Page MenuHomePhabricator
Differential D18507

[AArch64] Do not lower scalar sdiv/udiv to a shifts + mul sequence when optimizing for minsize
ClosedPublic
Actions

Authored by haicheng on Mar 27 2016, 8:22 PM.

Details

Reviewers
gberry
jmolloy
mssimpso
mcrosier
Summary

Mimic what x86 does when optimizing sdiv/udiv for minsize.

Diff Detail

Repository
rL LLVM

Event Timeline

haicheng updated this revision to Diff 51768.Mar 27 2016, 8:22 PM
haicheng retitled this revision from to [AArch64] Do not lower scalar sdiv/udiv to a shifts + mul sequence when optimizing for minsize.
haicheng updated this object.
haicheng added reviewers: mcrosier, gberry, mssimpso.
haicheng set the repository for this revision to rL LLVM.
haicheng added a subscriber: llvm-commits.
Herald added subscribers: mcrosier, rengolin, aemerson. · View Herald TranscriptMar 27 2016, 8:22 PM
jmolloy accepted this revision.Mar 28 2016, 2:45 AM
jmolloy added a reviewer: jmolloy.
jmolloy added a subscriber: jmolloy.

LGTM, thanks!

This revision is now accepted and ready to land.Mar 28 2016, 2:45 AM
flyingforyou added a subscriber: flyingforyou.Mar 28 2016, 3:03 AM

LGTM! too.

haicheng closed this revision.Mar 29 2016, 6:24 AM

Committed in r264606. Thank you for reviewing this, James and Junmo.

Revision Contents
Changeset List

PathSize
lib/
Target/
AArch64/
2 lines
17 lines
test/
CodeGen/
AArch64/
45 lines

Diff 51768

View Options

lib/Target/AArch64/AArch64ISelLowering.h

Loading...
View Options

lib/Target/AArch64/AArch64ISelLowering.cpp

Loading...
View Options

test/CodeGen/AArch64/div_minsize.ll

Loading...