This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Inline DOT_PRODUCT revisited - to libgfortran/m4 gurus


On Sun, Feb 26, 2006 at 04:19:13PM +0100, Paul Thomas wrote:
> >See the original message from November 2005 here:
> >http://gcc.gnu.org/ml/gcc-patches/2005-11/msg00686.html
> >
> >On Thursday 10 November 2005 09:13, Paul Thomas wrote:
> >
> >>an inline version of DOT_PRODUCT, which is never slower that the
> >>original library version and is very much faster for small vectors.

I looked at this patch yesterday, with the plan to integrate it with
the calling BLAS thing. I agree that it should be committed as it
seems to work, a few questions tough.

1) What's the problem with kind > 8? I removed the 

if (expr->ts.kind > 8)
+     {
+       gfc_conv_intrinsic_funcall (se, expr);
+       return;
+     }

thing and it seems to work for kind=10 (work, as in your testcase
modified for kind=10 ran without crashing). Tough I didn't make a
testcase to check whether the results were sensible, that should
certainly be made.

2) What about dot_product for logicals? The patch doesn't seem to
address this? But I don't think it's worth worrying about the
performance of this, I don't know of any real usage of this (if I'm
wrong please correct me). So the easiest thing would be to use the
existing library implementation.

-- 
Janne Blomqvist

Attachment: pgp00000.pgp
Description: PGP signature


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]