experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Updated for changes in LLVM 7a45aeacf3a2.

Browse Source 
- The deprecated CreateCall(Value*, ...) IRBuilder API has been removed.
- Renamed applyPatternsGreedily to applyPatternsAndFoldGreedily in MLIR.
- Update MLIR users after adding support for optional operands/results to ODS (upstream aba1acc89c653b2cc08cccfb754ff16994a05332)
- Other updates to BUILD files for upstream changes.

PiperOrigin-RevId: 306177884
Change-Id: Idae1009ba89caf296758748ab7aa57815d946a0c
This commit is contained in:
A. Unique TensorFlower 2020-04-12 22:11:54 -07:00 committed by TensorFlower Gardener
parent 700ff4897f
commit f5ee735428
24 changed files with 63 additions and 29 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
tensorflow/compiler/mlir/lite/converter_gen.cc
View File

@ -496,7 +496,7 @@ static bool RuntimeVerifierWriterMain(raw_ostream &os, RecordKeeper &records) {
auto &value = op.getOperand(i);
// Skip from from first variadic operands for now. Else getOperand index
// used below doesn't match.
if (value.isVariadic()) break;
if (value.isVariableLength()) break;
if (!value.name.empty())
verify_ctx.addSubst(value.name, formatv("op->getOperand({0})", i));
}
@ -504,7 +504,7 @@ static bool RuntimeVerifierWriterMain(raw_ostream &os, RecordKeeper &records) {
auto &value = op.getResult(i);
// Skip from from first variadic results for now. Else getResult index
// used below doesn't match.
if (value.isVariadic()) break;
if (value.isVariableLength()) break;
if (!value.name.empty())
verify_ctx.addSubst(value.name, formatv("op->getResult({0})", i));
}

2
tensorflow/compiler/mlir/lite/quantization/tensorflow/tf_to_quant.cc
View File

@ -146,7 +146,7 @@ void LegalizeTFToQuant::runOnFunction() {
auto func = getFunction();
auto *ctx = func.getContext();
patterns.insert<PreparePerTensorFakeQuant, PreparePerChannelFakeQuant>(ctx);
applyPatternsGreedily(func, patterns);
applyPatternsAndFoldGreedily(func, patterns);
}
} // namespace

2
tensorflow/compiler/mlir/lite/transforms/dilated_conv.cc
View File

@ -30,7 +30,7 @@ void IdentifyDilatedConvPass::runOnFunction() {
patterns.insert<ConvertTFDilatedConvOp<TF::Conv2DOp>,
ConvertTFDilatedConvOp<TF::DepthwiseConv2dNativeOp>>(
&getContext());
applyPatternsGreedily(func, patterns);
applyPatternsAndFoldGreedily(func, patterns);
}
} // namespace

4
tensorflow/compiler/mlir/lite/transforms/optimize.cc
View File

@ -711,7 +711,7 @@ void Optimize::runOnFunction() {
TFL::populateWithGenerated(ctx, &patterns);
patterns.insert<FuseFullyConnectedAndAdd, FuseFullyConnectedAndRelu,
FuseFullyConnectedAndMul>(ctx);
applyPatternsGreedily(func, patterns);
applyPatternsAndFoldGreedily(func, patterns);
// Fuse the binary ops with the following ops.
patterns.insert<
@ -719,7 +719,7 @@ void Optimize::runOnFunction() {
FuseBinaryOpToFollowingFullyConnected, FuseConv2DAndMulWithQDQs,
FuseDepthwiseConv2DAndMulWithQDQs, ConvertTrivialTransposeOpToReshapeOp>(
ctx);
applyPatternsGreedily(func, patterns);
applyPatternsAndFoldGreedily(func, patterns);
}
} // namespace

2
tensorflow/compiler/mlir/lite/transforms/optimize_functional_ops.cc
View File

@ -187,7 +187,7 @@ void OptimizeFunctionalOpsPass::runOnOperation() {
patterns.insert<FoldIfOp>(&getContext(), &inlined_funcs);
ModuleOp module = getOperation();
applyPatternsGreedily(module, patterns);
applyPatternsAndFoldGreedily(module, patterns);
// Erase inlined functions that don't have any references.
//

