This is the mail archive of the
fortran@gcc.gnu.org
mailing list for the GNU Fortran project.
Re: gfortran.dg/vect failures at -m64 on darwin
- From: Dorit Nuzman <DORIT at il dot ibm dot com>
- To: dominiq at lps dot ens dot fr (Dominique Dhumieres)
- Cc: bromo dot msbb dot uc dot edu at lps dot ens dot fr, dominiq at lps dot ens dot fr, fortran at gcc dot gnu dot org
- Date: Sat, 4 Aug 2007 00:56:52 +0300
- Subject: Re: gfortran.dg/vect failures at -m64 on darwin
dominiq@lps.ens.fr (Dominique Dhumieres) wrote on 04/08/2007 00:27:24:
> > Do you see the following in the dump file (when compiled with
> > -fdump-tree-vect-details):
> > "vector alignment may not be reachable"
>
> I don't see such a line.
>
I think this is because you probably added "-ftree-vect-details" on top of
the default test flags "-fdump-tre-vect-stats -ftree-vectorizer-verbose=4"
(it should be instead, sorry about that). Anyhow, in the meantime Jack
Howarth sent me a dump file that seems to confirm my suspicion. I'll work
on a patch for that.
thanks,
dorit
> > and
> > "Alignment of access forced using versioning."?
>
> I do:
>
> ...
> (res = {{1, +, 1}_1, +, 1}_2))
> (get_loop_exit_condition
> if (j_3 == pretmp.57_24))
>
> /opt/gcc/gcc-4.3-work/gcc/testsuite/gfortran.dg/vect/vect-5.f90:14:
> note: Alignment of access forced using versioning.
> /opt/gcc/gcc-4.3-work/gcc/testsuite/gfortran.dg/vect/vect-5.f90:14:
> note: Vectorizing an unaligned access.(analyze_scalar_evolution
> (loop_nb = 2)
> (scalar = j_3)
> (get_scalar_evolution
> (scalar = j_3)
> (scalar_evolution = {i_1, +, 1}_2))
> (set_scalar_evolution
> (scalar = j_3)
> (scalar_evolution = {i_1, +, 1}_2))
> )
> ..
>
> Do you need more?
>
> Dominique