https://gcc.gnu.org/bugzilla/show_bug.cgi?id=122207
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |NEW
Last reconfirmed| |2025-10-09
Ever confirmed|0 |1
--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Note loop header copying is critical to make store-motion work. The PHI
visible at SCCP time then doens't compute ctz() but ctz() offsetted by *bit_no.
Not
sure why niter analysis fails for
<bb 4> [local count: 105119324]:
bit_no__lsm.3_6 = *bit_no_11(D);
<bb 5> [local count: 955630224]:
# bits_16 = PHI <bits_10(8), bits_8(D)(4)>
# bit_no__lsm.3_14 = PHI <_2(8), bit_no__lsm.3_6(4)>
bits_10 = bits_16 >> 1;
_2 = bit_no__lsm.3_14 + 1;
_3 = bits_10 & 1;
if (_3 == 0)
goto <bb 8>; [89.00%]
else
goto <bb 6>; [11.00%]
<bb 8> [local count: 850510900]:
goto <bb 5>; [100.00%]
<bb 6> [local count: 105119324]:
# _19 = PHI <_2(5)>
