projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: d4a8c73) | patch
[mlir][sparse] disable nonunit stride dense vectorization
authorAart Bik <ajcbik@google.com>
Fri, 12 Mar 2021 22:56:23 +0000 (14:56 -0800)
committerAart Bik <ajcbik@google.com>
Sat, 13 Mar 2021 00:49:32 +0000 (16:49 -0800)
commite7ee4eaaf7f051d74f0706491a1c16594a832b7d
treeace4eb13e2a5876e9f4a6d1eb1d30a5aad3df744tree | snapshot
parentd4a8c7359b57bafc7bfa2a9dea30017fb0153c1acommit | diff
[mlir][sparse] disable nonunit stride dense vectorization

This is a temporary work-around to get our all-annotations-all-flags
stress testing effort run clean. In the long run, we want to provide
efficient implementations of strided loads and stores though

Reviewed By: bixia

Differential Revision: https://reviews.llvm.org/D98563
mlir/lib/Dialect/Linalg/Transforms/Sparsification.cpp diff | blob | history
Domain: System / Toolchain;