2
tensorflow/compiler/mlir/lite/transforms/post_quantize.cc
View File

@ -125,7 +125,7 @@ void PostQuantizePass::runOnFunction() {
auto func = getFunction();
auto* ctx = func.getContext();
TFL::populateWithGenerated(ctx, &patterns);
applyPatternsGreedily(func, patterns);
applyPatternsAndFoldGreedily(func, patterns);
if (!emit_quant_adaptor_ops_) {
RemoveQuantizationAdaptorOps(getFunction());

2
tensorflow/compiler/mlir/lite/transforms/prepare_quantize.cc
View File

@ -267,7 +267,7 @@ void PrepareQuantizePass::runOnFunction() {
// Currently, only activation stats are imported, so narrow_range = false.
patterns.insert<PrepareQuantStats>(8, false, false, ctx);
}
applyPatternsGreedily(func, patterns);
applyPatternsAndFoldGreedily(func, patterns);
SanityCheckAndAdjustment(func);

8
tensorflow/compiler/mlir/lite/transforms/prepare_tf.cc
View File

@ -619,8 +619,8 @@ void PrepareTFPass::runOnFunction() {
// This pattern was intented to uses TFL QDQs to preserve the quantization
// parameters from the TF Quant ops, thus this pattern should run with the
// first `applyPatternsGreedily` method, which would otherwise removes the
// TF FakeQuant ops by the constant folding.
// first `applyPatternsAndFoldGreedily` method, which would otherwise removes
// the TF FakeQuant ops by the constant folding.
patterns.insert<PreparePerTensorFakeQuant, PreparePerChannelFakeQuant>(ctx);
// This pattern will try to identify and optimize for dilated convolution.
@ -634,7 +634,7 @@ void PrepareTFPass::runOnFunction() {
// This will allow optimizing any TF_Mul->TF_Conv in the graph
// and any expanded from FusedBatchNorm. We need to do this
// before converting TF_Conv to TFL_Conv
applyPatternsGreedily(func, patterns);
applyPatternsAndFoldGreedily(func, patterns);
// Load the generated pattern again, so new quantization pass-through
// will be applied.
@ -646,7 +646,7 @@ void PrepareTFPass::runOnFunction() {
}
patterns.insert<TF::ConvertTFEinsumOp, ConvertTFConv2D,
ConvertTFDepthwiseConv2dNative, ConvertTFStridedSlice>(ctx);
applyPatternsGreedily(func, patterns);
applyPatternsAndFoldGreedily(func, patterns);
}
} // namespace

2
tensorflow/compiler/mlir/lite/transforms/quantize.cc
View File

@ -88,7 +88,7 @@ void QuantizePass::runOnFunction() {
TFL::populateWithGenerated(ctx, &patterns);
patterns.insert<TFLFullQuantization>(
ctx, enable_numeric_verify, error_tolerance, enable_single_layer_verify);
applyPatternsGreedily(func, patterns);
applyPatternsAndFoldGreedily(func, patterns);
}
} // namespace

2
tensorflow/compiler/mlir/tensorflow/transforms/batchmatmul_to_einsum.cc
View File

@ -55,7 +55,7 @@ void BatchMatMulToEinsumPass::runOnFunction() {
patterns.insert<ConvertTFBatchMatMulToEinsumOp<TF::BatchMatMulOp>,
ConvertTFBatchMatMulToEinsumOp<TF::BatchMatMulV2Op>>(
&getContext());
applyPatternsGreedily(func, patterns);
applyPatternsAndFoldGreedily(func, patterns);
}
} // namespace

2
tensorflow/compiler/mlir/tensorflow/transforms/decompose_resource_ops_pass.cc
View File

@ -45,7 +45,7 @@ struct DecomposeResourceOps
OwningRewritePatternList patterns;
mlir::TF::PopulateDecomposeResourceOpsPatterns(&getContext(), &patterns);
applyPatternsGreedily(getFunction(), patterns);
applyPatternsAndFoldGreedily(getFunction(), patterns);
}
};

