings().empty();

    for (gl::ShaderType shaderType : linkedShaderStages)
    {
        ANGLE_TRY(initGraphicsShaderProgram(
            contextVk, shaderType, shaderType == lastPreFragmentStage, isTransformFeedbackProgram,
            transformOptions, &programInfo, mVariableInfoMap));
    }

    *shaderProgramOut = programInfo.getShaderProgram();
    ASSERT(*shaderProgramOut);

    return angle::Result::Continue;
}

angle::Result ProgramExecutableVk::createGraphicsPipelineImpl(
    ContextVk *contextVk,
    ProgramTransformOptions transformOptions,
    vk::GraphicsPipelineSubset pipelineSubset,
    vk::PipelineCacheAccess *pipelineCache,
    PipelineSource source,
    const vk::GraphicsPipelineDesc &desc,
    const gl::ProgramExecutable &glExecutable,
    const vk::GraphicsPipelineDesc **descPtrOut,
    vk::PipelineHelper **pipelineOut)
{
    vk::ShaderProgramHelper *shaderProgram = nullptr;
    ANGLE_TRY(
        initGraphicsShaderPrograms(contextVk, transformOptions, glExecutable, &shaderProgram));

    const uint8_t programIndex = GetGraphicsProgramIndex(transformOptions);

    // Set specialization constants.  These are also a part of GraphicsPipelineDesc, so that a
    // change in specialization constants also results in a new pipeline.
    vk::SpecializationConstants specConsts;
    specConsts.surfaceRotation = transformOptions.surfaceRotation;
    specConsts.dither          = desc.getEmulatedDitherControl();

    // Pull in a compatible RenderPass.
    const vk::RenderPass *compatibleRenderPass = nullptr;
    ANGLE_TRY(contextVk->getRenderPassCache().getCompatibleRenderPass(
        contextVk, desc.getRenderPassDesc(), &compatibleRenderPass));

    if (pipelineSubset == vk::GraphicsPipelineSubset::Complete)
    {
        CompleteGraphicsPipelineCache &pipelines = mCompleteGraphicsPipelines[programIndex];
        return shaderProgram->createGraphicsPipeline(
            contextVk, &pipelines, pipelineCache, *compatibleRenderPass, getPipelineLayout(),
            source, desc, specConsts, descPtrOut, pipelineOut);
    }
    else
    {
        // Vertex input and fragment output subsets are independent of shaders, and are not created
        // through the program executable.
        ASSERT(pipelineSubset == vk::GraphicsPipelineSubset::Shaders);

        ShadersGraphicsPipelineCache &pipelines = mShadersGraphicsPipelines[programIndex];
        return shaderProgram->createGraphicsPipeline(
            contextVk, &pipelines, pipelineCache, *compatibleRenderPass, getPipelineLayout(),
            source, desc, specConsts, descPtrOut, pipelineOut);
    }
}

angle::Result ProgramExecutableVk::getGraphicsPipeline(ContextVk *contextVk,
                                                       vk::GraphicsPipelineSubset pipelineSubset,
                                                       const vk::GraphicsPipelineDesc &desc,
                                                       const gl::ProgramExecutable &glExecutable,
                                                       const vk::GraphicsPipelineDesc **descPtrOut,
                                                       vk::PipelineHelper **pipelineOut)
{
    ProgramTransformOptions transformOptions = getTransformOptions(contextVk, desc, glExecutable);

    vk::ShaderProgramHelper *shaderProgram = nullptr;
    ANGLE_TRY(
        initGraphicsShaderPrograms(contextVk, transformOptions, glExecutable, &shaderProgram));

    const uint8_t programIndex = GetGraphicsProgramIndex(transformOptions);

    *descPtrOut  = nullptr;
    *pipelineOut = nullptr;

    if (pipelineSubset == vk::GraphicsPipelineSubset::Complete)
    {
        mCompleteGraphicsPipelines[programIndex].getPipeline(desc, descPtrOut, pipelineOut);
    }
    else
    {
        // Vertex input and fragment output subsets are independent of shaders, and are not created
        // through the program executable.
        ASSERT(pipelineSubset == vk::GraphicsPipelineSubset::Shaders);

        mShadersGraphicsPipelines[programIndex].getPipeline(desc, descPtrOut, pipelineOut);
    }

    return angle::Result::Continue;
}

angle::Result ProgramExecutableVk::createGraphicsPipeline(
    ContextVk *contextVk,
    vk::GraphicsPipelineSubset pipelineSubset,
    vk::PipelineCacheAccess *pipelineCache,
    PipelineSource source,
    const vk::GraphicsPipelineDesc &desc,
    const gl::ProgramExecutable &glExecutable,
    const vk::GraphicsPipelineDesc **descPtrOut,
    vk::PipelineHelper **pipelineOut)
{
    ProgramTransformOptions transformOptions = getTransformOptions(contextVk, desc, glExecutable);

