renderer/d3d/ShaderD3D.cpp

//
// Copyright 2014 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.
//

// ShaderD3D.cpp: Defines the rx::ShaderD3D class which implements rx::ShaderImpl.

#include "libANGLE/renderer/d3d/ShaderD3D.h"

#include "common/utilities.h"
#include "libANGLE/Caps.h"
#include "libANGLE/Compiler.h"
#include "libANGLE/Context.h"
#include "libANGLE/Shader.h"
#include "libANGLE/features.h"
#include "libANGLE/renderer/d3d/ProgramD3D.h"
#include "libANGLE/renderer/d3d/RendererD3D.h"
#include "libANGLE/trace.h"

namespace rx
{

class TranslateTaskD3D : public angle::Closure
{
  public:
    TranslateTaskD3D(ShHandle handle,
                     ShCompileOptions options,
                     const std::string &source,
                     const std::string &sourcePath)
        : mHandle(handle),
          mOptions(options),
          mSource(source),
          mSourcePath(sourcePath),
          mResult(false)
    {}

    void operator()() override
    {
        ANGLE_TRACE_EVENT1("gpu.angle", "TranslateTask::run", "source", mSource);
        std::vector<const char *> srcStrings;
        if (!mSourcePath.empty())
        {
            srcStrings.push_back(mSourcePath.c_str());
        }
        srcStrings.push_back(mSource.c_str());

        mResult = sh::Compile(mHandle, &srcStrings[0], srcStrings.size(), mOptions);
    }

    bool getResult() { return mResult; }

  private:
    ShHandle mHandle;
    ShCompileOptions mOptions;
    std::string mSource;
    std::string mSourcePath;
    bool mResult;
};

using PostTranslateFunctor =
    std::function<bool(gl::ShCompilerInstance *compiler, std::string *infoLog)>;

class WaitableCompileEventD3D final : public WaitableCompileEvent
{
  public:
    WaitableCompileEventD3D(std::shared_ptr<angle::WaitableEvent> waitableEvent,
                            gl::ShCompilerInstance *compilerInstance,
                            PostTranslateFunctor &&postTranslateFunctor,
                            std::shared_ptr<TranslateTaskD3D> translateTask)
        : WaitableCompileEvent(waitableEvent),
          mCompilerInstance(compilerInstance),
          mPostTranslateFunctor(std::move(postTranslateFunctor)),
          mTranslateTask(translateTask)
    {}

    bool getResult() override { return mTranslateTask->getResult(); }

    bool postTranslate(std::string *infoLog) override
    {
        return mPostTranslateFunctor(mCompilerInstance, infoLog);
    }

  private:
    gl::ShCompilerInstance *mCompilerInstance;
    PostTranslateFunctor mPostTranslateFunctor;
    std::shared_ptr<TranslateTaskD3D> mTranslateTask;
};

ShaderD3D::ShaderD3D(const gl::ShaderState &state,
                     const angle::FeaturesD3D &features,
                     const gl::Extensions &extensions)
    : ShaderImpl(state), mAdditionalOptions(0)
{
    uncompile();

    if (features.expandIntegerPowExpressions.enabled)
    {
        mAdditionalOptions |= SH_EXPAND_SELECT_HLSL_INTEGER_POW_EXPRESSIONS;
    }

    if (features.getDimensionsIgnoresBaseLevel.enabled)
    {
        mAdditionalOptions |= SH_HLSL_GET_DIMENSIONS_IGNORES_BASE_LEVEL;
    }

