This is an archive of the discontinued LLVM Phabricator instance.

Restructuring changes in preparation for enabling the x86 call frame optimization for 64-bit targets
ClosedPublic

Authored by DavidKreitzer on Apr 18 2016, 6:37 AM.

Details

Summary

This is step 1 of 2 for fixing pr27241.

The changes to generalize the x86 call frame optimization to support 64-bit targets are fairly small, but to make the review easier, I separated these restructuring changes from the functional changes, which will come in the next patch. No functional changes are intended from this patch.

Diff Detail

Event Timeline

DavidKreitzer retitled this revision from to Restructuring changes in preparation for enabling the x86 call frame optimization for 64-bit targets.
DavidKreitzer updated this object.
DavidKreitzer added reviewers: mkuper, hans, rnk, delena.
DavidKreitzer added a subscriber: llvm-commits.
DavidKreitzer added inline comments.Apr 18 2016, 6:43 AM
lib/Target/X86/X86CallFrameOptimization.cpp
108–110

The deletion of "const" here was accidental. I'll need to remove it in the next patch, but I'll add it back for this one.

rnk accepted this revision.Apr 18 2016, 10:59 AM
rnk edited edge metadata.

lgtm

lib/Target/X86/X86CallFrameOptimization.cpp
109–110

Total bikeshedding: I think you can omit "Stack" and use SlotSize and Log2SlotSize. If you grep around LLVM for SlotSize, you'll find that this variable name is used very frequently with this meaning. It's nice to be consistent.

This revision is now accepted and ready to land.Apr 18 2016, 10:59 AM
DavidKreitzer edited edge metadata.

Thanks for the review, Reid! I'm all for consistency. I updated the patch to reflect the variable name changes and to restore "const" on the MRI member.

This revision was automatically updated to reflect the committed changes.