    // When creating monolithic pipelines, the renderer's pipeline cache is used as passed in.
    // When creating the shaders subset of pipelines, the program's own pipeline cache is used.
    vk::PipelineCacheAccess perProgramPipelineCache;
    const bool useProgramPipelineCache = pipelineSubset == vk::GraphicsPipelineSubset::Shaders;
    if (useProgramPipelineCache)
    {
        perProgramPipelineCache.init(&mPipelineCache, nullptr);
        pipelineCache = &perProgramPipelineCache;
    }

    ANGLE_TRY(createGraphicsPipelineImpl(contextVk, transformOptions, pipelineSubset, pipelineCache,
                                         source, desc, glExecutable, descPtrOut, pipelineOut));

    if (useProgramPipelineCache &&
        contextVk->getFeatures().mergeProgramPipelineCachesToGlobalCache.enabled)
    {
        ANGLE_TRY(contextVk->getRenderer()->mergeIntoPipelineCache(mPipelineCache));
    }

    return angle::Result::Continue;
}

angle::Result ProgramExecutableVk::linkGraphicsPipelineLibraries(
    ContextVk *contextVk,
    vk::PipelineCacheAccess *pipelineCache,
    const vk::GraphicsPipelineDesc &desc,
    const gl::ProgramExecutable &glExecutable,
    const vk::PipelineHelper &vertexInputPipeline,
    const vk::PipelineHelper &shadersPipeline,
    const vk::PipelineHelper &fragmentOutputPipeline,
    const vk::GraphicsPipelineDesc **descPtrOut,
    vk::PipelineHelper **pipelineOut)
{
    ProgramTransformOptions transformOptions = getTransformOptions(contextVk, desc, glExecutable);
    const uint8_t programIndex               = GetGraphicsProgramIndex(transformOptions);

    ANGLE_TRY(mCompleteGraphicsPipelines[programIndex].linkLibraries(
        contextVk, pipelineCache, desc, getPipelineLayout(), vertexInputPipeline, shadersPipeline,
        fragmentOutputPipeline, descPtrOut, pipelineOut));

    return angle::Result::Continue;
}

angle::Result ProgramExecutableVk::getOrCreateComputePipeline(
    ContextVk *contextVk,
    vk::PipelineCacheAccess *pipelineCache,
    PipelineSource source,
    const gl::ProgramExecutable &glExecutable,
    vk::PipelineHelper **pipelineOut)
{
    ASSERT(glExecutable.hasLinkedShaderStage(gl::ShaderType::Compute));

    ANGLE_TRY(initComputeProgram(contextVk, &mComputeProgramInfo, mVariableInfoMap));

    vk::ShaderProgramHelper *shaderProgram = mComputeProgramInfo.getShaderProgram();
    ASSERT(shaderProgram);
    return shaderProgram->getOrCreateComputePipeline(
        contextVk, &mComputePipelines, pipelineCache, getPipelineLayout(),
        contextVk->getComputePipelineFlags(), source, pipelineOut);
}

angle::Result ProgramExecutableVk::createPipelineLayout(
    ContextVk *contextVk,
    const gl::ProgramExecutable &glExecutable,
    gl::ActiveTextureArray<TextureVk *> *activeTextures)
{
    gl::TransformFeedback *transformFeedback = contextVk->getState().getCurrentTransformFeedback();
    const gl::ShaderBitSet &linkedShaderStages = glExecutable.getLinkedShaderStages();

    resetLayout(contextVk);

    // Store a reference to the pipeline and descriptor set layouts. This will create them if they
    // don't already exist in the cache.