    if (features.preAddTexelFetchOffsets.enabled)
    {
        mAdditionalOptions |= SH_REWRITE_TEXELFETCHOFFSET_TO_TEXELFETCH;
    }
    if (features.rewriteUnaryMinusOperator.enabled)
    {
        mAdditionalOptions |= SH_REWRITE_INTEGER_UNARY_MINUS_OPERATOR;
    }
    if (features.emulateIsnanFloat.enabled)
    {
        mAdditionalOptions |= SH_EMULATE_ISNAN_FLOAT_FUNCTION;
    }
    if (features.skipVSConstantRegisterZero.enabled &&
        mState.getShaderType() == gl::ShaderType::Vertex)
    {
        mAdditionalOptions |= SH_SKIP_D3D_CONSTANT_REGISTER_ZERO;
    }
    if (features.forceAtomicValueResolution.enabled)
    {
        mAdditionalOptions |= SH_FORCE_ATOMIC_VALUE_RESOLUTION;
    }
    if (features.allowTranslateUniformBlockToStructuredBuffer.enabled)
    {
        mAdditionalOptions |= SH_ALLOW_TRANSLATE_UNIFORM_BLOCK_TO_STRUCTUREDBUFFER;
    }
    if (extensions.multiview || extensions.multiview2)
    {
        mAdditionalOptions |= SH_INITIALIZE_BUILTINS_FOR_INSTANCED_MULTIVIEW;
    }
}

ShaderD3D::~ShaderD3D() {}

std::string ShaderD3D::getDebugInfo() const
{
    if (mDebugInfo.empty())
    {
        return "";
    }

    return mDebugInfo + std::string("\n// ") + gl::GetShaderTypeString(mState.getShaderType()) +
           " SHADER END\n";
}

// initialize/clean up previous state
void ShaderD3D::uncompile()
{
    // set by compileToHLSL
    mCompilerOutputType = SH_ESSL_OUTPUT;

    mUsesMultipleRenderTargets   = false;
    mUsesFragColor               = false;
    mUsesFragData                = false;
    mUsesSecondaryColor          = false;
    mUsesFragCoord               = false;
    mUsesFrontFacing             = false;
    mUsesHelperInvocation        = false;
    mUsesPointSize               = false;
    mUsesPointCoord              = false;
    mUsesDepthRange              = false;
    mUsesFragDepth               = false;
    mHasANGLEMultiviewEnabled    = false;
    mUsesVertexID                = false;
    mUsesViewID                  = false;
    mUsesDiscardRewriting        = false;
    mUsesNestedBreak             = false;
    mRequiresIEEEStrictCompiling = false;

    mDebugInfo.clear();
}

void ShaderD3D::generateWorkarounds(angle::CompilerWorkaroundsD3D *workarounds) const
{
    if (mUsesDiscardRewriting)
    {
        // ANGLE issue 486:
        // Work-around a D3D9 compiler bug that presents itself when using conditional discard, by
        // disabling optimization
        workarounds->skipOptimization = true;
    }
    else if (mUsesNestedBreak)
    {
        // ANGLE issue 603:
        // Work-around a D3D9 compiler bug that presents itself when using break in a nested loop,
        // by maximizing optimization We want to keep the use of
        // ANGLE_D3D_WORKAROUND_MAX_OPTIMIZATION minimal to prevent hangs, so usesDiscard takes
        // precedence
        workarounds->useMaxOptimization = true;
    }

    if (mRequiresIEEEStrictCompiling)
    {
        // IEEE Strictness for D3D compiler needs to be enabled for NaNs to work.
        workarounds->enableIEEEStrictness = true;
    }
}

unsigned int ShaderD3D::getUniformRegister(const std::string &uniformName) const
{
    ASSERT(mUniformRegisterMap.count(uniformName) > 0);
    return mUniformRegisterMap.find(uniformName)->second;
}

unsigned int ShaderD3D::getUniformBlockRegister(const std::string &blockName) const
{
    ASSERT(mUniformBlockRegisterMap.count(blockName) > 0);
    return mUniformBlockRegisterMap.find(blockName)->second;
}

bool ShaderD3D::shouldUniformBlockUseStructuredBuffer(const std::string &blockName) const
{
    ASSERT(mUniformBlockUseStructuredBufferMap.count(blockName) > 0);
    return mUniformBlockUseStructuredBufferMap.find(blockName)->second;
}

unsigned int ShaderD3D::getShaderStorageBlockRegister(const std::string &blockName) const
{
    ASSERT(mShaderStorageBlockRegisterMap.count(blockName) > 0);
    return mShaderStorageBlockRegisterMap.find(blockName)->second;
}

