https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90713

--- Comment #3 from rguenther at suse dot de <rguenther at suse dot de> ---
On Mon, 3 Jun 2019, segher at gcc dot gnu.org wrote:

> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90713
> 
> Segher Boessenkool <segher at gcc dot gnu.org> changed:
> 
>            What    |Removed                     |Added
> ----------------------------------------------------------------------------
>              Status|UNCONFIRMED                 |NEW
>    Last reconfirmed|                            |2019-06-03
>      Ever confirmed|0                           |1
> 
> --- Comment #2 from Segher Boessenkool <segher at gcc dot gnu.org> ---
> Confirmed.  And confirmed that the testcase works fine with -maltivec.
> 
> vect_float is the wrong selector to see if vectors of float are enabled by
> current compiler options:
> 
> # Return 1 if the target supports hardware vectors of float when
> # -funsafe-math-optimizations is enabled, 0 otherwise.
> #
> # This won't change for different subtargets so cache the result.
> 
> This testcase will need a test to see if there are 16 byte vector floats.
> Something that just tries to create a
>   typedef float v4sf __attribute__((vector_size(16)));
> variable should do fine.
> 
> Oh, and the gimple front end shouldn't ICE the compiler, of course :-/

The GIMPLE FE expects valid GIMPLE - but yes, the GIMPLE verifier
lacks verification of certain vector target constraints so we end
up ICEing in the expander or lowering things unexpectedly.

The GIMPLE FE is a unit-testing / debugging tool, not a "proper"
frontend ;)

I guess I'll add { dg-additional-options "-maltivec" { target powerpc*-*-* 
} }

Reply via email to