[Bug middle-end/116815] Make better use of overflow flags in codegen of min/max(a, add/sub(a, b))

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

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116815

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|normal                      |enhancement
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2024-09-23
     Ever confirmed|0                           |1
          Component|target                      |middle-end

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
The easiest way to fix this is transform (late in gimple):
  _1 = a_2(D) + b_3(D);
  _5 = MAX_EXPR <_1, a_2(D)>;

into:
  _tmp = .ADD_OVERFLOW (a_2(D), b_3(D));
  _tmp1 = IMAGPART_EXPR <_tmp>;
  _tmp2 = _tmp1 != 0;
  _tmp3 = REALPART_EXPR <_tmp>;
  _5 = _tmp2 ? _tmp3 : a_2(D);

If we have addc support.

Likewise the others. Which should be too hard to add to tree-ssa-math-opts.cc
with the other ADD_OVERFLOW optimizations.