ShShaderOutput ShaderD3D::getCompilerOutputType() const
{
    return mCompilerOutputType;
}

bool ShaderD3D::useImage2DFunction(const std::string &functionName) const
{
    if (mUsedImage2DFunctionNames.empty())
    {
        return false;
    }

    return mUsedImage2DFunctionNames.find(functionName) != mUsedImage2DFunctionNames.end();
}

const std::set<std::string> &ShaderD3D::getSlowCompilingUniformBlockSet() const
{
    return mSlowCompilingUniformBlockSet;
}

const std::map<std::string, unsigned int> &GetUniformRegisterMap(
    const std::map<std::string, unsigned int> *uniformRegisterMap)
{
    ASSERT(uniformRegisterMap);
    return *uniformRegisterMap;
}

const std::set<std::string> &GetSlowCompilingUniformBlockSet(
    const std::set<std::string> *slowCompilingUniformBlockSet)
{
    ASSERT(slowCompilingUniformBlockSet);
    return *slowCompilingUniformBlockSet;
}

const std::set<std::string> &GetUsedImage2DFunctionNames(
    const std::set<std::string> *usedImage2DFunctionNames)
{
    ASSERT(usedImage2DFunctionNames);
    return *usedImage2DFunctionNames;
}

std::shared_ptr<WaitableCompileEvent> ShaderD3D::compile(const gl::Context *context,
                                                         gl::ShCompilerInstance *compilerInstance,
                                                         ShCompileOptions options)
{
    std::string sourcePath;
    uncompile();

    ShCompileOptions additionalOptions = 0;

    const std::string &source = mState.getSource();

#if !defined(ANGLE_ENABLE_WINDOWS_UWP)
    if (gl::DebugAnnotationsActive())
    {
        sourcePath = getTempPath();
        writeFile(sourcePath.c_str(), source.c_str(), source.length());
        additionalOptions |= SH_LINE_DIRECTIVES | SH_SOURCE_PATH;
    }
#endif

    additionalOptions |= mAdditionalOptions;

    options |= additionalOptions;

    auto postTranslateFunctor = [this](gl::ShCompilerInstance *compiler, std::string *infoLog) {
        // TODO(jmadill): We shouldn't need to cache this.
        mCompilerOutputType = compiler->getShaderOutputType();

        const std::string &translatedSource = mState.getTranslatedSource();

        mUsesMultipleRenderTargets = translatedSource.find("GL_USES_MRT") != std::string::npos;
        mUsesFragColor      = translatedSource.find("GL_USES_FRAG_COLOR") != std::string::npos;
        mUsesFragData       = translatedSource.find("GL_USES_FRAG_DATA") != std::string::npos;
        mUsesSecondaryColor = translatedSource.find("GL_USES_SECONDARY_COLOR") != std::string::npos;
        mUsesFragCoord      = translatedSource.find("GL_USES_FRAG_COORD") != std::string::npos;
        mUsesFrontFacing    = translatedSource.find("GL_USES_FRONT_FACING") != std::string::npos;
        mUsesHelperInvocation =
            translatedSource.find("GL_USES_HELPER_INVOCATION") != std::string::npos;
        mUsesPointSize  = translatedSource.find("GL_USES_POINT_SIZE") != std::string::npos;
        mUsesPointCoord = translatedSource.find("GL_USES_POINT_COORD") != std::string::npos;
        mUsesDepthRange = translatedSource.find("GL_USES_DEPTH_RANGE") != std::string::npos;
        mUsesFragDepth  = translatedSource.find("GL_USES_FRAG_DEPTH") != std::string::npos;
        mHasANGLEMultiviewEnabled =
            translatedSource.find("GL_ANGLE_MULTIVIEW_ENABLED") != std::string::npos;
        mUsesVertexID = translatedSource.find("GL_USES_VERTEX_ID") != std::string::npos;
        mUsesViewID   = translatedSource.find("GL_USES_VIEW_ID") != std::string::npos;
        mUsesDiscardRewriting =
            translatedSource.find("ANGLE_USES_DISCARD_REWRITING") != std::string::npos;
        mUsesNestedBreak = translatedSource.find("ANGLE_USES_NESTED_BREAK") != std::string::npos;
        mRequiresIEEEStrictCompiling =
            translatedSource.find("ANGLE_REQUIRES_IEEE_STRICT_COMPILING") != std::string::npos;

