blob: 18ff4df792d48bc61391134b09542ef912390c52 [file] [log] [blame]
//
// Copyright 2015 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.
//
// ProgramGL.cpp: Implements the class methods for ProgramGL.
#include "libANGLE/renderer/gl/ProgramGL.h"
#include "common/WorkerThread.h"
#include "common/angleutils.h"
#include "common/bitset_utils.h"
#include "common/debug.h"
#include "common/string_utils.h"
#include "common/utilities.h"
#include "libANGLE/Context.h"
#include "libANGLE/ProgramLinkedResources.h"
#include "libANGLE/Uniform.h"
#include "libANGLE/queryconversions.h"
#include "libANGLE/renderer/gl/ContextGL.h"
#include "libANGLE/renderer/gl/FunctionsGL.h"
#include "libANGLE/renderer/gl/RendererGL.h"
#include "libANGLE/renderer/gl/ShaderGL.h"
#include "libANGLE/renderer/gl/StateManagerGL.h"
#include "libANGLE/trace.h"
#include "platform/PlatformMethods.h"
#include "platform/autogen/FeaturesGL_autogen.h"
namespace rx
{
namespace
{
// Returns mapped name of a transform feedback varying. The original name may contain array
// brackets with an index inside, which will get copied to the mapped name. The varying must be
// known to be declared in the shader.
std::string GetTransformFeedbackVaryingMappedName(const gl::SharedCompiledShaderState &shaderState,
const std::string &tfVaryingName)
{
ASSERT(shaderState->shaderType != gl::ShaderType::Fragment &&
shaderState->shaderType != gl::ShaderType::Compute);
const auto &varyings = shaderState->outputVaryings;
auto bracketPos = tfVaryingName.find("[");
if (bracketPos != std::string::npos)
{
auto tfVaryingBaseName = tfVaryingName.substr(0, bracketPos);
for (const auto &varying : varyings)
{
if (varying.name == tfVaryingBaseName)
{
std::string mappedNameWithArrayIndex =
varying.mappedName + tfVaryingName.substr(bracketPos);
return mappedNameWithArrayIndex;
}
}
}
else
{
for (const auto &varying : varyings)
{
if (varying.name == tfVaryingName)
{
return varying.mappedName;
}
else if (varying.isStruct())
{
GLuint fieldIndex = 0;
const auto *field = varying.findField(tfVaryingName, &fieldIndex);
if (field == nullptr)
{
continue;
}
ASSERT(field != nullptr && !field->isStruct() &&
(!field->isArray() || varying.isShaderIOBlock));
std::string mappedName;
// If it's an I/O block without an instance name, don't include the block name.
if (!varying.isShaderIOBlock || !varying.name.empty())
{
mappedName = varying.isShaderIOBlock ? varying.mappedStructOrBlockName
: varying.mappedName;
mappedName += '.';
}
return mappedName + field->mappedName;
}
}
}
UNREACHABLE();
return std::string();
}
} // anonymous namespace
class ProgramGL::LinkTaskGL final : public LinkTask
{
public:
LinkTaskGL(ProgramGL *program,
bool hasNativeParallelCompile,
const FunctionsGL *functions,
const gl::Extensions &extensions,
GLuint programID)
: mProgram(program),
mHasNativeParallelCompile(hasNativeParallelCompile),
mFunctions(functions),
mExtensions(extensions),
mProgramID(programID)
{}
~LinkTaskGL() override = default;
std::vector<std::shared_ptr<LinkSubTask>> link(const gl::ProgramLinkedResources &resources,
const gl::ProgramMergedVaryings &mergedVaryings,
bool *areSubTasksOptionalOut) override
{
mProgram->linkJobImpl(mExtensions);
// If there is no native parallel compile, do the post-link right away.