    // Default uniforms and transform feedback:
    vk::DescriptorSetLayoutDesc uniformsAndXfbSetDesc;
    mNumDefaultUniformDescriptors = 0;
    for (gl::ShaderType shaderType : linkedShaderStages)
    {
        const ShaderInterfaceVariableInfo &info =
            mVariableInfoMap.getDefaultUniformInfo(shaderType);
        if (info.isDuplicate || !info.activeStages[shaderType])
        {
            continue;
        }

        uniformsAndXfbSetDesc.update(info.binding, VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC, 1,
                                     gl_vk::kShaderStageMap[shaderType], nullptr);
        mNumDefaultUniformDescriptors++;
    }

    gl::ShaderType linkedTransformFeedbackStage = glExecutable.getLinkedTransformFeedbackStage();
    bool hasXfbVaryings = linkedTransformFeedbackStage != gl::ShaderType::InvalidEnum &&
                          !glExecutable.getLinkedTransformFeedbackVaryings().empty();
    if (transformFeedback && hasXfbVaryings)
    {
        size_t xfbBufferCount                    = glExecutable.getTransformFeedbackBufferCount();
        TransformFeedbackVk *transformFeedbackVk = vk::GetImpl(transformFeedback);
        transformFeedbackVk->updateDescriptorSetLayout(contextVk, mVariableInfoMap, xfbBufferCount,
                                                       &uniformsAndXfbSetDesc);
    }

    ANGLE_TRY(contextVk->getDescriptorSetLayoutCache().getDescriptorSetLayout(
        contextVk, uniformsAndXfbSetDesc,
        &mDescriptorSetLayouts[DescriptorSetIndex::UniformsAndXfb]));

    // Uniform and storage buffers, atomic counter buffers and images:
    vk::DescriptorSetLayoutDesc resourcesSetDesc;

    // Count the number of active uniform buffer descriptors.
    uint32_t numActiveUniformBufferDescriptors = 0;
    for (gl::ShaderType shaderType : linkedShaderStages)
    {
        const std::vector<gl::InterfaceBlock> &blocks = glExecutable.getUniformBlocks();
        for (uint32_t bufferIndex = 0; bufferIndex < blocks.size();)
        {
            const gl::InterfaceBlock &block = blocks[bufferIndex];
            const uint32_t arraySize        = GetInterfaceBlockArraySize(blocks, bufferIndex);
            bufferIndex += arraySize;

            if (!block.isActive(shaderType))
            {
                continue;
            }

            numActiveUniformBufferDescriptors += arraySize;
        }
    }

    // Decide if we should use dynamic or fixed descriptor types.
    VkPhysicalDeviceLimits limits = contextVk->getRenderer()->getPhysicalDeviceProperties().limits;
    uint32_t totalDynamicUniformBufferCount = numActiveUniformBufferDescriptors +
                                              mNumDefaultUniformDescriptors +
                                              kReservedDriverUniformBindingCount;
    if (totalDynamicUniformBufferCount <= limits.maxDescriptorSetUniformBuffersDynamic)
    {
        mUniformBufferDescriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC;
    }
    else
    {
        mUniformBufferDescriptorType = VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER;
    }

    for (gl::ShaderType shaderType : linkedShaderStages)
    {
        addInterfaceBlockDescriptorSetDesc(glExecutable.getUniformBlocks(), shaderType,
                                           ShaderVariableType::UniformBuffer,
                                           mUniformBufferDescriptorType, &resourcesSetDesc);
        addInterfaceBlockDescriptorSetDesc(glExecutable.getShaderStorageBlocks(), shaderType,
                                           ShaderVariableType::ShaderStorageBuffer,
                                           vk::kStorageBufferDescriptorType, &resourcesSetDesc);
        addAtomicCounterBufferDescriptorSetDesc(glExecutable.getAtomicCounterBuffers(), shaderType,
                                                &resourcesSetDesc);
    }

    for (gl::ShaderType shaderType : linkedShaderStages)
    {
        addImageDescriptorSetDesc(glExecutable, &resourcesSetDesc);
        addInputAttachmentDescriptorSetDesc(glExecutable, shaderType, &resourcesSetDesc);
    }

    ANGLE_TRY(contextVk->getDescriptorSetLayoutCache().getDescriptorSetLayout(
        contextVk, resourcesSetDesc, &mDescriptorSetLayouts[DescriptorSetIndex::ShaderResource]));

    // Textures:
    vk::DescriptorSetLayoutDesc texturesSetDesc;
    ANGLE_TRY(
        addTextureDescriptorSetDesc(contextVk, glExecutable, activeTextures, &texturesSetDesc));

