Hash :
89a69a03
Author :
Date :
2017-10-23T12:20:45
Generate performance warnings in HLSL translation Generate performance warnings for some code that undergoes heavy emulation when translated to HLSL: 1. Dynamic indexing of vectors and matrices. 2. Non-empty fall-through cases in switch/case. The warnings are generated only when code is translated to HLSL. Generating them in the parsing stage would add too much maintenance burden. Improves switch statement fall-through handling in cases where an empty fall-through case follows a non-empty one so that extra performance warnings are not generated. BUG=angleproject:1116 Change-Id: I7c85d78fe7c4f8e6042bda72ceaaf6e37dadfe6c Reviewed-on: https://chromium-review.googlesource.com/732986 Commit-Queue: Olli Etuaho <oetuaho@nvidia.com> Reviewed-by: Corentin Wallez <cwallez@chromium.org>
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161
//
// Copyright (c) 2002-2013 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.
//
#include "compiler/translator/TranslatorHLSL.h"
#include "compiler/translator/AddDefaultReturnStatements.h"
#include "compiler/translator/ArrayReturnValueToOutParameter.h"
#include "compiler/translator/BreakVariableAliasingInInnerLoops.h"
#include "compiler/translator/EmulatePrecision.h"
#include "compiler/translator/ExpandIntegerPowExpressions.h"
#include "compiler/translator/IntermNodePatternMatcher.h"
#include "compiler/translator/OutputHLSL.h"
#include "compiler/translator/RemoveDynamicIndexing.h"
#include "compiler/translator/RemoveNoOpCasesFromEndOfSwitchStatements.h"
#include "compiler/translator/RewriteElseBlocks.h"
#include "compiler/translator/RewriteTexelFetchOffset.h"
#include "compiler/translator/RewriteUnaryMinusOperatorInt.h"
#include "compiler/translator/SeparateArrayInitialization.h"
#include "compiler/translator/SeparateDeclarations.h"
#include "compiler/translator/SeparateExpressionsReturningArrays.h"
#include "compiler/translator/SimplifyLoopConditions.h"
#include "compiler/translator/SplitSequenceOperator.h"
#include "compiler/translator/UnfoldShortCircuitToIf.h"
#include "compiler/translator/WrapSwitchStatementsInBlocks.h"
namespace sh
{
TranslatorHLSL::TranslatorHLSL(sh::GLenum type, ShShaderSpec spec, ShShaderOutput output)
: TCompiler(type, spec, output)
{
}
void TranslatorHLSL::translate(TIntermBlock *root,
ShCompileOptions compileOptions,
PerformanceDiagnostics *perfDiagnostics)
{
const ShBuiltInResources &resources = getResources();
int numRenderTargets = resources.EXT_draw_buffers ? resources.MaxDrawBuffers : 1;
sh::AddDefaultReturnStatements(root);
// Note that SimplifyLoopConditions needs to be run before any other AST transformations that
// may need to generate new statements from loop conditions or loop expressions.
// Note that SeparateDeclarations has already been run in TCompiler::compileTreeImpl().
SimplifyLoopConditions(root,
IntermNodePatternMatcher::kExpressionReturningArray |
IntermNodePatternMatcher::kUnfoldedShortCircuitExpression |
IntermNodePatternMatcher::kDynamicIndexingOfVectorOrMatrixInLValue,
&getSymbolTable(), getShaderVersion());
SplitSequenceOperator(root,
IntermNodePatternMatcher::kExpressionReturningArray |
IntermNodePatternMatcher::kUnfoldedShortCircuitExpression |
IntermNodePatternMatcher::kDynamicIndexingOfVectorOrMatrixInLValue,
&getSymbolTable(), getShaderVersion());
// Note that SeparateDeclarations needs to be run before UnfoldShortCircuitToIf.
UnfoldShortCircuitToIf(root, &getSymbolTable());
SeparateExpressionsReturningArrays(root, &getSymbolTable());
// Note that SeparateDeclarations needs to be run before SeparateArrayInitialization.