if (!mHasNativeParallelCompile)
{
mResult = mProgram->postLinkJobImpl(resources);
}
// See comment on mResources
mResources = &resources;
return {};
}
angle::Result getResult(const gl::Context *context, gl::InfoLog &infoLog) override
{
ANGLE_TRACE_EVENT0("gpu.angle", "LinkTaskGL::getResult");
if (mHasNativeParallelCompile)
{
mResult = mProgram->postLinkJobImpl(*mResources);
}
return mResult;
}
bool isLinkingInternally() override
{
GLint completionStatus = GL_TRUE;
if (mHasNativeParallelCompile)
{
mFunctions->getProgramiv(mProgramID, GL_COMPLETION_STATUS, &completionStatus);
}
return completionStatus == GL_FALSE;
}
private:
ProgramGL *mProgram;
const bool mHasNativeParallelCompile;
const FunctionsGL *mFunctions;
const gl::Extensions &mExtensions;
const GLuint mProgramID;
angle::Result mResult = angle::Result::Continue;
// Note: resources are kept alive by the front-end for the entire duration of the link,
// including during resolve when getResult() and postLink() are called.
const gl::ProgramLinkedResources *mResources = nullptr;
};
ProgramGL::ProgramGL(const gl::ProgramState &data,
const FunctionsGL *functions,
const angle::FeaturesGL &features,
StateManagerGL *stateManager,
const std::shared_ptr<RendererGL> &renderer)
: ProgramImpl(data),
mFunctions(functions),
mFeatures(features),
mStateManager(stateManager),
mProgramID(0),
mRenderer(renderer)
{
ASSERT(mFunctions);
ASSERT(mStateManager);
mProgramID = mFunctions->createProgram();
}
ProgramGL::~ProgramGL() = default;
void ProgramGL::destroy(const gl::Context *context)
{
mFunctions->deleteProgram(mProgramID);
mProgramID = 0;
}
angle::Result ProgramGL::load(const gl::Context *context,
gl::BinaryInputStream *stream,
std::shared_ptr<LinkTask> *loadTaskOut,
egl::CacheGetResult *resultOut)
{
ANGLE_TRACE_EVENT0("gpu.angle", "ProgramGL::load");
ProgramExecutableGL *executableGL = getExecutable();
// Read the binary format, size and blob
GLenum binaryFormat = stream->readInt<GLenum>();
GLint binaryLength = stream->readInt<GLint>();
const uint8_t *binary = stream->data() + stream->offset();
stream->skip(binaryLength);
// Load the binary
mFunctions->programBinary(mProgramID, binaryFormat, binary, binaryLength);
// Verify that the program linked. Ensure failure if program binary is intentionally corrupted,
// even if the corruption didn't really cause a failure.
if (!checkLinkStatus() ||
GetImplAs<ContextGL>(context)->getFeaturesGL().corruptProgramBinaryForTesting.enabled)
{
return angle::Result::Continue;
}
executableGL->postLink(mFunctions, mStateManager, mFeatures, mProgramID);
reapplyUBOBindingsIfNeeded(context);
*loadTaskOut = {};
*resultOut = egl::CacheGetResult::GetSuccess;
return angle::Result::Continue;
}
void ProgramGL::save(const gl::Context *context, gl::BinaryOutputStream *stream)
{
GLint binaryLength = 0;
mFunctions->getProgramiv(mProgramID, GL_PROGRAM_BINARY_LENGTH, &binaryLength);
std::vector<uint8_t> binary(std::max(binaryLength, 1));
GLenum binaryFormat = GL_NONE;
mFunctions->getProgramBinary(mProgramID, binaryLength, &binaryLength, &binaryFormat,
binary.data());
stream->writeInt(binaryFormat);
stream->writeInt(binaryLength);
if (GetImplAs<ContextGL>(context)->getFeaturesGL().corruptProgramBinaryForTesting.enabled)
{
// Random corruption of the binary data. Corrupting the first byte has proven to be enough
// to later cause the binary load to fail on most platforms.
++binary[0];
}
stream->writeBytes(binary.data(), binaryLength);
reapplyUBOBindingsIfNeeded(context);
}
void ProgramGL::reapplyUBOBindingsIfNeeded(const gl::Context *context)
{
// Re-apply UBO bindings to work around driver bugs.
const angle::FeaturesGL &features = GetImplAs<ContextGL>(context)->getFeaturesGL();
if (features.reapplyUBOBindingsAfterUsingBinaryProgram.enabled)
{
const auto &blocks = mState.getExecutable().getUniformBlocks();
for (size_t blockIndex : mState.getExecutable().getActiveUniformBlockBindings())
{
setUniformBlockBinding(static_cast<GLuint>(blockIndex), blocks[blockIndex].pod.binding);
}
}
}
void ProgramGL::setBinaryRetrievableHint(bool retrievable)
{
// glProgramParameteri isn't always available on ES backends.
