[Boost-users] Re: Odd uBLAS limitation

From: Angus Leeming (angus.leeming_at_[hidden])
Date: 2004-10-29 06:01:38

ajb_at_[hidden] wrote:

> G'day all.
> 
> Looks like the uBLAS questions are going here, since we can't post to
> the proper list.  So might as well ask here.
> 
> I'm on a 32-bit platform, and I can't create a sparse matrix of
> dimensions 65536*65536.  I get a run-time assertion failure, because
> uBLAS requires that for an m by n matrix, m*n fits in a machine word.
> 
> For a dense matrix, that's an understandable limitation, but for a
> sparse matrix it's impractically small.
> 
> I'm trying to use uBLAS to solve finite element Poisson problems with
> odd boundary conditions, which involves solving a ver sparse linear
> system with the grid points as unknowns.  This limitation means I can't
> use anything larger than a 40*40*40 voxel grid.
> 
> Any thoughts?

I believe that this limitation has been removed in the
to-be-released-imminently 1.32 version.

See
http://article.gmane.org/gmane.comp.lib.boost.ublas:1460
which talks of a 1000000x1000000 matrix.

To answer the original poster, the Ublas mailing list can be browsed here.
I'd say that there are 4 or 5 active developers ATM.

http://news.gmane.org/gmane.comp.lib.boost.ublas

Regards,
Angus