[PATCH] Add vectorization of sqrt to the i386 backend

[Richard Guenther]

On Tue, 28 Nov 2006, Dorit Nuzman wrote:

> >
> > This adds vectorization capability of sqrt to the i386 backend and fixes
> > up the sqrt vectorization testcase from the pattern vectorization
> > enhacement.
> >
> ...
> > Index: testsuite/gcc.dg/vect/no-math-errno-vect-pow-1.c
> > ===================================================================
> > *** testsuite/gcc.dg/vect/no-math-errno-vect-pow-1.c   (revision 119115)
> > --- testsuite/gcc.dg/vect/no-math-errno-vect-pow-1.c   (working copy)
> > ***************
> > *** 1,5 ****
> > ! /* { dg-do compile } */
> > ! /* { dg-options "-O2 -ftree-vectorize -fno-math-errno -fdump-tree-
> > vect-details" } */
> >
> >   double x[256];
> >
> > --- 1,5 ----
> > ! /* { dg-do compile { target x86_64-*-* i?86-*-* } } */
> > ! /* { dg-require-effective-target vect_double } */
> >
> >   double x[256];
> >
> 
> I was hoping that restricting the testcase for 'vect_double' targets would
> suffice, along with xfailing the 'pattern recognized' test on targets that
> don't have a vector sqrt builtin yet.

Ok, that would have the benefit that we notice if it starts passing once
support for vectorized sqrt becomes more widespread.  Any idea of the
initial XFAIL target list?

Thanks,
Richard.

--
Richard Guenther <rguenther@suse.de>
Novell / SUSE Labs