if (mFunctions->programParameteri)
{
mFunctions->programParameteri(mProgramID, GL_PROGRAM_BINARY_RETRIEVABLE_HINT,
retrievable ? GL_TRUE : GL_FALSE);
}
}
void ProgramGL::setSeparable(bool separable)
{
mFunctions->programParameteri(mProgramID, GL_PROGRAM_SEPARABLE, separable ? GL_TRUE : GL_FALSE);
}
void ProgramGL::prepareForLink(const gl::ShaderMap<ShaderImpl *> &shaders)
{
for (gl::ShaderType shaderType : gl::AllShaderTypes())
{
mAttachedShaders[shaderType] = 0;
if (shaders[shaderType] != nullptr)
{
const ShaderGL *shaderGL = GetAs<ShaderGL>(shaders[shaderType]);
mAttachedShaders[shaderType] = shaderGL->getShaderID();
}
}
}
angle::Result ProgramGL::link(const gl::Context *context, std::shared_ptr<LinkTask> *linkTaskOut)
{
ANGLE_TRACE_EVENT0("gpu.angle", "ProgramGL::link");
*linkTaskOut = std::make_shared<LinkTaskGL>(this, mRenderer->hasNativeParallelCompile(),
mFunctions, context->getExtensions(), mProgramID);
return angle::Result::Continue;
}
void ProgramGL::linkJobImpl(const gl::Extensions &extensions)
{
ANGLE_TRACE_EVENT0("gpu.angle", "ProgramGL::linkJobImpl");
const gl::ProgramExecutable &executable = mState.getExecutable();
ProgramExecutableGL *executableGL = getExecutable();
if (mAttachedShaders[gl::ShaderType::Compute] != 0)
{
mFunctions->attachShader(mProgramID, mAttachedShaders[gl::ShaderType::Compute]);
}
else
{
// Set the transform feedback state
std::vector<std::string> transformFeedbackVaryingMappedNames;
const gl::ShaderType tfShaderType =
executable.hasLinkedShaderStage(gl::ShaderType::Geometry) ? gl::ShaderType::Geometry
: gl::ShaderType::Vertex;
const gl::SharedCompiledShaderState &tfShaderState = mState.getAttachedShader(tfShaderType);
for (const auto &tfVarying : mState.getTransformFeedbackVaryingNames())
{
std::string tfVaryingMappedName =
GetTransformFeedbackVaryingMappedName(tfShaderState, tfVarying);
transformFeedbackVaryingMappedNames.push_back(tfVaryingMappedName);
}
if (transformFeedbackVaryingMappedNames.empty())
{
// Only clear the transform feedback state if transform feedback varyings have already
// been set.
if (executableGL->mHasAppliedTransformFeedbackVaryings)
{
ASSERT(mFunctions->transformFeedbackVaryings);
mFunctions->transformFeedbackVaryings(mProgramID, 0, nullptr,
mState.getTransformFeedbackBufferMode());
executableGL->mHasAppliedTransformFeedbackVaryings = false;
}
}
else
{
ASSERT(mFunctions->transformFeedbackVaryings);
std::vector<const GLchar *> transformFeedbackVaryings;
for (const auto &varying : transformFeedbackVaryingMappedNames)
{
transformFeedbackVaryings.push_back(varying.c_str());
}
mFunctions->transformFeedbackVaryings(
mProgramID, static_cast<GLsizei>(transformFeedbackVaryingMappedNames.size()),
&transformFeedbackVaryings[0], mState.getTransformFeedbackBufferMode());
executableGL->mHasAppliedTransformFeedbackVaryings = true;
}
for (const gl::ShaderType shaderType : gl::kAllGraphicsShaderTypes)
{
if (mAttachedShaders[shaderType] != 0)
{
mFunctions->attachShader(mProgramID, mAttachedShaders[shaderType]);
}
}
// Bind attribute locations to match the GL layer.
for (const gl::ProgramInput &attribute : executable.getProgramInputs())
{
if (!attribute.isActive() || attribute.isBuiltIn())
{
continue;
}
mFunctions->bindAttribLocation(mProgramID, attribute.getLocation(),
attribute.mappedName.c_str());
}
// Bind the secondary fragment color outputs defined in EXT_blend_func_extended. We only use
// the API to bind fragment output locations in case EXT_blend_func_extended is enabled.
