https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117978
ktkachov at gcc dot gnu.org changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |rsandifo at gcc dot gnu.org
--- Comment #2 from ktkachov at gcc dot gnu.org ---
(In reply to Andrew Pinski from comment #1)
> Confirmed. Could this be done at the gimple level? At least for the simple
> cases it would work. I am not sure if there are any other cases where it
> would matter.
The GIMPLE folding hooks for the intrinsics would be a natural place to do this
indeed. I think we'll need to agree on whether we're comfortable making this
transformation in spite of the user's intent when writing intrinsics.
I think it does fit into the class of transformations that we should be
comfortable.
