This is an archive of the discontinued LLVM Phabricator instance.

[mlir][sparse] fix a crash when generating sparse convolution with nchw input
ClosedPublic

Authored by Peiming on May 30 2023, 12:02 PM.

Diff Detail

Event Timeline

Peiming created this revision.May 30 2023, 12:02 PM
Herald added a project: Restricted Project. · View Herald Transcript
Peiming requested review of this revision.May 30 2023, 12:02 PM
aartbik added inline comments.May 30 2023, 12:59 PM
mlir/lib/Dialect/SparseTensor/Transforms/Sparsification.cpp
800–801

why no util call here?

aartbik added inline comments.May 30 2023, 1:03 PM
mlir/lib/Dialect/SparseTensor/Transforms/Sparsification.cpp
521

so the f == t skip should only apply to a few branches right, otherwise it cannot even occur

but it seems acceptable behavior to skip self-order

Peiming updated this revision to Diff 526755.May 30 2023, 1:04 PM

address comments

aartbik accepted this revision.May 30 2023, 1:09 PM
This revision is now accepted and ready to land.May 30 2023, 1:09 PM
This revision was landed with ongoing or failed builds.May 30 2023, 1:17 PM
This revision was automatically updated to reflect the committed changes.