// Otherwise shader-assigned locations will work.
if (extensions.blendFuncExtendedEXT)
{
const gl::SharedCompiledShaderState &fragmentShader =
mState.getAttachedShader(gl::ShaderType::Fragment);
if (fragmentShader && fragmentShader->shaderVersion == 100 &&
mFunctions->standard == STANDARD_GL_DESKTOP)
{
const auto &shaderOutputs = fragmentShader->activeOutputVariables;
for (const auto &output : shaderOutputs)
{
// TODO(http://anglebug.com/1085) This could be cleaner if the transformed names
// would be set correctly in ShaderVariable::mappedName. This would require some
// refactoring in the translator. Adding a mapped name dictionary for builtins
// into the symbol table would be one fairly clean way to do it.
if (output.name == "gl_SecondaryFragColorEXT")
{
mFunctions->bindFragDataLocationIndexed(mProgramID, 0, 0,
"webgl_FragColor");
mFunctions->bindFragDataLocationIndexed(mProgramID, 0, 1,
"webgl_SecondaryFragColor");
}
else if (output.name == "gl_SecondaryFragDataEXT")
{
// Basically we should have a loop here going over the output
// array binding "webgl_FragData[i]" and "webgl_SecondaryFragData[i]" array
// indices to the correct color buffers and color indices.
// However I'm not sure if this construct is legal or not, neither ARB or
// EXT version of the spec mention this. They only mention that
// automatically assigned array locations for ESSL 3.00 output arrays need
// to have contiguous locations.
//
// In practice it seems that binding array members works on some drivers and
// fails on others. One option could be to modify the shader translator to
// expand the arrays into individual output variables instead of using an
// array.
//
// For now we're going to have a limitation of assuming that
// GL_MAX_DUAL_SOURCE_DRAW_BUFFERS is *always* 1 and then only bind the
// basename of the variable ignoring any indices. This appears to work
// uniformly.
ASSERT(output.isArray() && output.getOutermostArraySize() == 1);
mFunctions->bindFragDataLocationIndexed(mProgramID, 0, 0, "webgl_FragData");
mFunctions->bindFragDataLocationIndexed(mProgramID, 0, 1,
"webgl_SecondaryFragData");
}
}
}
else if (fragmentShader && fragmentShader->shaderVersion >= 300)
{
// ESSL 3.00 and up.
const auto &outputLocations = executable.getOutputLocations();
const auto &secondaryOutputLocations = executable.getSecondaryOutputLocations();
for (size_t outputLocationIndex = 0u; outputLocationIndex < outputLocations.size();
++outputLocationIndex)
{
const gl::VariableLocation &outputLocation =
outputLocations[outputLocationIndex];
if (outputLocation.arrayIndex == 0 && outputLocation.used() &&
!outputLocation.ignored)
{
const gl::ProgramOutput &outputVar =
executable.getOutputVariables()[outputLocation.index];
if (outputVar.pod.location == -1 || outputVar.pod.index == -1)
{
// We only need to assign the location and index via the API in case the
// variable doesn't have a shader-assigned location and index. If a
// variable doesn't have its location set in the shader it doesn't have
// the index set either.
ASSERT(outputVar.pod.index == -1);
mFunctions->bindFragDataLocationIndexed(
mProgramID, static_cast<int>(outputLocationIndex), 0,
outputVar.mappedName.c_str());
}
}
}
for (size_t outputLocationIndex = 0u;
outputLocationIndex < secondaryOutputLocations.size(); ++outputLocationIndex)
{
const gl::VariableLocation &outputLocation =
secondaryOutputLocations[outputLocationIndex];
if (outputLocation.arrayIndex == 0 && outputLocation.used() &&
!outputLocation.ignored)
{
const gl::ProgramOutput &outputVar =
executable.getOutputVariables()[outputLocation.index];
if (outputVar.pod.location == -1 || outputVar.pod.index == -1)
{
// We only need to assign the location and index via the API in case the
// variable doesn't have a shader-assigned location and index. If a
// variable doesn't have its location set in the shader it doesn't have
// the index set either.
