This is an archive of the discontinued LLVM Phabricator instance.

[AMDGPU] Fix incorrect register pressure calculation
ClosedPublic

Authored by rampitec on May 10 2017, 7:21 PM.

Details

Summary

Earlier fix D32572 introduced a bug where live-ins were calculated
for basic block instead of scheduling region. This change fixes it.

Diff Detail

Repository
rL LLVM

Event Timeline

rampitec created this revision.May 10 2017, 7:21 PM
vpykhtin added inline comments.May 11 2017, 8:29 AM
lib/Target/AMDGPU/GCNSchedStrategy.cpp
426

you can use skipDebugInstructionsForward from MachineBasicBlock.h for the same purpose

rampitec updated this revision to Diff 98636.May 11 2017, 8:42 AM
rampitec marked an inline comment as done.

Changed to use skipDebugInstructionsForward().

lib/Target/AMDGPU/GCNSchedStrategy.cpp
426

Thanks.

This revision is now accepted and ready to land.May 11 2017, 10:21 AM
This revision was automatically updated to reflect the committed changes.