HomePhabricator

[PowerPC] combine rlwinm+rlwinm to rlwinm combine x3 = rlwinm x3, 27, 5, 31 x3…

Authored by shchenz on Nov 21 2019, 9:00 PM.

Description

[PowerPC] combine rlwinm+rlwinm to rlwinm combine x3 = rlwinm x3, 27, 5, 31 x3 = rlwinm x3, 19, 0, 12

to
x3 = rlwinm x3, 14, 0, 12

Reviewed by: steven.zhang

Differential Revision: https://reviews.llvm.org/D70374

Event Timeline

anil9 added a subscriber: anil9.Nov 22 2019, 10:42 PM

Hello,

This change set is breaking the buildbot :

clang-ppc64be-linux-multistage http://lab.llvm.org:8011/builders/clang-ppc64be-linux-multistage/builds/21918

Thanks,
Anil Mahmud

shchenz added a comment.EditedNov 24 2019, 6:11 PM

Hello,

This change set is breaking the buildbot :

clang-ppc64be-linux-multistage http://lab.llvm.org:8011/builders/clang-ppc64be-linux-multistage/builds/21918

Thanks,
Anil Mahmud

Hi @anil9 could you please help to provide the reproduced steps? I run make check-all & ninja check-all on Power9 64 bit, get no error. Thanks. I use latest commit e8a4e5892bd83a21451b74ebd0f41129be9f3267.

lkail added a subscriber: lkail.Nov 24 2019, 6:27 PM

What it broke seems to be tests on Power8.

anil9 added a comment.EditedNov 24 2019, 6:59 PM

It seems it needs a big endian machine to fail as all the little endian machines seem to be unaffected. And also it will fail in stage 2 not stage 1 as shown in the link.

It seems it needs a big endian machine to fail as all the little endian machines seem to be unaffected. And also it will fail in stage 2 not stage 1 as shown in the link.

Thanks for your info. I will revert the commit first and see how to fix it.