    ANGLE_TRY(contextVk->getDescriptorSetLayoutCache().getDescriptorSetLayout(
        contextVk, texturesSetDesc, &mDescriptorSetLayouts[DescriptorSetIndex::Texture]));

    // Create pipeline layout with these 3 descriptor sets.
    vk::PipelineLayoutDesc pipelineLayoutDesc;
    pipelineLayoutDesc.updateDescriptorSetLayout(DescriptorSetIndex::UniformsAndXfb,
                                                 uniformsAndXfbSetDesc);
    pipelineLayoutDesc.updateDescriptorSetLayout(DescriptorSetIndex::ShaderResource,
                                                 resourcesSetDesc);
    pipelineLayoutDesc.updateDescriptorSetLayout(DescriptorSetIndex::Texture, texturesSetDesc);

    // Set up driver uniforms as push constants. The size is set for a graphics pipeline, as there
    // are more driver uniforms for a graphics pipeline than there are for a compute pipeline. As
    // for the shader stages, both graphics and compute stages are used.
    VkShaderStageFlags pushConstantShaderStageFlags =
        contextVk->getRenderer()->getSupportedVulkanShaderStageMask();

    uint32_t pushConstantSize = contextVk->getDriverUniformSize(PipelineType::Graphics);
    pipelineLayoutDesc.updatePushConstantRange(pushConstantShaderStageFlags, 0, pushConstantSize);

    ANGLE_TRY(contextVk->getPipelineLayoutCache().getPipelineLayout(
        contextVk, pipelineLayoutDesc, mDescriptorSetLayouts, &mPipelineLayout));

    // Initialize descriptor pools.
    ANGLE_TRY(contextVk->bindCachedDescriptorPool(
        DescriptorSetIndex::UniformsAndXfb, uniformsAndXfbSetDesc, 1,
        &mDescriptorPools[DescriptorSetIndex::UniformsAndXfb]));
    ANGLE_TRY(contextVk->bindCachedDescriptorPool(DescriptorSetIndex::Texture, texturesSetDesc,
                                                  mImmutableSamplersMaxDescriptorCount,
                                                  &mDescriptorPools[DescriptorSetIndex::Texture]));
    ANGLE_TRY(
        contextVk->bindCachedDescriptorPool(DescriptorSetIndex::ShaderResource, resourcesSetDesc, 1,
                                            &mDescriptorPools[DescriptorSetIndex::ShaderResource]));

    mDynamicUniformDescriptorOffsets.clear();
    mDynamicUniformDescriptorOffsets.resize(glExecutable.getLinkedShaderStageCount(), 0);

    // If the program uses framebuffer fetch and this is the first time this happens, switch the
    // context to "framebuffer fetch mode".  In this mode, all render passes assume framebuffer
    // fetch may be used, so they are prepared to accept a program that uses input attachments.
    // This is done only when a program with framebuffer fetch is created to avoid potential
    // performance impact on applications that don't use this extension.  If other contexts in the
    // share group use this program, they will lazily switch to this mode.
    if (contextVk->getFeatures().permanentlySwitchToFramebufferFetchMode.enabled &&
        glExecutable.usesFramebufferFetch())
    {
        ANGLE_TRY(contextVk->switchToFramebufferFetchMode(true));
    }

    // Initialize the pipeline cache if not already.  The cache might have been initialized from
    // data from the blob cache.  The per-program cache is used only when the context is not GLES1,
    // and either warmUpPipelineCacheAtLink or supportsGraphicsPipelineLibrary are enabled.
    //
    const bool usePerProgramCache =
        !contextVk->getState().isGLES1() &&
        (contextVk->getFeatures().warmUpPipelineCacheAtLink.enabled ||
         contextVk->getFeatures().supportsGraphicsPipelineLibrary.enabled);
    if (!mPipelineCache.valid() && usePerProgramCache)
    {
        VkPipelineCacheCreateInfo pipelineCacheCreateInfo = {};
        pipelineCacheCreateInfo.sType = VK_STRUCTURE_TYPE_PIPELINE_CACHE_CREATE_INFO;

        if (contextVk->getFeatures().supportsPipelineCreationCacheControl.enabled)
        {
            pipelineCacheCreateInfo.flags |=
                VK_PIPELINE_CACHE_CREATE_EXTERNALLY_SYNCHRONIZED_BIT_EXT;
        }

