Page MenuHomePhabricator

[mlir][Vector] NFC - Compress vector to outerproduct lowering.
ClosedPublic

Authored by nicolasvasilache on Jul 2 2021, 9:34 AM.

Details

Summary

The implementation has become too unwieldy and cognitive overhead wins.
Instead compress the implementation in preparation for additional lowering paths.

Diff Detail

Event Timeline

nicolasvasilache requested review of this revision.Jul 2 2021, 9:34 AM
Herald added a project: Restricted Project. · View Herald TranscriptJul 2 2021, 9:34 AM
ThomasRaoux accepted this revision.Jul 2 2021, 9:38 AM
This revision is now accepted and ready to land.Jul 2 2021, 9:38 AM
This revision was landed with ongoing or failed builds.Jul 2 2021, 9:42 AM
This revision was automatically updated to reflect the committed changes.

This change broke the Windows mlir bot:

https://lab.llvm.org/buildbot/#/builders/13/builds/9355

It also broke the GCC testing. I suspect the output code is made dependent on the order of evaluation of function arguments.

Reverted for now.

mehdi_amini added inline comments.Jul 2 2021, 11:00 AM
mlir/lib/Dialect/Vector/VectorTransforms.cpp
1930

Here: e.t(lhs), e.t(rhs) are not ordered unfortunately.

1943

(same here)

hmm the build was green on all platforms in this diff's buildbot..
trying again in https://reviews.llvm.org/D105367

Hmm actually it was not, sorry about that, I must have misinterpreted the green checkbox, I'll be extra careful with the followup.