https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103771
--- Comment #34 from Jeffrey A. Law <law at gcc dot gnu.org> ---
I've always wanted to see us be able to push something like those matched
conversions down through the PHI.
That would make the code look like:
if (x.1_1 > 255)
goto <bb 3>; [INV]
else
goto <bb 4>; [INV]
<bb 3> :
_2 = -x_5(D);
_3 = _2 >> 31;
goto <bb 5>; [INV]
<bb 4> :
<bb 5> :
# tmp = PHI <_3, x_5(4)>
iftmp.0_4 = (uint8_t) tmp
And presumably we'd clean up the empty bb4 which would in turn unblock other
optimizations.
Is that what you're working on?