projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: e8b98a5) | patch
[mlir][linalg][bufferize][NFC] Simplify InsertSliceOp bufferization
authorMatthias Springer <springerm@google.com>
Thu, 6 Jan 2022 08:34:01 +0000 (17:34 +0900)
committerMatthias Springer <springerm@google.com>
Thu, 6 Jan 2022 08:35:45 +0000 (17:35 +0900)
commit0e5f258452b053cc3374754efaeabe3c30f42482
tree5f68d433262e6442f0d76d8aeb9bc560ce028faetree | snapshot
parente8b98a5216dbfdaa31f7016955f9586cef94a626commit | diff
[mlir][linalg][bufferize][NFC] Simplify InsertSliceOp bufferization

No need to keep track of equivalent extract_slice / insert_slice tensors during bufferization. Just emit a copy, it will fold away.

Note: The analysis still keeps track of equivalent tensors to make the correct inplace bufferization decisions.

Differential Revision: https://reviews.llvm.org/D116684
mlir/include/mlir/Dialect/Linalg/ComprehensiveBufferize/TensorInterfaceImpl.h diff | blob | history
mlir/lib/Dialect/Linalg/ComprehensiveBufferize/TensorInterfaceImpl.cpp diff | blob | history
mlir/lib/Dialect/Linalg/Transforms/ComprehensiveBufferizePass.cpp diff | blob | history
mlir/test/lib/Dialect/Linalg/TestComprehensiveBufferize.cpp diff | blob | history
Domain: System / Toolchain;