This is an archive of the discontinued LLVM Phabricator instance.

[mlir][linalg][bufferize][NFC] Simplify allocateBuffersForResults
ClosedPublic

Authored by springerm on Oct 1 2021, 8:01 AM.

Details

Summary

Reuse getResultBuffer.

Nicolas: Please check if it is OK to use output directly instead of:

Value dimTensor = bvm.lookupOrDefault(output);

Depends On D110940

Event Timeline

springerm created this revision.Oct 1 2021, 8:01 AM
springerm requested review of this revision.Oct 1 2021, 8:01 AM
Herald added a project: Restricted Project. · View Herald TranscriptOct 1 2021, 8:01 AM
nicolasvasilache accepted this revision.Oct 5 2021, 3:25 AM

this seems fine, I cannot remember the layering evolution that made me not reuse code.
Thanks for improving!

mlir/lib/Dialect/Linalg/Transforms/ComprehensiveBufferize.cpp
1593

nit: wants

This revision is now accepted and ready to land.Oct 5 2021, 3:25 AM

this seems fine, I cannot remember the layering evolution that made me not reuse code.
Thanks for improving!

Ah it's because you added that function to refactor stuff and avoid code duplication, makes sense!

springerm updated this revision to Diff 378545.Oct 10 2021, 6:50 PM
springerm marked an inline comment as done.

address comments

This revision was landed with ongoing or failed builds.Oct 12 2021, 5:36 PM
This revision was automatically updated to reflect the committed changes.