https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80520
Jeffrey A. Law <law at redhat dot com> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |law at redhat dot com
--- Comment #8 from Jeffrey A. Law <law at redhat dot com> ---
I'd look more at split-paths rather than DOM in this case.
Prior to split-paths we have:
[ ... ]
_20 = y_33 & 1;
if (_20 != 0)
goto <bb 4>; [50.00%]
else
goto <bb 5>; [50.00%]
<bb 4> [local count: 531502203]:
_18 = _25 ^ 2567483615;
<bb 5> [local count: 1063004407]:
# prephitmp_49 = PHI <_25(3), _18(4)>
_2 = (void *) ivtmp.8_30;
[ ... ]
Split-paths comes along and decides it'd like split BB5 resulting in:
_20 = y_33 & 1;
if (_20 != 0)
goto <bb 4>; [50.00%]
else
goto <bb 6>; [50.00%]
<bb 4> [local count: 531502203]:
_18 = _25 ^ 2567483615;
<bb 5> [local count: 531502203]:
# prephitmp_43 = PHI <_18(4)>
_47 = (void *) ivtmp.8_30;
MEM[base: _47, offset: 0B] = prephitmp_43;
ivtmp.8_42 = ivtmp.8_30 + 8;
if (ivtmp.8_42 != _6)
goto <bb 3>; [98.99%]
else
goto <bb 7>; [1.01%]
<bb 6> [local count: 531502204]:
# prephitmp_49 = PHI <_25(3)>
_2 = (void *) ivtmp.8_30;
MEM[base: _2, offset: 0B] = prephitmp_49;
ivtmp.8_29 = ivtmp.8_30 + 8;
if (ivtmp.8_29 != _6)
goto <bb 3>; [98.99%]
else
goto <bb 7>; [1.01%]
I don't see any simplifications that happen as a result of that transformation
and we're not really even going to be able to simplify the branching structure.
So ISTM that split-paths is where we ought to concentrate. It's been
problematic in this space before -- it's got a number of hacks already to try
and discourage path splitting in cases where doing so it going to inhibit
if-conversion.