        ShHandle compilerHandle = compiler->getHandle();

        mUniformRegisterMap = GetUniformRegisterMap(sh::GetUniformRegisterMap(compilerHandle));
        mReadonlyImage2DRegisterIndex = sh::GetReadonlyImage2DRegisterIndex(compilerHandle);
        mImage2DRegisterIndex         = sh::GetImage2DRegisterIndex(compilerHandle);
        mUsedImage2DFunctionNames =
            GetUsedImage2DFunctionNames(sh::GetUsedImage2DFunctionNames(compilerHandle));

        for (const sh::InterfaceBlock &interfaceBlock : mState.getUniformBlocks())
        {
            if (interfaceBlock.active)
            {
                unsigned int index = static_cast<unsigned int>(-1);
                bool blockRegisterResult =
                    sh::GetUniformBlockRegister(compilerHandle, interfaceBlock.name, &index);
                ASSERT(blockRegisterResult);
                bool useStructuredBuffer =
                    sh::ShouldUniformBlockUseStructuredBuffer(compilerHandle, interfaceBlock.name);

                mUniformBlockRegisterMap[interfaceBlock.name]            = index;
                mUniformBlockUseStructuredBufferMap[interfaceBlock.name] = useStructuredBuffer;
            }
        }

        mSlowCompilingUniformBlockSet =
            GetSlowCompilingUniformBlockSet(sh::GetSlowCompilingUniformBlockSet(compilerHandle));

        for (const sh::InterfaceBlock &interfaceBlock : mState.getShaderStorageBlocks())
        {
            if (interfaceBlock.active)
            {
                unsigned int index = static_cast<unsigned int>(-1);
                bool blockRegisterResult =
                    sh::GetShaderStorageBlockRegister(compilerHandle, interfaceBlock.name, &index);
                ASSERT(blockRegisterResult);

                mShaderStorageBlockRegisterMap[interfaceBlock.name] = index;
            }
        }

        mDebugInfo += std::string("// ") + gl::GetShaderTypeString(mState.getShaderType()) +
                      " SHADER BEGIN\n";
        mDebugInfo += "\n// GLSL BEGIN\n\n" + mState.getSource() + "\n\n// GLSL END\n\n\n";
        mDebugInfo +=
            "// INITIAL HLSL BEGIN\n\n" + translatedSource + "\n// INITIAL HLSL END\n\n\n";
        // Successive steps will append more info
        return true;
    };

    auto workerThreadPool = context->getWorkerThreadPool();
    auto translateTask = std::make_shared<TranslateTaskD3D>(compilerInstance->getHandle(), options,
                                                            source, sourcePath);

    return std::make_shared<WaitableCompileEventD3D>(
        angle::WorkerThreadPool::PostWorkerTask(workerThreadPool, translateTask), compilerInstance,
        std::move(postTranslateFunctor), translateTask);
}

bool ShaderD3D::hasUniform(const std::string &name) const
{
    return mUniformRegisterMap.find(name) != mUniformRegisterMap.end();
}

}  // namespace rx