From: Christian Stimming <stimming@tu...> - 2010-11-06 22:12:00
please ask on the mailing list lapackpp-devel@... (already
in CC). I personally haven't been using lapackpp for quite some time by now.
As for large matrices and LAPACK (the underlying one, without pp): No, I have
no information about its implicit size limits. Your 5000x5000 double example
corresponds to approx. 200 MByte of memory, which sounds still reasonable for
today's amounts of RAM that people have. I thought people have reported before
that somewhat larger matrices also work, but I'm not sure. I haven't worked
with such large matrices before.
Also, I have no information about debugging aids in LAPACK. Sorry for that.
Am Donnerstag, 4. November 2010 schrieb ddupuis@...:
> Hi Mr. Stimming,
> I am a software engineer, and I have been using Lapack++ to solve large
> linear systems for my work, but have been finding that some operations
> fail (basic addition, multiplication, and calls to "LaLinearSolve") if I
> use matrices that are too large (on the order of 5000x5000 doubles, give
> or take). I found your email off of the Lapack++ documentation website,
> so I thought I'd contact you about the problem.
> I was wondering if you knew where I could find more information about what
> the hard matrix size limits are, and if there are any work-arounds for
> this. I would look into the memory limitations of the underlying BLAS
> library, but it wasn't clear to me which functions from this library were
> being called. Perhaps there is some discussion group or thread focusing on
> this topic?
> Also, does Lapack have any error/debugging support so I can determine
> details about thrown exceptions, etc. ?
> Thanks very much for your help!
> Donna Dupuis
> Software Development Engineer
> Vorum Research Corporation
> Vancouver, BC