[llvm-bugs] [Bug 38197] New: Compiler producing suboptimal code for vector packed fp operation followed by a vector insert

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=38197

            Bug ID: 38197
           Summary: Compiler producing suboptimal code for vector packed
                    fp operation followed by a vector insert
           Product: new-bugs
           Version: trunk
          Hardware: PC
                OS: Linux
            Status: NEW
          Severity: normal
          Priority: P
         Component: new bugs
          Assignee: unassignedbugs at nondot.org
          Reporter: douglas_yung at playstation.sony.com
                CC: llvm-bugs at lists.llvm.org

Change r336971 caused a regression in the codegen for a certain pattern that
was fixed previously in r197145.

Consider the following code:

/* test.c */
#include <x86intrin.h>

__m128 foo(__m128 a, __m128 b) {
  __m128 c = a + b;

  return (__m128) { c[0], a[1], a[2], a[3] };
}

Prior to upstream r197145, the compiler would generate the following code for
foo() when compiled with optimizations (-O2):

addps %xmm0, %xmm1
movss %xmm1, %xmm0

After the fix in r197145, the compiler generated the more optimal:

addss %xmm1, %xmm0

But now after r336971, we are no longer generating the optimal code and are now
generating the original code

addps   %xmm0, %xmm1
movss   %xmm1, %xmm0

-- 
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/20180717/3bc7db2f/attachment.html>