Page MenuHomePhabricator

[AMDGPU] performMinMaxCombine should not optimize patterns of vectors to min3/max3
ClosedPublic

Authored by FarhanaAleen on Apr 3 2018, 11:31 AM.

Details

Summary

There are no packed instructions for min3 or max3. So, performMinMaxCombine should not optimize vectors of f16 to min3/max3.

Diff Detail

Repository
rL LLVM

Event Timeline

FarhanaAleen created this revision.Apr 3 2018, 11:31 AM

LGTM except needs check lines

test/CodeGen/AMDGPU/fmax3.ll
89 ↗(On Diff #140834)

Needs check lines

test/CodeGen/AMDGPU/fmin3.ll
87 ↗(On Diff #140834)

Ditto

This revision was not accepted when it landed; it landed in state Needs Review.Apr 3 2018, 4:03 PM
This revision was automatically updated to reflect the committed changes.