[llvm-bugs] [Bug 51149] New: Performance regression saw after https://reviews.llvm.org/rG9c4c2f24725e9f98b96fb360894276d342c3ba50

via llvm-bugs llvm-bugs at lists.llvm.org
Tue Jul 20 17:08:40 PDT 2021


https://bugs.llvm.org/show_bug.cgi?id=51149

            Bug ID: 51149
           Summary: Performance regression saw after
                    https://reviews.llvm.org/rG9c4c2f24725e9f98b96fb360894
                    276d342c3ba50
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: All
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Scalar Optimizations
          Assignee: unassignedbugs at nondot.org
          Reporter: wmi at google.com
                CC: llvm-bugs at lists.llvm.org

Created attachment 25039
  --> https://bugs.llvm.org/attachment.cgi?id=25039&action=edit
compare.png

We have an internal micro-benchmark seeing 11% performance regression after
https://reviews.llvm.org/rG9c4c2f24725e9f98b96fb360894276d342c3ba50.

We extract a testcase to show the problem:
clang-without-the-patch -fexperimental-new-pass-manager -O3 1.ll -S -o /tmp/1.s
clang-with-the-patch -fexperimental-new-pass-manager -O3 1.ll -S -o /tmp/2.s

1.ll, 1.s and 2.s are attached. 

In the compare.png attached, for the same code region, we show there are 13
instructions in the 2.s while there are 10 instructions in the 1.s.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20210721/de303641/attachment.html>


More information about the llvm-bugs mailing list