This is an archive of the discontinued LLVM Phabricator instance.

Differential D75264

Flip if condition of SimplifyRedundantTranspose.
ClosedPublic

Authored by matthiaskramm on Feb 27 2020, 8:32 AM.

Details

Reviewers
rriddle
ftynse
Commits
rGda0257563f5f: [mlir][Tutorial] Fix comment position in SimplifyRedundantTranspose.
Summary

This is a cosmetic change to make the "bingo" comment be in the
right place.

Diff Detail

Event Timeline

matthiaskramm created this revision.Feb 27 2020, 8:32 AM
Herald added a project: Restricted Project. · View Herald TranscriptFeb 27 2020, 8:32 AM
Herald added subscribers: llvm-commits, Joonsoo, liufengdb and 9 others. · View Herald Transcript
ftynse added a subscriber: ftynse.Feb 27 2020, 8:54 AM

LLVM encourages the inverse https://llvm.org/docs/CodingStandards.html#use-early-exits-and-continue-to-simplify-code of your change. Maybe rather move the comment?

Harbormaster failed remote builds in B47437: Diff 246961!Feb 27 2020, 8:55 AM
matthiaskramm updated this revision to Diff 246979.Feb 27 2020, 9:12 AM

Ran clang-format.

matthiaskramm updated this revision to Diff 246983.Feb 27 2020, 9:21 AM

Undo, adjust comments.

matthiaskramm updated this revision to Diff 246985.Feb 27 2020, 9:24 AM

Adjust commit message.

matthiaskramm added a comment.Feb 27 2020, 9:24 AM

Good point. Adjusted the comments instead.

rriddle accepted this revision.Feb 27 2020, 9:43 AM

Thanks Matthias!

This revision is now accepted and ready to land.Feb 27 2020, 9:43 AM
Harbormaster failed remote builds in B47447: Diff 246979!Feb 27 2020, 9:48 AM
Harbormaster failed remote builds in B47452: Diff 246985!Feb 27 2020, 10:14 AM
ftynse accepted this revision.Feb 27 2020, 10:16 AM
Harbormaster failed remote builds in B47451: Diff 246983!Feb 27 2020, 10:37 AM
Closed by commit rGda0257563f5f: [mlir][Tutorial] Fix comment position in SimplifyRedundantTranspose. (authored by matthiaskramm, committed by rriddle). · Explain WhyFeb 27 2020, 5:59 PM
This revision was automatically updated to reflect the committed changes.

Revision Contents

PathSize
mlir/
docs/
Tutorials/
Toy/
5 lines
examples/
toy/
Ch3/
mlir/
4 lines
Ch4/
mlir/
4 lines
Ch5/
mlir/
4 lines
Ch6/
mlir/
4 lines
Ch7/
mlir/
4 lines

