RE: [RFC, vectorizer] Allow single element vector types for vector reduction operations

[Richard Biener <rguenther at suse dot de>]

On Mon, 28 Aug 2017, Jon Beniston wrote:

> Hi Richard,
> 
> >-  if (nunits < 1) /* Support V1SI.  */
> >+  if (nunits < 1 || (nunits == 1 && !reduct_p))
> >     return NULL_TREE;
> >
> >doesn't seem to be against trunk which has
> >
> >  if (nunits <= 1)
> >    return NULL_TREE;
> >
> >so what happens if you just change that to
> >
> >  if (nunits < 1)
> >    return NULL_TREE;
> 
> Ah, that's what I first tried, and mistakenly sent the patch against that.
> 
> That did help the initial problem, but then I needed to add a lot of support
> for the V1SI type in the backend (which just duplicated SI patterns) and
> there are a few places where GCC seems to have sanity checks against vectors
> that are only one element wide. A dot product producing a scalar result
> seems natural.

Where do you need V1SI types?  The vectorizer should perform a SI extract
from V1SI here.  You need to elaborate a bit here.

> 
> Also, as well as ARC benefitting from this, I think the TI c6x port would
> too. That currently has a sdot_prodv2hi pattern that uses SI and V2HI types.

The vectorizer doesn't really support vector->scalar ops so V1SI feels
more natural here.  That is, your patch looks a bit ugly -- there's
nothing wrong with V1SI vector types.  So maybe instead of adjusting
that function the respective caller when facing a lane-reducing op
such as DOT_PROD needs to consider scalar result types.

Richard.

> Cheers,
> Jon
> 
> 
> 

-- 
Richard Biener <rguenther@suse.de>
SUSE LINUX GmbH, GF: Felix Imendoerffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nuernberg)