[Bug middle-end/118994] GCC fails to optimize (a >> 1) + (b >> 1) + ((a | b) & 1) to PAVGB/PAVGW (or equivalent instruction)

[pinskia at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118994

--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Richard Biener from comment #2)
> Sounds reasonable, note the vectorizer has a pattern for this already.  The
> issue with match.pd patterns is when we want to introduce those internal fns
> (late, close to ISEL).

There is PROP_last_full_fold now which I think is the kind of things to be used
here ...