Diff 247153

mlir/docs/Tutorials/Toy/Ch-3.md

Show First 20 Lines • Show All 87 Lines • ▼ Show 20 Linesstruct SimplifyRedundantTranspose : public mlir::OpRewritePattern<TransposeOp> {
/// to interact with it to perform any changes to the IR from here. /// to interact with it to perform any changes to the IR from here.
mlir::PatternMatchResult mlir::PatternMatchResult
matchAndRewrite(TransposeOp op, matchAndRewrite(TransposeOp op,
mlir::PatternRewriter &rewriter) const override { mlir::PatternRewriter &rewriter) const override {
// Look through the input of the current transpose. // Look through the input of the current transpose.
mlir::Value transposeInput = op.getOperand(); mlir::Value transposeInput = op.getOperand();
TransposeOp transposeInputOp = TransposeOp transposeInputOp =
llvm::dyn_cast_or_null<TransposeOp>(transposeInput.getDefiningOp()); llvm::dyn_cast_or_null<TransposeOp>(transposeInput.getDefiningOp());
// If the input is defined by another Transpose, bingo!
// Input defined by another transpose? If not, no match.
if (!transposeInputOp) if (!transposeInputOp)
return matchFailure(); return matchFailure();
// Use the rewriter to perform the replacement // Otherwise, we have a redundant transpose. Use the rewriter.
rewriter.replaceOp(op, {transposeInputOp.getOperand()}, {transposeInputOp}); rewriter.replaceOp(op, {transposeInputOp.getOperand()}, {transposeInputOp});
return matchSuccess(); return matchSuccess();
} }
}; };
``` ```
The implementation of this rewriter is in `ToyCombine.cpp`. The The implementation of this rewriter is in `ToyCombine.cpp`. The
[canonicalization pass](../../Canonicalization.md) applies transformations [canonicalization pass](../../Canonicalization.md) applies transformations
▲ Show 20 Lines • Show All 154 Lines • Show Last 20 Lines

mlir/examples/toy/Ch3/mlir/ToyCombine.cpp

Show All 37 Linesstruct SimplifyRedundantTranspose : public mlir::OpRewritePattern<TransposeOp> {
mlir::PatternMatchResult mlir::PatternMatchResult
matchAndRewrite(TransposeOp op, matchAndRewrite(TransposeOp op,
mlir::PatternRewriter &rewriter) const override { mlir::PatternRewriter &rewriter) const override {
// Look through the input of the current transpose. // Look through the input of the current transpose.
mlir::Value transposeInput = op.getOperand(); mlir::Value transposeInput = op.getOperand();
TransposeOp transposeInputOp = TransposeOp transposeInputOp =
llvm::dyn_cast_or_null<TransposeOp>(transposeInput.getDefiningOp()); llvm::dyn_cast_or_null<TransposeOp>(transposeInput.getDefiningOp());
// If the input is defined by another Transpose, bingo! // Input defined by another transpose? If not, no match.
if (!transposeInputOp) if (!transposeInputOp)
return matchFailure(); return matchFailure();
// Use the rewriter to perform the replacement. // Otherwise, we have a redundant transpose. Use the rewriter.
rewriter.replaceOp(op, {transposeInputOp.getOperand()}); rewriter.replaceOp(op, {transposeInputOp.getOperand()});
return matchSuccess(); return matchSuccess();
} }
}; };
/// Register our patterns as "canonicalization" patterns on the TransposeOp so /// Register our patterns as "canonicalization" patterns on the TransposeOp so
/// that they can be picked up by the Canonicalization framework. /// that they can be picked up by the Canonicalization framework.
void TransposeOp::getCanonicalizationPatterns(OwningRewritePatternList &results, void TransposeOp::getCanonicalizationPatterns(OwningRewritePatternList &results,
Show All 11 Lines

mlir/examples/toy/Ch4/mlir/ToyCombine.cpp

Show First 20 Lines • Show All 42 Lines • ▼ Show 20 Linesstruct SimplifyRedundantTranspose : public mlir::OpRewritePattern<TransposeOp> {
mlir::PatternMatchResult mlir::PatternMatchResult
matchAndRewrite(TransposeOp op, matchAndRewrite(TransposeOp op,
mlir::PatternRewriter &rewriter) const override { mlir::PatternRewriter &rewriter) const override {
// Look through the input of the current transpose. // Look through the input of the current transpose.
mlir::Value transposeInput = op.getOperand(); mlir::Value transposeInput = op.getOperand();
TransposeOp transposeInputOp = TransposeOp transposeInputOp =
llvm::dyn_cast_or_null<TransposeOp>(transposeInput.getDefiningOp()); llvm::dyn_cast_or_null<TransposeOp>(transposeInput.getDefiningOp());
// If the input is defined by another Transpose, bingo! // Input defined by another transpose? If not, no match.
if (!transposeInputOp) if (!transposeInputOp)
return matchFailure(); return matchFailure();
// Use the rewriter to perform the replacement. // Otherwise, we have a redundant transpose. Use the rewriter.
rewriter.replaceOp(op, {transposeInputOp.getOperand()}); rewriter.replaceOp(op, {transposeInputOp.getOperand()});
return matchSuccess(); return matchSuccess();
} }
}; };
/// Register our patterns as "canonicalization" patterns on the TransposeOp so /// Register our patterns as "canonicalization" patterns on the TransposeOp so
/// that they can be picked up by the Canonicalization framework. /// that they can be picked up by the Canonicalization framework.
void TransposeOp::getCanonicalizationPatterns(OwningRewritePatternList &results, void TransposeOp::getCanonicalizationPatterns(OwningRewritePatternList &results,
Show All 11 Lines

