[llvm-bugs] [Bug 32558] New: r299387 introduced regression in performance of generate code for test-suite/SingleSource/Benchmarks/Misc/matmul_f64_4x4.c

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=32558

            Bug ID: 32558
           Summary: r299387 introduced regression in performance of
                    generate code for
                    test-suite/SingleSource/Benchmarks/Misc/matmul_f64_4x4
                    .c
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Windows NT
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Backend: X86
          Assignee: unassignedbugs at nondot.org
          Reporter: zvi.rackover at intel.com
                CC: llvm-bugs at lists.llvm.org

Created attachment 18242
  --> https://bugs.llvm.org/attachment.cgi?id=18242&action=edit
Module for which generated code has regressed

The attached LLVM-IR module was compiled from the LLVM test-suite file
test-suite/SingleSource/Benchmarks/Misc/matmul_-f64_4x4.c, with the option
-march=skx.

The performance quality of the generated code for the function 'wrap_mul4' has
regressed.

By examining the difference in the generated codes it seems that movsd+vpermpd
sequences are being generated instead single vbroadcastsd instructions.

Note: the commit log in r299387 mentions that 
"There are a couple of minor regressions this patch unearths due to some
missing MOVDDUP/BROADCAST folds that I will address in a future patch."

-- 
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/20170406/65dda450/attachment.html>