[llvm-bugs] [Bug 27435] New: [X86][SSE] Failure to vectorize __builtin_sqrt/__builtin_sqrtf

via llvm-bugs llvm-bugs at lists.llvm.org
Wed Apr 20 07:21:54 PDT 2016


https://llvm.org/bugs/show_bug.cgi?id=27435

            Bug ID: 27435
           Summary: [X86][SSE] Failure to vectorize
                    __builtin_sqrt/__builtin_sqrtf
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Windows NT
            Status: NEW
          Severity: normal
          Priority: P
         Component: Backend: X86
          Assignee: unassignedbugs at nondot.org
          Reporter: llvm-dev at redking.me.uk
                CC: llvm-bugs at lists.llvm.org
    Classification: Unclassified

Trunk fails to vectorize:

#include <x86intrin.h>

__m128d sqrtd2(double* v) {
  return _mm_setr_pd(__builtin_sqrt(v[0]), __builtin_sqrt(v[1]));
}

__m128 sqrtf4(float* v) {
  return _mm_setr_ps(__builtin_sqrtf(v[0]), __builtin_sqrtf(v[1]),
__builtin_sqrtf(v[2]), __builtin_sqrtf(v[3]));
}

clang -S -O3 test.c -o -

sqrtd2:
 vsqrtsd (%rdi), %xmm0, %xmm0
 vsqrtsd 8(%rdi), %xmm0, %xmm1
 vunpcklpd %xmm1, %xmm0, %xmm0 # xmm0 = xmm0[0],xmm1[0]
 retq

sqrtf4:
 vsqrtss (%rdi), %xmm0, %xmm0
 vsqrtss 4(%rdi), %xmm0, %xmm1
 vsqrtss 8(%rdi), %xmm0, %xmm2
 vsqrtss 12(%rdi), %xmm0, %xmm3
 vinsertps $16, %xmm1, %xmm0, %xmm0 # xmm0 = xmm0[0],xmm1[0],xmm0[2,3]
 vinsertps $32, %xmm2, %xmm0, %xmm0 # xmm0 = xmm0[0,1],xmm2[0],xmm0[3]
 vinsertps $48, %xmm3, %xmm0, %xmm0 # xmm0 = xmm0[0,1,2],xmm3[0]
 retq

Note that this is for __builtin_sqrt not ::sqrt()!

I'd expect:

sqrtd2:
 vsqrtpd (%rdi), %xmm0
 retq

sqrtf4:
 vsqrtps (%rdi), %xmm0
 retq

-- 
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/20160420/8c1071fb/attachment-0001.html>


More information about the llvm-bugs mailing list