2
tensorflow/compiler/mlir/tensorflow/transforms/einsum.cc
View File

@ -364,7 +364,7 @@ void TransformEinsumPass::runOnFunction() {
auto func = getFunction();
patterns.insert<ConvertTFEinsumOp>(&getContext());
applyPatternsGreedily(func, patterns);
applyPatternsAndFoldGreedily(func, patterns);
}
static PassRegistration<TransformEinsumPass> pass(

2
tensorflow/compiler/mlir/tensorflow/transforms/gpu_fusion.cc
View File

@ -118,7 +118,7 @@ void GpuOpFusionPass::runOnFunction() {
FuncOp func = getFunction();
OwningRewritePatternList patterns;
patterns.insert<ReluToFusedBatchNorm>(&getContext());
applyPatternsGreedily(func, patterns);
applyPatternsAndFoldGreedily(func, patterns);
}
} // namespace

2
tensorflow/compiler/mlir/tensorflow/transforms/lower_tf_pass.cc
View File

@ -29,7 +29,7 @@ struct LowerTF : public PassWrapper<LowerTF, FunctionPass> {
OwningRewritePatternList patterns;
mlir::TF::PopulateLoweringTFPatterns(&getContext(), &patterns);
applyPatternsGreedily(getFunction(), patterns);
applyPatternsAndFoldGreedily(getFunction(), patterns);
}
};

2
tensorflow/compiler/mlir/tensorflow/transforms/optimize.cc
View File

@ -38,7 +38,7 @@ struct TFOptimizePass : public PassWrapper<TFOptimizePass, FunctionPass> {
OwningRewritePatternList patterns;
auto func = getFunction();
populateWithGenerated(&getContext(), &patterns);
applyPatternsGreedily(func, patterns);
applyPatternsAndFoldGreedily(func, patterns);
}
};

2
tensorflow/compiler/mlir/tensorflow/transforms/unroll_batch_matmul.cc
View File

@ -55,7 +55,7 @@ void UnrollBatchMatMulPass::runOnFunction() {
patterns.insert<ConvertTFBatchMatMulOp<TF::BatchMatMulOp>,
ConvertTFBatchMatMulOp<TF::BatchMatMulV2Op>>(&getContext());
applyPatternsGreedily(func, patterns);
applyPatternsAndFoldGreedily(func, patterns);
}
} // namespace

4
tensorflow/compiler/mlir/xla/operator_writer_gen.cc
View File

@ -81,7 +81,7 @@ static void BuildOperator(const Operator& op, raw_ostream* output) {
// Emit an argument for an operand.
if (auto* operand_cst = arg.dyn_cast<NamedTypeConstraint*>()) {
// Handle a non-variadic operand.
if (!operand_cst->isVariadic()) {
if (!operand_cst->isVariableLength()) {
os << " auto xla_arg_" << index
<< " = value_map[*xla_op.getODSOperands(" << operand_number++
<< ").begin()];\n";
@ -108,7 +108,7 @@ static void BuildOperator(const Operator& op, raw_ostream* output) {
// If all operands are variadic, then pass the builder explicitly to xla
// client API call
if (op.getNumOperands() == op.getNumVariadicOperands()) {
if (op.getNumOperands() == op.getNumVariableLengthOperands()) {
os << "lowering_context.builder";
if (op.getNumArgs() != 0) os << ", ";
}

2
tensorflow/compiler/mlir/xla/transforms/legalize_to_standard.cc
View File

@ -198,7 +198,7 @@ void PopulateXlaToStdPatterns(OwningRewritePatternList *patterns,
void LegalizeToStandard::runOnFunction() {
OwningRewritePatternList patterns;
mlir::xla_hlo::PopulateXlaToStdPatterns(&patterns, &getContext());
applyPatternsGreedily(getFunction(), patterns);
applyPatternsAndFoldGreedily(getFunction(), patterns);
}
static PassRegistration<LegalizeToStandard> legalize_pass(

