This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Fwd: [patch] Fix PR19294
- From: ja2morri at csclub dot uwaterloo dot ca (James A. Morrison)
- To: Thomas Koenig <Thomas dot Koenig at online dot de>
- Cc: gcc-patches at gcc dot gnu dot org, fortran at gcc dot gnu dot org
- Date: 16 Jan 2005 08:27:57 -0500
- Subject: Re: Fwd: [patch] Fix PR19294
- References: <94DD311C-6786-11D9-B0FF-000A95D692F4@physics.uc.edu><20050116071230.GA4995@meiner.onlinehome.de>
Thomas Koenig <Thomas.Koenig@online.de> writes:
> > >From: ja2morri@csclub.uwaterloo.ca (James Morrison)
>
> > >@@ -1343,7 +1347,10 @@ gfc_resolve_transpose (gfc_expr * f, gfc
> > > switch (matrix->ts.type)
> > > {
> > > case BT_COMPLEX:
> > >- kind = matrix->ts.kind * 2;
> > >+ if (STRICT_ALIGNMENT)
> > >+ kind = 0;
> > >+ else
> > >+ kind = matrix->ts.kind * 2;
> > > break;
>
> What about machines with SLOW_ALIGNMENT? The code would for them,
> but probably with a performance penalty for using misaligned
> integers.
It probably is best to use gfortrans_transpose_[48] on those targets,
but I don't know. When kind is set to 0 then the generic byte by byte
transpose function is used. So for SLOW_ALIGNMENT targets copying a
larger amount is probably better.
> Thomas
>
--
Thanks,
Jim
http://www.student.cs.uwaterloo.ca/~ja2morri/
http://phython.blogspot.com
http://open.nit.ca/wiki/?page=jim