patch: xfail some altivec tests

[Janis Johnson]

On Wed, Apr 21, 2004 at 09:52:29PM -0700, Mark Mitchell wrote:
> Zack Weinberg wrote:
>
> >Aldy Hernandez <aldyh@redhat.com> writes:
> >
> >>On Wed, Apr 21, 2004 at 07:38:20PM -0700, Zack Weinberg wrote:
> >>
> >>>Aldy Hernandez <aldyh@redhat.com> writes:
> >>>>Powerpc-eabispe has conflicting opcodes with AltiVec.  They can never
> >>>>be in the same executable.  Parts of the backend rely on this
> >>>>(particularly the %y operand).
> >>>>I'm XFAILing these tests for said architecture.
> >>>
> >>>Better to not run them at all, surely?
> >>
> >>Uh yeah.  How do I do that?
> >
> >Uh, I dunno either.  Anyone?
>
> Write dg-require-altivec.  Have it check the target triplet.  There's 
> existing dg-require goo that does similar stuff.

As I just learned yesterday, the dg-require* procs in gcc-dg.exp use the
check_*_available procs in target-supports.exp.  My patch to make the
default of the vmx tests 'compile' instead of 'run' if the target does
not have Altivec hardware adds check_vmx_hw_available, which could be
used by a new dg-require-vmx-hw.  It returns 0 for powerpc-eabispe.

Janis