diff options
author | thomasraoux <thomasraoux@google.com> | 2021-05-03 10:04:12 -0700 |
---|---|---|
committer | thomasraoux <thomasraoux@google.com> | 2021-05-03 10:23:45 -0700 |
commit | be8e2801a4f305213fef99ae3f7e80cfeb2a7c08 (patch) | |
tree | 6e1a6b401ad26b960e5e2a45018dd9f156caec62 /mlir/lib/Conversion/VectorToLLVM | |
parent | 9779b664b6a8d8a047d52061b23a4a3737446e45 (diff) |
[mlir][vector][NFC] split TransposeOp lowerning out of contractLowering
Move TransposeOp lowering in its own populate function as in some cases
it is better to keep it during ContractOp lowering to better
canonicalize it rather than emiting scalar insert/extract.
Differential Revision: https://reviews.llvm.org/D101647
Diffstat (limited to 'mlir/lib/Conversion/VectorToLLVM')
-rw-r--r-- | mlir/lib/Conversion/VectorToLLVM/ConvertVectorToLLVMPass.cpp | 1 |
1 files changed, 1 insertions, 0 deletions
diff --git a/mlir/lib/Conversion/VectorToLLVM/ConvertVectorToLLVMPass.cpp b/mlir/lib/Conversion/VectorToLLVM/ConvertVectorToLLVMPass.cpp index ec6c707a3656..85cd06516259 100644 --- a/mlir/lib/Conversion/VectorToLLVM/ConvertVectorToLLVMPass.cpp +++ b/mlir/lib/Conversion/VectorToLLVM/ConvertVectorToLLVMPass.cpp @@ -64,6 +64,7 @@ void LowerVectorToLLVMPass::runOnOperation() { populateVectorToVectorCanonicalizationPatterns(patterns); populateVectorSlicesLoweringPatterns(patterns); populateVectorContractLoweringPatterns(patterns); + populateVectorTransposeLoweringPatterns(patterns); (void)applyPatternsAndFoldGreedily(getOperation(), std::move(patterns)); } |