Re: [tree-ssa-lno] vectorizer patch.

[Richard Henderson <rth at redhat dot com>]

On Wed, Jan 07, 2004 at 11:05:28AM -0500, Daniel Berlin wrote:
> On Jan 7, 2004, at 11:04 AM, Dorit Naishlos wrote:
> >Regarding the new build_vector_type() utility you added in tree.c to
> >replace the target hook - where did you envision that the check to 
> >verify
> >that the vector type is supported would take place? With your change, 
> >it
> >allows vector types that are not supported to go through. Is it ok if 
> >I add
> >the check (test VECTOR_MODE_SUPPORTED_P) in tree.c:build_vector_type()?
> >
> 
> I'm not sure that is the right place, because we can generate scalar 
> code for unsupported vector types.
> You probably want to check for hardware support (IE 
> VECTOR_MODE_SUPPORTED_P) *before* calling build_vector_type.

Absolutely.  The code in tree.c should be able to build any 
representable type.


r~