2
tensorflow/compiler/mlir/xla/transforms/lhlo_legalize_to_affine.cc
View File

@ -87,7 +87,7 @@ struct LhloLegalizeToAffine
OwningRewritePatternList patterns;
auto func = getFunction();
populateLHLOToAffineConversionPattern(func.getContext(), &patterns);
applyPatternsGreedily(func, patterns);
applyPatternsAndFoldGreedily(func, patterns);
}
};

2
tensorflow/compiler/mlir/xla/transforms/lower_complex.cc
View File

@ -71,7 +71,7 @@ void LowerComplex::runOnFunction() {
OwningRewritePatternList patterns;
mlir::xla::PopulateComplexLoweringPatterns(&getContext(), &patterns);
applyPatternsGreedily(getFunction(), patterns);
applyPatternsAndFoldGreedily(getFunction(), patterns);
}
static PassRegistration<LowerComplex> pass(

2
tensorflow/compiler/mlir/xla/transforms/lower_general_dot.cc
View File

@ -178,7 +178,7 @@ struct LegalizeGeneralDot
OwningRewritePatternList patterns;
mlir::xla_hlo::PopulateGeneralDotOpLoweringPatterns(&patterns,
&getContext());
applyPatternsGreedily(getFunction(), patterns);
applyPatternsAndFoldGreedily(getFunction(), patterns);
}
};

2
tensorflow/compiler/mlir/xla/transforms/unfuse_batch_norm_pass.cc
View File

@ -32,7 +32,7 @@ struct TestUnfuseBatchNormPass
void runOnOperation() override {
OwningRewritePatternList patterns;
PopulateUnfuseBatchNormPatterns(&getContext(), &patterns);
applyPatternsGreedily(getOperation(), patterns);
applyPatternsAndFoldGreedily(getOperation(), patterns);
}
};

22
tensorflow/compiler/xla/service/llvm_ir/ir_builder_mixin.h
View File

@ -79,11 +79,31 @@ class IrBuilderMixin {
return mixin_builder()->CreateBr(std::forward<Args>(args)...);
}
llvm::CallInst* Call(llvm::FunctionCallee func_callee,
llvm::ArrayRef<llvm::Value*> args = llvm::None,
const llvm::Twine& name = "",
llvm::MDNode* fp_math_tag = nullptr) {
return mixin_builder()->CreateCall(func_callee, args, name, fp_math_tag);
}
llvm::CallInst* Call(llvm::FunctionType* func_type, llvm::Value* callee,
llvm::ArrayRef<llvm::Value*> args = llvm::None,
const llvm::Twine& name = "",
llvm::MDNode* fp_math_tag = nullptr) {
return mixin_builder()->CreateCall(func_type, callee, args, name,
fp_math_tag);
}
// DEPRECATED. LLVM is removing getPointerElementType, so calls to this must
// be transitioned to one of the other overloads.
llvm::CallInst* Call(llvm::Value* callee,
llvm::ArrayRef<llvm::Value*> args = llvm::None,
const llvm::Twine& name = "",
llvm::MDNode* fp_math_tag = nullptr) {
return mixin_builder()->CreateCall(callee, args, name, fp_math_tag);
return mixin_builder()->CreateCall(
llvm::cast<llvm::FunctionType>(
callee->getType()->getPointerElementType()),
callee, args, name, fp_math_tag);
}
template <class... Args>

14
third_party/mlir/BUILD vendored
View File

@ -2666,6 +2666,20 @@ cc_binary(
],
)
cc_binary(
name = "mlir-linalg-ods-gen",
srcs = glob([
"tools/mlir-linalg-ods-gen/mlir-linalg-ods-gen.cpp",
]),
deps = [
":IR",
":Support",
"@llvm-project//llvm:config",
"@llvm-project//llvm:support",
"@llvm-project//llvm:tablegen",
],
)
## OpenMP dialect
gentbl(
name = "OpenMPOpsIncGen",