ASSERT(outputVar.pod.index == -1);
mFunctions->bindFragDataLocationIndexed(
mProgramID, static_cast<int>(outputLocationIndex), 1,
outputVar.mappedName.c_str());
}
}
}
}
}
}
mFunctions->linkProgram(mProgramID);
}
angle::Result ProgramGL::postLinkJobImpl(const gl::ProgramLinkedResources &resources)
{
ANGLE_TRACE_EVENT0("gpu.angle", "ProgramGL::postLinkJobImpl");
if (mAttachedShaders[gl::ShaderType::Compute] != 0)
{
mFunctions->detachShader(mProgramID, mAttachedShaders[gl::ShaderType::Compute]);
}
else
{
for (const gl::ShaderType shaderType : gl::kAllGraphicsShaderTypes)
{
if (mAttachedShaders[shaderType] != 0)
{
mFunctions->detachShader(mProgramID, mAttachedShaders[shaderType]);
}
}
}
// Verify the link
if (!checkLinkStatus())
{
return angle::Result::Stop;
}
if (mFeatures.alwaysCallUseProgramAfterLink.enabled)
{
mStateManager->forceUseProgram(mProgramID);
}
linkResources(resources);
getExecutable()->postLink(mFunctions, mStateManager, mFeatures, mProgramID);
return angle::Result::Continue;
}
GLboolean ProgramGL::validate(const gl::Caps & /*caps*/)
{
// TODO(jmadill): implement validate
return true;
}
void ProgramGL::setUniformBlockBinding(GLuint uniformBlockIndex, GLuint uniformBlockBinding)
{
const gl::ProgramExecutable &executable = mState.getExecutable();
ProgramExecutableGL *executableGL = getExecutable();
// Lazy init
if (executableGL->mUniformBlockRealLocationMap.empty())
{
executableGL->mUniformBlockRealLocationMap.reserve(executable.getUniformBlocks().size());
for (const gl::InterfaceBlock &uniformBlock : executable.getUniformBlocks())
{
const std::string &mappedNameWithIndex = uniformBlock.mappedNameWithArrayIndex();
GLuint blockIndex =
mFunctions->getUniformBlockIndex(mProgramID, mappedNameWithIndex.c_str());
executableGL->mUniformBlockRealLocationMap.push_back(blockIndex);
}
}
GLuint realBlockIndex = executableGL->mUniformBlockRealLocationMap[uniformBlockIndex];
if (realBlockIndex != GL_INVALID_INDEX)
{
mFunctions->uniformBlockBinding(mProgramID, realBlockIndex, uniformBlockBinding);
}
}
bool ProgramGL::getUniformBlockSize(const std::string & /* blockName */,
const std::string &blockMappedName,
size_t *sizeOut) const
{
ASSERT(mProgramID != 0u);
GLuint blockIndex = mFunctions->getUniformBlockIndex(mProgramID, blockMappedName.c_str());
if (blockIndex == GL_INVALID_INDEX)
{
*sizeOut = 0;
return false;
}
GLint dataSize = 0;
mFunctions->getActiveUniformBlockiv(mProgramID, blockIndex, GL_UNIFORM_BLOCK_DATA_SIZE,
&dataSize);
*sizeOut = static_cast<size_t>(dataSize);
return true;
}
bool ProgramGL::getUniformBlockMemberInfo(const std::string & /* memberUniformName */,
const std::string &memberUniformMappedName,
sh::BlockMemberInfo *memberInfoOut) const
{
GLuint uniformIndex;
const GLchar *memberNameGLStr = memberUniformMappedName.c_str();
mFunctions->getUniformIndices(mProgramID, 1, &memberNameGLStr, &uniformIndex);
if (uniformIndex == GL_INVALID_INDEX)
{
*memberInfoOut = sh::kDefaultBlockMemberInfo;
return false;
}
mFunctions->getActiveUniformsiv(mProgramID, 1, &uniformIndex, GL_UNIFORM_OFFSET,
&memberInfoOut->offset);
mFunctions->getActiveUniformsiv(mProgramID, 1, &uniformIndex, GL_UNIFORM_ARRAY_STRIDE,
&memberInfoOut->arrayStride);
mFunctions->getActiveUniformsiv(mProgramID, 1, &uniformIndex, GL_UNIFORM_MATRIX_STRIDE,
&memberInfoOut->matrixStride);
// TODO(jmadill): possibly determine this at the gl::Program level.
