This is the mail archive of the
gsl-discuss@sources.redhat.com
mailing list for the GSL project.
Re: GSL and CLAPACK together
- From: Brian Gough <bjg at network-theory dot co dot uk>
- To: Andrew Fredricks <fredra at comcast dot net>
- Cc: gsl-discuss at sources dot redhat dot com
- Date: Mon, 24 Nov 2003 16:05:09 +0000
- Subject: Re: GSL and CLAPACK together
- References: <1069253453.10485.10.camel@Taurus>
Andrew Fredricks writes:
> Hello,
>
> I have recently been using GSL and CLAPACK together on a project and
> was happy to see they work easily. But it was not obvious that they
> would. I suggest that there be a place in documentation that list hints
> or trick to doing things like that.
Hi,
Thanks for your email.
I've avoided mentioning calling LAPACK in the past, for the following
reason:
When the problem only involves one square matrix, and the result is a
vector, it's straightforward to call LAPACK, but with several matrices
it's more complicated. The transposes have to be done out of
place. And then result matrices have to be transposed back. The call
has to be done with the appropriate lda/tda values. So it gets messy.
>From time to time I think about writing a wrapper to LAPACK to handle
it all, but it becomes very inelegant (unlike wrapping BLAS).
The general problem is hard, but there are some cases (like solving
Ax=b) which could be documented as an example.
It's unfortunate that LAPACK doesn't support transpose arguments, like
BLAS, as an alternative to physically transposing the matrix.
--
Brian