        ANGLE_VK_TRY(contextVk,
                     mPipelineCache.init(contextVk->getDevice(), pipelineCacheCreateInfo));
    }

    return angle::Result::Continue;
}

void ProgramExecutableVk::resolvePrecisionMismatch(const gl::ProgramMergedVaryings &mergedVaryings)
{
    for (const gl::ProgramVaryingRef &mergedVarying : mergedVaryings)
    {
        if (!mergedVarying.frontShader || !mergedVarying.backShader)
        {
            continue;
        }

        GLenum frontPrecision = mergedVarying.frontShader->precision;
        GLenum backPrecision  = mergedVarying.backShader->precision;
        if (frontPrecision == backPrecision)
        {
            continue;
        }

        ASSERT(frontPrecision >= GL_LOW_FLOAT && frontPrecision <= GL_HIGH_INT);
        ASSERT(backPrecision >= GL_LOW_FLOAT && backPrecision <= GL_HIGH_INT);

        if (frontPrecision > backPrecision)
        {
            // The output is higher precision than the input
            ShaderInterfaceVariableInfo &info = mVariableInfoMap.getMutable(
                mergedVarying.frontShaderStage, ShaderVariableType::Varying,
                mergedVarying.frontShader->mappedName);
            info.varyingIsOutput     = true;
            info.useRelaxedPrecision = true;
        }
        else
        {
            // The output is lower precision than the input, adjust the input
            ASSERT(backPrecision > frontPrecision);
            ShaderInterfaceVariableInfo &info = mVariableInfoMap.getMutable(
                mergedVarying.backShaderStage, ShaderVariableType::Varying,
                mergedVarying.backShader->mappedName);
            info.varyingIsInput      = true;
            info.useRelaxedPrecision = true;
        }
    }
}

angle::Result ProgramExecutableVk::getOrAllocateDescriptorSet(
    vk::Context *context,
    UpdateDescriptorSetsBuilder *updateBuilder,
    vk::CommandBufferHelperCommon *commandBufferHelper,
    const vk::DescriptorSetDescBuilder &descriptorSetDesc,
    DescriptorSetIndex setIndex,
    vk::SharedDescriptorSetCacheKey *newSharedCacheKeyOut)
{
    ANGLE_TRY(mDescriptorPools[setIndex].get().getOrAllocateDescriptorSet(
        context, commandBufferHelper, descriptorSetDesc.getDesc(),
        mDescriptorSetLayouts[setIndex].get(), &mDescriptorPoolBindings[setIndex],
        &mDescriptorSets[setIndex], newSharedCacheKeyOut));
    ASSERT(mDescriptorSets[setIndex] != VK_NULL_HANDLE);

    if (*newSharedCacheKeyOut != nullptr)
    {
        // Cache miss. A new cache entry has been created.
        descriptorSetDesc.updateDescriptorSet(updateBuilder, mDescriptorSets[setIndex]);
    }
    else
    {
        commandBufferHelper->retainResource(&mDescriptorPoolBindings[setIndex].get());
    }

    return angle::Result::Continue;
}

angle::Result ProgramExecutableVk::updateShaderResourcesDescriptorSet(
    vk::Context *context,
    UpdateDescriptorSetsBuilder *updateBuilder,
    vk::CommandBufferHelperCommon *commandBufferHelper,
    const vk::DescriptorSetDescBuilder &shaderResourcesDesc,
    vk::SharedDescriptorSetCacheKey *newSharedCacheKeyOut)
{
    if (!mDescriptorPools[DescriptorSetIndex::ShaderResource].get().valid())
    {
        *newSharedCacheKeyOut = nullptr;
        return angle::Result::Continue;
    }

    ANGLE_TRY(getOrAllocateDescriptorSet(context, updateBuilder, commandBufferHelper,
                                         shaderResourcesDesc, DescriptorSetIndex::ShaderResource,
                                         newSharedCacheKeyOut));

    size_t numOffsets = shaderResourcesDesc.getDynamicOffsetsSize();
    mDynamicShaderResourceDescriptorOffsets.resize(numOffsets);
    if (numOffsets > 0)
    {
        memcpy(mDynamicShaderResourceDescriptorOffsets.data(),
               shaderResourcesDesc.getDynamicOffsets(), numOffsets * sizeof(uint32_t));
    }

