[llvm-bugs] [Bug 32841] New: [X86][SSE] Use PHMINPOSUW for i8/i16 min/max reductions

[via llvm-bugs]

http://bugs.llvm.org/show_bug.cgi?id=32841

            Bug ID: 32841
           Summary: [X86][SSE] Use PHMINPOSUW for i8/i16 min/max
                    reductions
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Windows NT
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Backend: X86
          Assignee: unassignedbugs at nondot.org
          Reporter: llvm-dev at redking.me.uk
                CC: filcab at gmail.com, llvm-bugs at lists.llvm.org,
                    spatel+llvm at rotateright.com

With suitable bit twiddling, PHMINPOSUW can be used to determine the
signed/unsigned min/max element of v16i8/v8i16 vectors, not just umin v8i16.

e.g.

unsigned char umin_v16i8(__m128i a) {
  __m128i result = a;
  result = _mm_min_epu8( result, _mm_srli_epi16( result, 8 ) );
  result = _mm_minpos_epu16( result );
  return _mm_extract_epi8(result, 0);
}

char smax_v16i8(__m128i a) {
  __m128i k127 = _mm_set1_epi8(127);
  __m128i result = a;
  result = _mm_xor_si128( k127, result );
  result = _mm_min_epu8( result, _mm_srli_epi16( result, 8 ) );
  result = _mm_minpos_epu16( result );
  result = _mm_xor_si128( k127, result );
  return _mm_extract_epi8(result, 0);
}

-- 
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/20170428/c7805c21/attachment.html>