This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [lno][patch] vectorizer updates
- From: Dorit Naishlos <DORIT at il dot ibm dot com>
- To: Richard Henderson <rth at redhat dot com>
- Cc: gcc-patches at gcc dot gnu dot org, Pop Sébastian <pop at gauvain dot u-strasbg dot fr>
- Date: Thu, 15 Jan 2004 12:34:52 +0200
- Subject: Re: [lno][patch] vectorizer updates
> /* { dg-do compile { target powerpc*-*-* i?86-*-* } } */
> /* { dg-options "-O2 -ftree-vectorize -fdump-tree-vect-stats -maltivec" {
target powerpc*-*-* } } */
> /* { dg-options "-O2 -ftree-vectorize -fdump-tree-vect-stats -msse" {
target i?86-*-* } } */
thanks
> Further, I really think you should break these into separate tests.
> Checking that 4 of 12 loops are vectorized does not mean that the same
> 4 loops are vectorized all the time.
You are right, but there's also value in testing multiple loops together,
at least until the monev analyzer and the vectorizer and the interaction
between them stabilize. I will also check in single loop tests to verify
that the loops that get vectorized are only those that are expected to.
thanks,
dorit
Richard Henderson
<rth@redhat.com> To: Dorit Naishlos/Haifa/IBM@IBMIL
cc: gcc-patches@gcc.gnu.org, Pop Sébastian <pop@gauvain.u-strasbg.fr>
15/01/2004 09:01 Subject: Re: [lno][patch] vectorizer updates
On Wed, Jan 14, 2004 at 07:14:08PM +0200, Dorit Naishlos wrote:
> /* { dg-do compile { target powerpc*-*-darwin* } } */
> /* { dg-options "-O2 -ftree-vectorize -fdump-tree-vect-stats -maltivec" }
> */
>
> The sse equivalent would be...?
/* { dg-do compile { target powerpc*-*-* i?86-*-* } } */
/* { dg-options "-O2 -ftree-vectorize -fdump-tree-vect-stats -maltivec" {
target powerpc*-*-* } } */
/* { dg-options "-O2 -ftree-vectorize -fdump-tree-vect-stats -msse" {
target i?86-*-* } } */
For *compile* tests, as these are, you shouldn't be restricting to just
darwin; with the right options I'd expect any ppc target should pass.
For *run* tests, you'll have to add runtime checks for x86 to verify
that the appropriate support is present at runtime. We might be able
to hack dejagnu to test this for us once; not sure.
Further, I really think you should break these into separate tests.
Checking that 4 of 12 loops are vectorized does not mean that the same
4 loops are vectorized all the time.
r~