Hi,
this is patch I comitted.
Thanks for looking into it,
Honza
PR middle-end/81030
* gcc.dg/pr81030.c: New test.
* cfgbuild.c (find_many_sub_basic_blocks): Update REG_BR_PROB note
when gimple level profile disagrees with what RTL expander did.
Index: tests
Hi,
thanks for looking into this issue. It is quite weird indeed.
> Commenting out this bit makes the compilation succeed.
so my understanding is that RTL expansions confuses itself and redistributes
probabilities to two jumps while immediately optimizing out the second...
I think in such case i
Hi,
this patch fixes PR81030, an P1 ICE in expand.
I.
For the test-case from the patch compiled at -O1 we have at .optimized:
...
_21 = 0;
_22 = c.5_6 != 0;
_23 = _21 | _22;
if (_23 != 0)
goto ; [73.27%] [count: INV]
else
goto ; [26.73%] [count: INV]
;;succ: 7 [73.3
