[Bug target/106736] [13 Regression] ICE in gen_movxo, at config/rs6000/mma.md:333

["linkw at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106736

--- Comment #12 from Kewen Lin <linkw at gcc dot gnu.org> ---
(In reply to Segher Boessenkool from comment #11)
> (In reply to Peter Bergner from comment #10)
> > (In reply to Segher Boessenkool from comment #9)
> > > When MMA is not enabled,
> > ...
> > > the __vector_{quad,pair} types should not exist, 
> > 
> > Unfortunately, target type initialization only occurs once at the very
> > beginning
> 
> That is what indirection is for (or copying more likely, in this case).
> 
> > and if we don't initialize them because of the command line
> > options in affect at the time, then we get problems like PR96125, so we have
> > to initialize these types always, just like we do for built-in functions.
> 
> I don't understand?

Since we need to support some target pragma/attribute usage, we have to
initialize those built-in functions globally, otherwise for example assuming
command line option specifying cpu type power8, we don't initialize mma bif,
meanwhile there is one function with mma target attribute using mma bif, it
will raise error for no such bif. Although we always initialize those bifs , we
will still check their validity when expanding.

But your question makes me consider that if we can apply the similar idea for
built-in type. Now we emit error msg when expanding bif by checking its
supported (ENB...), if we can find somewhere in expander for checking built-in
type, we can add one hook there. I'll have a further look on this.