[Bug rtl-optimization/85925] [7/8/9 regression] Mis-compilation at -02, masking with 257 goes wrong in combine

[ebotcazou at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85925

--- Comment #9 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
> I'm running this on all archs, to see what (if anything) degrades.

I have both wins and losses, unlike the reversion of the nonzero_bits1 change
where I have only losses, but I nevertheless think it's _not_ the way to go.