mlir/examples/toy/Ch5/mlir/ToyCombine.cpp

Show First 20 Lines • Show All 42 Lines • ▼ Show 20 Linesstruct SimplifyRedundantTranspose : public mlir::OpRewritePattern<TransposeOp> {
mlir::PatternMatchResult mlir::PatternMatchResult
matchAndRewrite(TransposeOp op, matchAndRewrite(TransposeOp op,
mlir::PatternRewriter &rewriter) const override { mlir::PatternRewriter &rewriter) const override {
// Look through the input of the current transpose. // Look through the input of the current transpose.
mlir::Value transposeInput = op.getOperand(); mlir::Value transposeInput = op.getOperand();
TransposeOp transposeInputOp = TransposeOp transposeInputOp =
llvm::dyn_cast_or_null<TransposeOp>(transposeInput.getDefiningOp()); llvm::dyn_cast_or_null<TransposeOp>(transposeInput.getDefiningOp());
// If the input is defined by another Transpose, bingo! // Input defined by another transpose? If not, no match.
if (!transposeInputOp) if (!transposeInputOp)
return matchFailure(); return matchFailure();
// Use the rewriter to perform the replacement. // Otherwise, we have a redundant transpose. Use the rewriter.
rewriter.replaceOp(op, {transposeInputOp.getOperand()}); rewriter.replaceOp(op, {transposeInputOp.getOperand()});
return matchSuccess(); return matchSuccess();
} }
}; };
/// Register our patterns as "canonicalization" patterns on the TransposeOp so /// Register our patterns as "canonicalization" patterns on the TransposeOp so
/// that they can be picked up by the Canonicalization framework. /// that they can be picked up by the Canonicalization framework.
void TransposeOp::getCanonicalizationPatterns(OwningRewritePatternList &results, void TransposeOp::getCanonicalizationPatterns(OwningRewritePatternList &results,
Show All 11 Lines

mlir/examples/toy/Ch6/mlir/ToyCombine.cpp

Show First 20 Lines • Show All 42 Lines • ▼ Show 20 Linesstruct SimplifyRedundantTranspose : public mlir::OpRewritePattern<TransposeOp> {
mlir::PatternMatchResult mlir::PatternMatchResult
matchAndRewrite(TransposeOp op, matchAndRewrite(TransposeOp op,
mlir::PatternRewriter &rewriter) const override { mlir::PatternRewriter &rewriter) const override {
// Look through the input of the current transpose. // Look through the input of the current transpose.
mlir::Value transposeInput = op.getOperand(); mlir::Value transposeInput = op.getOperand();
TransposeOp transposeInputOp = TransposeOp transposeInputOp =
llvm::dyn_cast_or_null<TransposeOp>(transposeInput.getDefiningOp()); llvm::dyn_cast_or_null<TransposeOp>(transposeInput.getDefiningOp());
// If the input is defined by another Transpose, bingo! // Input defined by another transpose? If not, no match.
if (!transposeInputOp) if (!transposeInputOp)
return matchFailure(); return matchFailure();
// Use the rewriter to perform the replacement. // Otherwise, we have a redundant transpose. Use the rewriter.
rewriter.replaceOp(op, {transposeInputOp.getOperand()}); rewriter.replaceOp(op, {transposeInputOp.getOperand()});
return matchSuccess(); return matchSuccess();
} }
}; };
/// Register our patterns as "canonicalization" patterns on the TransposeOp so /// Register our patterns as "canonicalization" patterns on the TransposeOp so
/// that they can be picked up by the Canonicalization framework. /// that they can be picked up by the Canonicalization framework.
void TransposeOp::getCanonicalizationPatterns(OwningRewritePatternList &results, void TransposeOp::getCanonicalizationPatterns(OwningRewritePatternList &results,
Show All 11 Lines

mlir/examples/toy/Ch7/mlir/ToyCombine.cpp

Show First 20 Lines • Show All 60 Lines • ▼ Show 20 Linesstruct SimplifyRedundantTranspose : public mlir::OpRewritePattern<TransposeOp> {
mlir::PatternMatchResult mlir::PatternMatchResult
matchAndRewrite(TransposeOp op, matchAndRewrite(TransposeOp op,
mlir::PatternRewriter &rewriter) const override { mlir::PatternRewriter &rewriter) const override {
// Look through the input of the current transpose. // Look through the input of the current transpose.
mlir::Value transposeInput = op.getOperand(); mlir::Value transposeInput = op.getOperand();
TransposeOp transposeInputOp = TransposeOp transposeInputOp =
llvm::dyn_cast_or_null<TransposeOp>(transposeInput.getDefiningOp()); llvm::dyn_cast_or_null<TransposeOp>(transposeInput.getDefiningOp());
// If the input is defined by another Transpose, bingo! // Input defined by another transpose? If not, no match.
if (!transposeInputOp) if (!transposeInputOp)
return matchFailure(); return matchFailure();
// Use the rewriter to perform the replacement. // Otherwise, we have a redundant transpose. Use the rewriter.
rewriter.replaceOp(op, {transposeInputOp.getOperand()}); rewriter.replaceOp(op, {transposeInputOp.getOperand()});
return matchSuccess(); return matchSuccess();
} }
}; };
/// Register our patterns as "canonicalization" patterns on the TransposeOp so /// Register our patterns as "canonicalization" patterns on the TransposeOp so
/// that they can be picked up by the Canonicalization framework. /// that they can be picked up by the Canonicalization framework.
void TransposeOp::getCanonicalizationPatterns(OwningRewritePatternList &results, void TransposeOp::getCanonicalizationPatterns(OwningRewritePatternList &results,
Show All 11 Lines