[Bug middle-end/106804] Poor codegen for selecting and incrementing value behind a reference

[pinskia at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106804

--- Comment #5 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Reference case:
  <bb 2> [local count: 1073741824]:
  _1 = *a_7(D);
  _2 = *b_8(D);
  if (_1 > _2)
    goto <bb 3>; [50.00%]
  else
    goto <bb 4>; [50.00%]

  <bb 3> [local count: 536870913]:
  _3 = _1 + 1;
  *a_7(D) = _3;
  goto <bb 5>; [100.00%]

  <bb 4> [local count: 536870913]:
  _4 = _2 + 1;
  *b_8(D) = _4;

  <bb 5> [local count: 1073741824]:

Pointer case:

  <bb 2> [local count: 1073741824]:
  _1 = *a_7(D);
  _2 = *b_8(D);
  if (_1 > _2)
    goto <bb 3>; [50.00%]
  else
    goto <bb 4>; [50.00%]

  <bb 3> [local count: 536870912]:

  <bb 4> [local count: 1073741824]:
  # iftmp.0_5 = PHI <a_7(D)(3), b_8(D)(2)>
  # prephitmp_10 = PHI <_1(3), _2(2)>
  _4 = prephitmp_10 + 1;
  *iftmp.0_5 = _4;


Pointercase 2 which does the reference case really:
    void increment_largest2(int* a, int* b) {
        (*a > *b ? ++*a : ++*b);
    }

x86_64 clang trunk does this though:
increment_largest(int&, int&):              # @increment_largest(int&, int&)
        movl    (%rdi), %eax
        movl    (%rsi), %ecx
        cmpl    %ecx, %eax
        cmovgl  %eax, %ecx
        cmovgq  %rdi, %rsi
        incl    %ecx
        movl    %ecx, (%rsi)
        retq

But 2 cmov's might be worse than a conditional branch on x86_64.