[Bug target/81739] Replace calls to memcmp with more efficient code if possible

[msebor at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81739

Martin Sebor <msebor at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2018-02-02
                 CC|                            |msebor at gcc dot gnu.org
     Ever confirmed|0                           |1

--- Comment #2 from Martin Sebor <msebor at gcc dot gnu.org> ---
Confirmed.  Clang 5.0 emits semi-optimal code.  For x86_64:

cmp1(unsigned long long, unsigned long long): # @cmp1(unsigned long long,
unsigned long long)

  mov qword ptr [rsp - 8], rdi
  mov qword ptr [rsp - 16], rsi
  bswap rdi
  bswap rsi
  xor ecx, ecx
  cmp rdi, rsi
  mov edx, -1
  mov eax, 1
  cmovb eax, edx
  cmove eax, ecx
  ret