    return angle::Result::Continue;
}

angle::Result ProgramExecutableVk::updateUniformsAndXfbDescriptorSet(
    vk::Context *context,
    UpdateDescriptorSetsBuilder *updateBuilder,
    vk::CommandBufferHelperCommon *commandBufferHelper,
    vk::BufferHelper *defaultUniformBuffer,
    vk::DescriptorSetDescBuilder *uniformsAndXfbDesc)
{
    mCurrentDefaultUniformBufferSerial =
        defaultUniformBuffer ? defaultUniformBuffer->getBufferSerial() : vk::kInvalidBufferSerial;

    vk::SharedDescriptorSetCacheKey newSharedCacheKey;
    ANGLE_TRY(getOrAllocateDescriptorSet(context, updateBuilder, commandBufferHelper,
                                         *uniformsAndXfbDesc, DescriptorSetIndex::UniformsAndXfb,
                                         &newSharedCacheKey));
    uniformsAndXfbDesc->updateImagesAndBuffersWithSharedCacheKey(newSharedCacheKey);
    return angle::Result::Continue;
}

angle::Result ProgramExecutableVk::updateTexturesDescriptorSet(
    vk::Context *context,
    const gl::ProgramExecutable &executable,
    const gl::ActiveTextureArray<TextureVk *> &textures,
    const gl::SamplerBindingVector &samplers,
    bool emulateSeamfulCubeMapSampling,
    PipelineType pipelineType,
    UpdateDescriptorSetsBuilder *updateBuilder,
    vk::CommandBufferHelperCommon *commandBufferHelper,
    const vk::DescriptorSetDesc &texturesDesc)
{
    vk::SharedDescriptorSetCacheKey newSharedCacheKey;
    ANGLE_TRY(mDescriptorPools[DescriptorSetIndex::Texture].get().getOrAllocateDescriptorSet(
        context, commandBufferHelper, texturesDesc,
        mDescriptorSetLayouts[DescriptorSetIndex::Texture].get(),
        &mDescriptorPoolBindings[DescriptorSetIndex::Texture],
        &mDescriptorSets[DescriptorSetIndex::Texture], &newSharedCacheKey));
    ASSERT(mDescriptorSets[DescriptorSetIndex::Texture] != VK_NULL_HANDLE);

    if (newSharedCacheKey != nullptr)
    {
        vk::DescriptorSetDescBuilder fullDesc;
        // Cache miss. A new cache entry has been created.
        ANGLE_TRY(fullDesc.updateFullActiveTextures(context, mVariableInfoMap, executable, textures,
                                                    samplers, emulateSeamfulCubeMapSampling,
                                                    pipelineType, newSharedCacheKey));
        fullDesc.updateDescriptorSet(updateBuilder, mDescriptorSets[DescriptorSetIndex::Texture]);
    }
    else
    {
        commandBufferHelper->retainResource(
            &mDescriptorPoolBindings[DescriptorSetIndex::Texture].get());
    }

    return angle::Result::Continue;
}

template <typename CommandBufferT>
angle::Result ProgramExecutableVk::bindDescriptorSets(
    vk::Context *context,
    vk::CommandBufferHelperCommon *commandBufferHelper,
    CommandBufferT *commandBuffer,
    PipelineType pipelineType)
{
    // Can probably use better dirty bits here.

    // Find the maximum non-null descriptor set.  This is used in conjunction with a driver
    // workaround to bind empty descriptor sets only for gaps in between 0 and max and avoid
    // binding unnecessary empty descriptor sets for the sets beyond max.
    DescriptorSetIndex lastNonNullDescriptorSetIndex = DescriptorSetIndex::InvalidEnum;
    for (DescriptorSetIndex descriptorSetIndex : angle::AllEnums<DescriptorSetIndex>())
    {
        if (mDescriptorSets[descriptorSetIndex] != VK_NULL_HANDLE)
        {
            lastNonNullDescriptorSetIndex = descriptorSetIndex;
        }
    }

    const VkPipelineBindPoint pipelineBindPoint = pipelineType == PipelineType::Compute
                                                      ? VK_PIPELINE_BIND_POINT_COMPUTE
                                                      : VK_PIPELINE_BIND_POINT_GRAPHICS;

    for (DescriptorSetIndex descriptorSetIndex : angle::AllEnums<DescriptorSetIndex>())
    {
        if (ToUnderlying(descriptorSetIndex) > ToUnderlying(lastNonNullDescriptorSetIndex))
        {
            continue;
        }