SeparateArrayInitialization(root);
// HLSL doesn't support arrays as return values, we'll need to make functions that have an array
// as a return value to use an out parameter to transfer the array data instead.
ArrayReturnValueToOutParameter(root, &getSymbolTable());
if (!shouldRunLoopAndIndexingValidation(compileOptions))
{
// HLSL doesn't support dynamic indexing of vectors and matrices.
RemoveDynamicIndexing(root, &getSymbolTable(), getShaderVersion(), perfDiagnostics);
}
// Work around D3D9 bug that would manifest in vertex shaders with selection blocks which
// use a vertex attribute as a condition, and some related computation in the else block.
if (getOutputType() == SH_HLSL_3_0_OUTPUT && getShaderType() == GL_VERTEX_SHADER)
{
sh::RewriteElseBlocks(root, &getSymbolTable());
}
// Work around an HLSL compiler frontend aliasing optimization bug.
// TODO(cwallez) The date is 2016-08-25, Microsoft said the bug would be fixed
// in the next release of d3dcompiler.dll, it would be nice to detect the DLL
// version and only apply the workaround if it is too old.
sh::BreakVariableAliasingInInnerLoops(root);
// WrapSwitchStatementsInBlocks should be called after any AST transformations that might
// introduce variable declarations inside the main scope of any switch statement.
if (WrapSwitchStatementsInBlocks(root))
{
// The WrapSwitchStatementsInBlocks step might introduce new no-op cases to the end of
// switch statements, so make sure to clean up the AST.
RemoveNoOpCasesFromEndOfSwitchStatements(root, &getSymbolTable());
}
bool precisionEmulation =
getResources().WEBGL_debug_shader_precision && getPragma().debugShaderPrecision;
if (precisionEmulation)
{
EmulatePrecision emulatePrecision(&getSymbolTable(), getShaderVersion());
root->traverse(&emulatePrecision);
emulatePrecision.updateTree();
emulatePrecision.writeEmulationHelpers(getInfoSink().obj, getShaderVersion(),
getOutputType());
}
if ((compileOptions & SH_EXPAND_SELECT_HLSL_INTEGER_POW_EXPRESSIONS) != 0)
{
sh::ExpandIntegerPowExpressions(root, &getSymbolTable());
}
if ((compileOptions & SH_REWRITE_TEXELFETCHOFFSET_TO_TEXELFETCH) != 0)
{
sh::RewriteTexelFetchOffset(root, getSymbolTable(), getShaderVersion());
}
if (((compileOptions & SH_REWRITE_INTEGER_UNARY_MINUS_OPERATOR) != 0) &&
getShaderType() == GL_VERTEX_SHADER)
{
sh::RewriteUnaryMinusOperatorInt(root);
}
sh::OutputHLSL outputHLSL(getShaderType(), getShaderVersion(), getExtensionBehavior(),
getSourcePath(), getOutputType(), numRenderTargets, getUniforms(),
compileOptions, &getSymbolTable(), perfDiagnostics);
outputHLSL.output(root, getInfoSink().obj);
mUniformBlockRegisterMap = outputHLSL.getUniformBlockRegisterMap();
mUniformRegisterMap = outputHLSL.getUniformRegisterMap();
}
bool TranslatorHLSL::shouldFlattenPragmaStdglInvariantAll()
{
// Not necessary when translating to HLSL.
return false;
}
bool TranslatorHLSL::hasUniformBlock(const std::string &uniformBlockName) const
{
return (mUniformBlockRegisterMap.count(uniformBlockName) > 0);
}
unsigned int TranslatorHLSL::getUniformBlockRegister(const std::string &uniformBlockName) const
{
ASSERT(hasUniformBlock(uniformBlockName));
return mUniformBlockRegisterMap.find(uniformBlockName)->second;
}
const std::map<std::string, unsigned int> *TranslatorHLSL::getUniformRegisterMap() const
{
return &mUniformRegisterMap;
}
} // namespace sh