On Friday 31 August 2007 20:12, Sergey Plis wrote:
> On Friday 31 August 2007 05:27:54 Bernd Paysan wrote:
> > Yes, found the bug, and fixed it. The new svn version should work
> > better. I've also changed a few calls in gsl.fs and atlas.fs, so
> > that it links ok to libblas and libgsl - I don't know if these
> > changes are correct, since I don't have an easy testcase for your
> > work.
>
> Just checked my previous example with compiling 1e5 - works now. It
> may have even saved me and eliminated the bugs I am having in
> conjunction with gsl bindings. Will check it later today.
>
> I am not sure about your changes to gsl.fs and atlas.fs. I have
> looked at atlas.fs - you've placed fortran lapack functions in place
> of C.

Yes, that's because my (or rather SuSE's) library doesn't contain the C 
functions.

> As I recall when I was writing these the ordering of parameters 
> is different for the two, so it matters how to call them now. I need
> to test the routines from gsl.fs which use lapack, but my prior is
> they will stop working, because of this order change.

Quite likely. However, in the long run it's better to only use the 
Fortran bindings, because they are always available.

> Thanks for fixing the bug!

Thanks for finding it.

-- 
Bernd Paysan
"If you want it done right, you have to do it yourself"
http://www.jwdt.com/~paysan/

Attachment: pgp3Uy4Rz6gxL.pgp
Description: PGP signature

Reply via email to