Page MenuHomePhabricator

Enable support for floating-point division reductions
Needs ReviewPublic

Authored by yuyichao on Jun 10 2017, 5:52 AM.

Details

Reviewers
mkuper
jmolloy
Summary

Similar to fsub, fdiv can also be vectorized using fmul.

Also http://llvm.org/viewvc/llvm-project?view=revision&revision=215200

Diff Detail

Event Timeline

yuyichao created this revision.Jun 10 2017, 5:52 AM
mkuper edited reviewers, added: jmolloy; removed: llvm-commits.Jun 19 2017, 5:38 PM
mkuper added a subscriber: llvm-commits.
delena added a subscriber: delena.Jun 22 2017, 12:24 AM

I suggest to add one more test that shows that vectorization does not happen when the reduction variable is the second argument of fdiv.

vchuravy added a project: Restricted Project.Jan 14 2020, 1:29 AM