projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: cbcdd74) | patch
[mlir][sparse] revert optimization for dense->csc conversion.
authorPeiming Liu <peiming@google.com>
Tue, 21 Feb 2023 21:26:46 +0000 (21:26 +0000)
committerPeiming Liu <peiming@google.com>
Tue, 21 Feb 2023 21:34:01 +0000 (21:34 +0000)
commitb7d86f3f1cfc8f7a0ac9e83d87767a4dac3832a0
tree6d60ad674afd3c8ec8f38152f4cd900520054a26tree | snapshot
parentcbcdd747e85b8d33b821d94d8114b971f31fd0d2commit | diff
[mlir][sparse] revert optimization for dense->csc conversion.

Eliminates the sort seems make the whole conversion slower (probably because loop rotation leads to bad locality).

Reviewed By: aartbik

Differential Revision: https://reviews.llvm.org/D144517
mlir/lib/Dialect/SparseTensor/Transforms/SparseTensorRewriting.cpp diff | blob | history
mlir/test/Dialect/SparseTensor/convert_dense2sparse.mlir diff | blob | history
Domain: System / Toolchain;