https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96094
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Ever confirmed|0 |1
Last reconfirmed| |2020-07-07
Status|UNCONFIRMED |NEW
--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
OK, we could do that based on the fact that we optimize 0 / a to a (but not
0 / 0, so we'd not optimize a / false then). Since the IL we see is
<bb 2> [local count: 1073741824]:
_1 = (int) a_4(D);
_2 = (int) b_5(D);
_3 = _1 / _2;
_6 = (_Bool) _3;
return _6;
the match.pd pattern should look at value-ranges, not types here.
