This is the mail archive of the
fortran@gcc.gnu.org
mailing list for the GNU Fortran project.
Re: Re: Polyhedron 2005 Benchmarks
- From: "Timothy C Prince" <tprince at myrealbox dot com>
- To: dberlin at dberlin dot org
- Cc: stevenb at suse dot de, dominiq at lps dot ens dot fr, fortran at gcc dot gnu dot org
- Date: Fri, 18 Nov 2005 18:38:59 +0000
- Subject: Re: Re: Polyhedron 2005 Benchmarks
-----Original Message-----
From: Daniel Berlin <dberlin@dberlin.org>
To: Steven Bosscher <stevenb@suse.de>
Date: Fri, 18 Nov 2005 09:10:15 -0500
Subject: Re: Polyhedron 2005 Benchmarks
On Fri, 2005-11-18 at 14:32 +0100, Steven Bosscher wrote:
> On Nov 18, 2005 01:57 PM, Dominique Dhumieres <dominiq@lps.ens.fr>
> wrote:
> > g95 and gfc are unable to reorder loops such as
> >
> > do i = 1, ni
> > do j = 1, nj
> > a(i,j) = b(i,j)
> > end do
> > end do
>
> Shouldn't -ftree-loop-linear perform the loop interchange here?
Probably :)
I'd be curious to know if it decides not to, and why.
-fdump-tree-ltrans-all
>
> Gr.
> Steven
>
>
>
For such a simple loop, it would be more interesting to see whether it is optimized in the form
a(:ni,:nj) = b(:ni,:nj)
Even ifort is not generally good at vectorization of 2D assignment.
Tim Prince