<html>
    <head>
      <base href="https://bugs.llvm.org/">
    </head>
    <body><table border="1" cellspacing="0" cellpadding="8">
        <tr>
          <th>Bug ID</th>
          <td><a class="bz_bug_link 
          bz_status_NEW "
   title="NEW - [X86][SSE] Improve equality memcmp support"
   href="https://bugs.llvm.org/show_bug.cgi?id=33325">33325</a>
          </td>
        </tr>

        <tr>
          <th>Summary</th>
          <td>[X86][SSE] Improve equality memcmp support
          </td>
        </tr>

        <tr>
          <th>Product</th>
          <td>libraries
          </td>
        </tr>

        <tr>
          <th>Version</th>
          <td>trunk
          </td>
        </tr>

        <tr>
          <th>Hardware</th>
          <td>PC
          </td>
        </tr>

        <tr>
          <th>OS</th>
          <td>Windows NT
          </td>
        </tr>

        <tr>
          <th>Status</th>
          <td>NEW
          </td>
        </tr>

        <tr>
          <th>Severity</th>
          <td>enhancement
          </td>
        </tr>

        <tr>
          <th>Priority</th>
          <td>P
          </td>
        </tr>

        <tr>
          <th>Component</th>
          <td>Backend: X86
          </td>
        </tr>

        <tr>
          <th>Assignee</th>
          <td>unassignedbugs@nondot.org
          </td>
        </tr>

        <tr>
          <th>Reporter</th>
          <td>llvm-dev@redking.me.uk
          </td>
        </tr>

        <tr>
          <th>CC</th>
          <td>filcab@gmail.com, llvm-bugs@lists.llvm.org, spatel+llvm@rotateright.com
          </td>
        </tr></table>
      <p>
        <div>
        <pre>int cmpeq16(const char *a, const char *b) {
        return 0 == __builtin_memcmp(a, b, 16);
}
int cmpeq32(const char *a, const char *b) {
        return 0 == __builtin_memcmp(a, b, 32);
}

On SSE2-AVX1 targets, equality memcmp of 16 bytes lowers using the SIMD unit
but remains scalarized for 32 byte compares:

cmpeq16(char const*, char const*):                       # @cmpeq16(char
const*, char const*)
        vmovdqu (%rdi), %xmm0
        xorl    %eax, %eax
        vpcmpeqb        (%rsi), %xmm0, %xmm0
        vpmovmskb       %xmm0, %ecx
        cmpl    $65535, %ecx            # imm = 0xFFFF
        sete    %al
        retq

cmpeq32(char const*, char const*):                       # @cmpeq32(char
const*, char const*)
        movq    16(%rdi), %rax
        movq    (%rdi), %rcx
        movq    8(%rdi), %rdx
        movq    24(%rdi), %rdi
        xorq    24(%rsi), %rdi
        xorq    8(%rsi), %rdx
        xorq    16(%rsi), %rax
        xorq    (%rsi), %rcx
        orq     %rax, %rcx
        orq     %rdi, %rdx
        xorl    %eax, %eax
        orq     %rdx, %rcx
        sete    %al
        retq

cmpeq32 is even worse on 32-bit targets....

Ideally it'd be something like:

cmpeq32(char const*, char const*):
        vmovdqu (%rdi), %xmm0
        vmovdqu 16(%rdi), %xmm1
        xorl    %eax, %eax
        vpcmpeqb        (%rsi), %xmm0, %xmm0
        vpcmpeqb        16(%rsi), %xmm1, %xmm1
        vpand        %xmm1, %xmm0, %xmm0
        vpmovmskb       %xmm0, %ecx
        cmpl    $65535, %ecx            # imm = 0xFFFF
        sete    %al
        retq

I'm not sure what the upper limit should be but 32-bytes on SSE2-AVX1 and
64-bytes on AVX2 should definitely be fine (no idea what the best solution is
on AVX512).</pre>
        </div>
      </p>


      <hr>
      <span>You are receiving this mail because:</span>

      <ul>
          <li>You are on the CC list for the bug.</li>
      </ul>
    </body>
</html>