        VkDescriptorSet descSet = mDescriptorSets[descriptorSetIndex];
        if (descSet == VK_NULL_HANDLE)
        {
            continue;
        }

        // Default uniforms are encompassed in a block per shader stage, and they are assigned
        // through dynamic uniform buffers (requiring dynamic offsets).  No other descriptor
        // requires a dynamic offset.
        if (descriptorSetIndex == DescriptorSetIndex::UniformsAndXfb)
        {
            commandBuffer->bindDescriptorSets(
                getPipelineLayout(), pipelineBindPoint, descriptorSetIndex, 1, &descSet,
                static_cast<uint32_t>(mDynamicUniformDescriptorOffsets.size()),
                mDynamicUniformDescriptorOffsets.data());
        }
        else if (descriptorSetIndex == DescriptorSetIndex::ShaderResource)
        {
            commandBuffer->bindDescriptorSets(
                getPipelineLayout(), pipelineBindPoint, descriptorSetIndex, 1, &descSet,
                static_cast<uint32_t>(mDynamicShaderResourceDescriptorOffsets.size()),
                mDynamicShaderResourceDescriptorOffsets.data());
        }
        else
        {
            commandBuffer->bindDescriptorSets(getPipelineLayout(), pipelineBindPoint,
                                              descriptorSetIndex, 1, &descSet, 0, nullptr);
        }
    }

    return angle::Result::Continue;
}

template angle::Result ProgramExecutableVk::bindDescriptorSets<vk::priv::SecondaryCommandBuffer>(
    vk::Context *context,
    vk::CommandBufferHelperCommon *commandBufferHelper,
    vk::priv::SecondaryCommandBuffer *commandBuffer,
    PipelineType pipelineType);
template angle::Result ProgramExecutableVk::bindDescriptorSets<vk::VulkanSecondaryCommandBuffer>(
    vk::Context *context,
    vk::CommandBufferHelperCommon *commandBufferHelper,
    vk::VulkanSecondaryCommandBuffer *commandBuffer,
    PipelineType pipelineType);

void ProgramExecutableVk::setAllDefaultUniformsDirty(const gl::ProgramExecutable &executable)
{
    mDefaultUniformBlocksDirty.reset();
    for (gl::ShaderType shaderType : executable.getLinkedShaderStages())
    {
        if (!mDefaultUniformBlocks[shaderType]->uniformData.empty())
        {
            mDefaultUniformBlocksDirty.set(shaderType);
        }
    }
}

angle::Result ProgramExecutableVk::updateUniforms(
    vk::Context *context,
    UpdateDescriptorSetsBuilder *updateBuilder,
    vk::CommandBufferHelperCommon *commandBufferHelper,
    vk::BufferHelper *emptyBuffer,
    const gl::ProgramExecutable &glExecutable,
    vk::DynamicBuffer *defaultUniformStorage,
    bool isTransformFeedbackActiveUnpaused,
    TransformFeedbackVk *transformFeedbackVk)
{
    ASSERT(hasDirtyUniforms());

    vk::BufferHelper *defaultUniformBuffer;
    bool anyNewBufferAllocated          = false;
    gl::ShaderMap<VkDeviceSize> offsets = {};  // offset to the beginning of bufferData
    uint32_t offsetIndex                = 0;
    size_t requiredSpace;

    // We usually only update uniform data for shader stages that are actually dirty. But when the
    // buffer for uniform data have switched, because all shader stages are using the same buffer,
    // we then must update uniform data for all shader stages to keep all shader stages' uniform
    // data in the same buffer.
    requiredSpace = calcUniformUpdateRequiredSpace(context, glExecutable, &offsets);
    ASSERT(requiredSpace > 0);

    // Allocate space from dynamicBuffer. Always try to allocate from the current buffer first.
    // If that failed, we deal with fall out and try again.
    if (!defaultUniformStorage->allocateFromCurrentBuffer(requiredSpace, &defaultUniformBuffer))
    {
        setAllDefaultUniformsDirty(glExecutable);

        requiredSpace = calcUniformUpdateRequiredSpace(context, glExecutable, &offsets);
        ANGLE_TRY(defaultUniformStorage->allocate(context, requiredSpace, &defaultUniformBuffer,
                                                  &anyNewBufferAllocated));
    }

