This is an archive of the discontinued LLVM Phabricator instance.

[ARM] Select MVE fp add and sub
ClosedPublic

Authored by dmgreen on Jun 13 2019, 5:46 AM.

Details

Summary

Same as integer addition, we can add simple floating point MVE addition and subtraction patterns.

Initial code by David Sherwood

Diff Detail

Repository
rL LLVM

Event Timeline

dmgreen created this revision.Jun 13 2019, 5:46 AM
dmgreen updated this revision to Diff 206164.Jun 24 2019, 1:29 AM
dmgreen added a reviewer: SjoerdMeijer.

Rebased for new patterns and moved next to instruction definition.

SjoerdMeijer added inline comments.Jun 24 2019, 2:14 AM
llvm/test/CodeGen/Thumb2/mve-simple-arith.ll
34 ↗(On Diff #206164)

do we need to test this when we only have -mattr=+mve?

dmgreen updated this revision to Diff 206672.EditedJun 26 2019, 7:56 AM

Now shows the awful code we generate from expansions when mve+fullfp16 is used (not mve.fp). The vdups will be cleared up later

This revision is now accepted and ready to land.Jun 28 2019, 12:28 AM
This revision was automatically updated to reflect the committed changes.