GLint isRowMajorMatrix = 0;
mFunctions->getActiveUniformsiv(mProgramID, 1, &uniformIndex, GL_UNIFORM_IS_ROW_MAJOR,
&isRowMajorMatrix);
memberInfoOut->isRowMajorMatrix = gl::ConvertToBool(isRowMajorMatrix);
return true;
}
bool ProgramGL::getShaderStorageBlockMemberInfo(const std::string & /* memberName */,
const std::string &memberUniformMappedName,
sh::BlockMemberInfo *memberInfoOut) const
{
const GLchar *memberNameGLStr = memberUniformMappedName.c_str();
GLuint index =
mFunctions->getProgramResourceIndex(mProgramID, GL_BUFFER_VARIABLE, memberNameGLStr);
if (index == GL_INVALID_INDEX)
{
*memberInfoOut = sh::kDefaultBlockMemberInfo;
return false;
}
constexpr int kPropCount = 5;
std::array<GLenum, kPropCount> props = {
{GL_ARRAY_STRIDE, GL_IS_ROW_MAJOR, GL_MATRIX_STRIDE, GL_OFFSET, GL_TOP_LEVEL_ARRAY_STRIDE}};
std::array<GLint, kPropCount> params;
GLsizei length;
mFunctions->getProgramResourceiv(mProgramID, GL_BUFFER_VARIABLE, index, kPropCount,
props.data(), kPropCount, &length, params.data());
ASSERT(kPropCount == length);
memberInfoOut->arrayStride = params[0];
memberInfoOut->isRowMajorMatrix = params[1] != 0;
memberInfoOut->matrixStride = params[2];
memberInfoOut->offset = params[3];
memberInfoOut->topLevelArrayStride = params[4];
return true;
}
bool ProgramGL::getShaderStorageBlockSize(const std::string &name,
const std::string &mappedName,
size_t *sizeOut) const
{
const GLchar *nameGLStr = mappedName.c_str();
GLuint index =
mFunctions->getProgramResourceIndex(mProgramID, GL_SHADER_STORAGE_BLOCK, nameGLStr);
if (index == GL_INVALID_INDEX)
{
*sizeOut = 0;
return false;
}
GLenum prop = GL_BUFFER_DATA_SIZE;
GLsizei length = 0;
GLint dataSize = 0;
mFunctions->getProgramResourceiv(mProgramID, GL_SHADER_STORAGE_BLOCK, index, 1, &prop, 1,
&length, &dataSize);
*sizeOut = static_cast<size_t>(dataSize);
return true;
}
void ProgramGL::getAtomicCounterBufferSizeMap(std::map<int, unsigned int> *sizeMapOut) const
{
if (mFunctions->getProgramInterfaceiv == nullptr)
{
return;
}
int resourceCount = 0;
mFunctions->getProgramInterfaceiv(mProgramID, GL_ATOMIC_COUNTER_BUFFER, GL_ACTIVE_RESOURCES,
&resourceCount);
for (int index = 0; index < resourceCount; index++)
{
constexpr int kPropCount = 2;
std::array<GLenum, kPropCount> props = {{GL_BUFFER_BINDING, GL_BUFFER_DATA_SIZE}};
std::array<GLint, kPropCount> params;
GLsizei length;
mFunctions->getProgramResourceiv(mProgramID, GL_ATOMIC_COUNTER_BUFFER, index, kPropCount,
props.data(), kPropCount, &length, params.data());
ASSERT(kPropCount == length);
int bufferBinding = params[0];
unsigned int bufferDataSize = params[1];
sizeMapOut->insert(std::pair<int, unsigned int>(bufferBinding, bufferDataSize));
}
}
bool ProgramGL::checkLinkStatus()
{
GLint linkStatus = GL_FALSE;
mFunctions->getProgramiv(mProgramID, GL_LINK_STATUS, &linkStatus);
if (linkStatus == GL_FALSE)
{
// Linking or program binary loading failed, put the error into the info log.
GLint infoLogLength = 0;
mFunctions->getProgramiv(mProgramID, GL_INFO_LOG_LENGTH, &infoLogLength);
// Info log length includes the null terminator, so 1 means that the info log is an empty
// string.
if (infoLogLength > 1)
{
std::vector<char> buf(infoLogLength);
mFunctions->getProgramInfoLog(mProgramID, infoLogLength, nullptr, &buf[0]);
mState.getExecutable().getInfoLog() << buf.data();
WARN() << "Program link or binary loading failed: " << buf.data();
}
else
{
WARN() << "Program link or binary loading failed with no info log.";
}
// This may happen under normal circumstances if we're loading program binaries and the
// driver or hardware has changed.
