This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug tree-optimization/79534] New: tree-ifcombine aarch64 regression with trunk@245151
- From: "brzycki at gmail dot com" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: Wed, 15 Feb 2017 17:18:55 +0000
- Subject: [Bug tree-optimization/79534] New: tree-ifcombine aarch64 regression with trunk@245151
- Auto-submitted: auto-generated
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79534
Bug ID: 79534
Summary: tree-ifcombine aarch64 regression with trunk@245151
Product: gcc
Version: 7.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: brzycki at gmail dot com
Target Milestone: ---
Created attachment 40752
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=40752&action=edit
Reduced test-case. Larger scores are better.
I am seeing about a 6-10% performance regression on a proprietary benchmark on
AArch64, depending on the target micro-architecture.
# SHA d2c3261, trunk@245149, patch immediately preceding regression
/gcc/2017-02-03-12-17-59-d2c3261-master/bin/aarch64-sarc-linux-gnu-g++ -O3
-mcpu=cortex-a57 -std=c++11 -c a.cpp -o a.o
/gcc/2017-02-03-12-17-59-d2c3261-master/bin/aarch64-sarc-linux-gnu-g++ -o a
a.o
(best of 3 runs)
SCORE_DIRECTION ==> larger is better
SCORE ==> 1619.832509
# SHA f56c861, trunk@245151, regressing patch
/gcc/2017-02-03-14-03-35-f56c861-master/bin/aarch64-sarc-linux-gnu-g++ -O3
-mcpu=cortex-a57 -std=c++11 -c a.cpp -o a.o
/gcc/2017-02-03-14-03-35-f56c861-master/bin/aarch64-sarc-linux-gnu-g++ -o a
a.o
(best of 3 runs)
SCORE_DIRECTION ==> larger is better
SCORE ==> 1468.440527
# SHA e502db7, a recent 2/14/2017 tip build
/gcc/2017-02-14-23-11-19-e502db7-master/bin/aarch64-sarc-linux-gnu-g++ -O3
-mcpu=cortex-a57 -std=c++11 -c a.cpp -o a.o
/gcc/2017-02-14-23-11-19-e502db7-master/bin/aarch64-sarc-linux-gnu-g++ -o a
a.o
SCORE_DIRECTION ==> larger is better
SCORE ==> 1483.952217
I do not see the slowdown when testing on x86_64.