https://gcc.gnu.org/bugzilla/show_bug.cgi?id=127306
Jeffrey A. Law <law at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |ASSIGNED
Ever confirmed|0 |1
Summary|RISC-V bclr_lowest_set_bit |[16/17 Regression] RISC-V
|pass produces wrong code |bclr_lowest_set_bit pass
| |produces wrong code
Last reconfirmed| |2026-09-10
--- Comment #3 from Jeffrey A. Law <law at gcc dot gnu.org> ---
Definitely an independent issue. I see the what's causing this to do the wrong
thing (the original AND is still in the IL, thus we clear two bits each
iteration). There's also a comment that I want to investigate. No idea how
that FIXME slid through, but it's worth investigation. Point is things are
well under control and I expect this will be fixed shortly. It's definitely a
gcc-16/gcc-17 regression.
