This is an archive of the discontinued LLVM Phabricator instance.

RegisterScavenger: Allow fail without spill
ClosedPublic

Authored by arsenm on Dec 4 2018, 10:01 AM.

Details

Reviewers
MatzeB
qcolombet
Summary

AMDGPU wants to use this in some contexts where
the spilling is either impossible, or a worse alternative
to doing something else.

Diff Detail

Event Timeline

arsenm created this revision.Dec 4 2018, 10:01 AM
qcolombet accepted this revision.Feb 25 2019, 8:32 AM

LGTM. Nitpick below.

include/llvm/CodeGen/RegisterScavenging.h
161

Add the same comment as what you put on scavengeRegisterBackwards.

This revision is now accepted and ready to land.Feb 25 2019, 8:32 AM
arsenm closed this revision.Feb 25 2019, 12:28 PM
arsenm marked an inline comment as done.

r354816