[llvm-bugs] [Bug 31629] New: performance regression in LoopStrengthReduce

via llvm-bugs llvm-bugs at lists.llvm.org
Fri Jan 13 07:28:11 PST 2017


            Bug ID: 31629
           Summary: performance regression in LoopStrengthReduce
           Product: new-bugs
           Version: 3.9
          Hardware: PC
                OS: Windows NT
            Status: NEW
          Severity: normal
          Priority: P
         Component: new bugs
          Assignee: unassignedbugs at nondot.org
          Reporter: jeroen.dobbelaere at synopsys.com
                CC: llvm-bugs at lists.llvm.org
    Classification: Unclassified

Created attachment 17837
  --> https://llvm.org/bugs/attachment.cgi?id=17837&action=edit
testcase showing bad behavior

During the 3.8-3.9 timeframe, a number of changes were done that exhibit some
severe quality-of-code regression.

For the testcase, I am using the xscale target to show the regression, but we
noticed it in our custom version of llvm. So this behavior is likely to be
present for most targets.

The main change that shows the regression is:

Author: Manuel Jacob <me at manueljacob.de>
Date:   Thu Jun 2 22:14:26 2016 +0000

    [PM] Schedule InstSimplify after late LICM run, to clean up LCSSA nodes.

    The module pass pipeline includes a late LICM run after loop
    unrolling.  LCSSA is implicitly run as a pass dependency of LICM.  However
    cleanup pass was run after this, so the LCSSA nodes ended in the optimized

    Reviewers: hfinkel, mehdi_amini

    Subscribers: majnemer, bruno, mzolotukhin, mehdi_amini, llvm-commits

    Differential Revision: http://reviews.llvm.org/D20606

    git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@271602

This change activates the instructionSimplifyPass, but, this change only allows
us to see when LoopStrengthReduce is breaking down.

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/20170113/6714d9b4/attachment.html>

More information about the llvm-bugs mailing list