[Bug tree-optimization/92734] Missing match.pd simplification done by fold_binary_loc on generic

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92734

--- Comment #7 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
I'm wondering if we shouldn't adjust also the
  /* (A +- B) - A       ->  +- B */
  /* (A +- B) -+ B      ->  A */
  /* A - (A +- B)       -> -+ B */
  /* A +- (B -+ A)      ->  +- B */
patterns.  Say testcases like:
int
f1 (int x, unsigned y)
{
  return (int) (x + y) - x;
}

int
f2 (int x, unsigned y)
{
  int a = x + y;
  return a - x;
}
aren't really optimized ATM until combine.  And they aren't really optimized
even with -fwrapv until combine, so the "turn on -fwrapv late in GIMPLE and
redoce a few optimizations" proposal wouldn't solve that.