ASSERT(mProgramID != 0);
return false;
}
return true;
}
void ProgramGL::markUnusedUniformLocations(std::vector<gl::VariableLocation> *uniformLocations,
std::vector<gl::SamplerBinding> *samplerBindings,
std::vector<gl::ImageBinding> *imageBindings)
{
const gl::ProgramExecutable &executable = mState.getExecutable();
const ProgramExecutableGL *executableGL = getExecutable();
GLint maxLocation = static_cast<GLint>(uniformLocations->size());
for (GLint location = 0; location < maxLocation; ++location)
{
if (executableGL->mUniformRealLocationMap[location] == -1)
{
auto &locationRef = (*uniformLocations)[location];
if (executable.isSamplerUniformIndex(locationRef.index))
{
GLuint samplerIndex = executable.getSamplerIndexFromUniformIndex(locationRef.index);
gl::SamplerBinding &samplerBinding = (*samplerBindings)[samplerIndex];
if (locationRef.arrayIndex <
static_cast<unsigned int>(samplerBinding.textureUnitsCount))
{
// Crop unused sampler bindings in the sampler array.
SetBitField(samplerBinding.textureUnitsCount, locationRef.arrayIndex);
}
}
else if (executable.isImageUniformIndex(locationRef.index))
{
GLuint imageIndex = executable.getImageIndexFromUniformIndex(locationRef.index);
gl::ImageBinding &imageBinding = (*imageBindings)[imageIndex];
if (locationRef.arrayIndex < imageBinding.boundImageUnits.size())
{
// Crop unused image bindings in the image array.
imageBinding.boundImageUnits.resize(locationRef.arrayIndex);
}
}
// If the location has been previously bound by a glBindUniformLocation call, it should
// be marked as ignored. Otherwise it's unused.
if (mState.getUniformLocationBindings().getBindingByLocation(location) != -1)
{
locationRef.markIgnored();
}
else
{
locationRef.markUnused();
}
}
}
}
void ProgramGL::linkResources(const gl::ProgramLinkedResources &resources)
{
// Gather interface block info.
auto getUniformBlockSize = [this](const std::string &name, const std::string &mappedName,
size_t *sizeOut) {
return this->getUniformBlockSize(name, mappedName, sizeOut);
};
auto getUniformBlockMemberInfo = [this](const std::string &name, const std::string &mappedName,
sh::BlockMemberInfo *infoOut) {
return this->getUniformBlockMemberInfo(name, mappedName, infoOut);
};
resources.uniformBlockLinker.linkBlocks(getUniformBlockSize, getUniformBlockMemberInfo);
auto getShaderStorageBlockSize = [this](const std::string &name, const std::string &mappedName,
size_t *sizeOut) {
return this->getShaderStorageBlockSize(name, mappedName, sizeOut);
};
auto getShaderStorageBlockMemberInfo = [this](const std::string &name,
const std::string &mappedName,
sh::BlockMemberInfo *infoOut) {
return this->getShaderStorageBlockMemberInfo(name, mappedName, infoOut);
};
resources.shaderStorageBlockLinker.linkBlocks(getShaderStorageBlockSize,
getShaderStorageBlockMemberInfo);
// Gather atomic counter buffer info.
std::map<int, unsigned int> sizeMap;
getAtomicCounterBufferSizeMap(&sizeMap);
resources.atomicCounterBufferLinker.link(sizeMap);
}
angle::Result ProgramGL::syncState(const gl::Context *context)
{
const gl::ProgramExecutable &executable = mState.getExecutable();
gl::ProgramExecutable::DirtyBits dirtyBits = executable.getAndResetDirtyBits();
for (size_t dirtyBit : dirtyBits)
{
ASSERT(dirtyBit <= gl::ProgramExecutable::DIRTY_BIT_UNIFORM_BLOCK_BINDING_MAX);
GLuint binding = static_cast<GLuint>(dirtyBit);
setUniformBlockBinding(binding, executable.getUniformBlockBinding(binding));
}
return angle::Result::Continue;
}
} // namespace rx