This is an archive of the discontinued LLVM Phabricator instance.

[MLIR] Expose atomicrmw and/or
ClosedPublic

Authored by wsmoses on Dec 28 2021, 3:41 PM.

Details

Summary

LLVM (dialect and IR) have atomics for and/or. This patch enables atomic_rmw ops in the standard dialect for and/or that lower to these (in addition to the existing atomics such as addi, etc).

Diff Detail

Event Timeline

wsmoses created this revision.Dec 28 2021, 3:41 PM
wsmoses requested review of this revision.Dec 28 2021, 3:41 PM
wsmoses updated this revision to Diff 396441.Dec 28 2021, 4:16 PM

Remove excess code

mehdi_amini accepted this revision.Dec 28 2021, 9:15 PM

LGTM.
I wonder why isn't this operation didn't move in the arith dialect as well?

mlir/lib/Analysis/AffineAnalysis.cpp
1

Nit: you should install clang-format on your machine

This revision is now accepted and ready to land.Dec 28 2021, 9:15 PM
This revision was automatically updated to reflect the committed changes.

Yeah I concur that this (and the other atomics) would be wise to move.