    ASSERT(defaultUniformBuffer);

    uint8_t *bufferData       = defaultUniformBuffer->getMappedMemory();
    VkDeviceSize bufferOffset = defaultUniformBuffer->getOffset();
    for (gl::ShaderType shaderType : glExecutable.getLinkedShaderStages())
    {
        if (mDefaultUniformBlocksDirty[shaderType])
        {
            const angle::MemoryBuffer &uniformData = mDefaultUniformBlocks[shaderType]->uniformData;
            memcpy(&bufferData[offsets[shaderType]], uniformData.data(), uniformData.size());
            mDynamicUniformDescriptorOffsets[offsetIndex] =
                static_cast<uint32_t>(bufferOffset + offsets[shaderType]);
            mDefaultUniformBlocksDirty.reset(shaderType);
        }
        ++offsetIndex;
    }
    ANGLE_TRY(defaultUniformBuffer->flush(context->getRenderer()));

    // Because the uniform buffers are per context, we can't rely on dynamicBuffer's allocate
    // function to tell us if you have got a new buffer or not. Other program's use of the buffer
    // might already pushed dynamicBuffer to a new buffer. We record which buffer (represented by
    // the unique BufferSerial number) we were using with the current descriptor set and then we
    // use that recorded BufferSerial compare to the current uniform buffer to quickly detect if
    // there is a buffer switch or not. We need to retrieve from the descriptor set cache or
    // allocate a new descriptor set whenever there is uniform buffer switch.
    if (mCurrentDefaultUniformBufferSerial != defaultUniformBuffer->getBufferSerial())
    {
        // We need to reinitialize the descriptor sets if we newly allocated buffers since we can't
        // modify the descriptor sets once initialized.
        vk::DescriptorSetDescBuilder uniformsAndXfbDesc;
        uniformsAndXfbDesc.updateUniformsAndXfb(
            context, glExecutable, *this, defaultUniformBuffer, *emptyBuffer,
            isTransformFeedbackActiveUnpaused,
            glExecutable.hasTransformFeedbackOutput() ? transformFeedbackVk : nullptr);

        ANGLE_TRY(updateUniformsAndXfbDescriptorSet(context, updateBuilder, commandBufferHelper,
                                                    defaultUniformBuffer, &uniformsAndXfbDesc));
    }

    return angle::Result::Continue;
}

size_t ProgramExecutableVk::calcUniformUpdateRequiredSpace(
    vk::Context *context,
    const gl::ProgramExecutable &glExecutable,
    gl::ShaderMap<VkDeviceSize> *uniformOffsets) const
{
    size_t requiredSpace = 0;
    for (gl::ShaderType shaderType : glExecutable.getLinkedShaderStages())
    {
        if (mDefaultUniformBlocksDirty[shaderType])
        {
            (*uniformOffsets)[shaderType] = requiredSpace;
            requiredSpace += getDefaultUniformAlignedSize(context, shaderType);
        }
    }
    return requiredSpace;
}

void ProgramExecutableVk::onProgramBind(const gl::ProgramExecutable &glExecutable)
{
    // Because all programs share default uniform buffers, when we switch programs, we have to
    // re-update all uniform data. We could do more tracking to avoid update if the context's
    // current uniform buffer is still the same buffer we last time used and buffer has not been
    // recycled. But statistics gathered on gfxbench shows that app always update uniform data on
    // program bind anyway, so not really worth it to add more tracking logic here.
    setAllDefaultUniformsDirty(glExecutable);
}

angle::Result ProgramExecutableVk::resizeUniformBlockMemory(
    ContextVk *contextVk,
    const gl::ProgramExecutable &glExecutable,
    const gl::ShaderMap<size_t> &requiredBufferSize)
{
    for (gl::ShaderType shaderType : glExecutable.getLinkedShaderStages())
    {
        if (requiredBufferSize[shaderType] > 0)
        {
            if (!mDefaultUniformBlocks[shaderType]->uniformData.resize(
                    requiredBufferSize[shaderType]))
            {
                ANGLE_VK_CHECK(contextVk, false, VK_ERROR_OUT_OF_HOST_MEMORY);
            }

            // Initialize uniform buffer memory to zero by default.
            mDefaultUniformBlocks[shaderType]->uniformData.fill(0);
            mDefaultUniformBlocksDirty.set(shaderType);
        }
    }

    return angle::Result::Continue;
}
}  // namespace rx