This is an archive of the discontinued LLVM Phabricator instance.

[AMDGPU] Fold copy (copy vgpr)
ClosedPublic

Authored by rampitec on Sep 26 2018, 2:38 PM.

Details

Summary

This allows to reduce a number of used VGPRs in some cases.

Diff Detail

Repository
rL LLVM

Event Timeline

rampitec created this revision.Sep 26 2018, 2:38 PM
mareko added a subscriber: mareko.Sep 27 2018, 10:00 AM

On a related note, another way to decrease VGPR usage is to fold immediates with more than 1 uses. The backend currently folds immediates with only 1 use.

This revision is now accepted and ready to land.Sep 27 2018, 11:28 AM
This revision was automatically updated to reflect the committed changes.

Why is this necessary? PeepholeOpt runs first and should have eliminated redundant copies