https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119912
Segher Boessenkool <segher at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Last reconfirmed| |2025-09-25
Status|UNCONFIRMED |NEW
Ever confirmed|0 |1
--- Comment #2 from Segher Boessenkool <segher at gcc dot gnu.org> ---
So, for a constant shift count we probably should sign-extend it to as many
bits as valid for that particular shift?
At least for sizes up to word. For double-word it is harder. Those insns
require arch 2.07 (power8), but we so not have xxsplti* until arch 3.0
(power9),
so there is a bit of a hole there. Of course we will survive that ;-)
