dolfin team mailing list archive

["Martin Sandve Alnæs" <martinal@xxxxxxxxx>]

2008/5/22 Anders Logg <logg@xxxxxxxxx>:
> On Mon, May 19, 2008 at 09:31:16AM +0200, Martin Sandve Alnæs wrote:
>> Regarding adding sparsity pattern information to an ufc::form, it seems to me
>> it should be part of the *integral classes, one
>> tabulate_sparsity_pattern to match
>> each tabulate_tensor, plus a num_nonzeros() function.
>
> I don't understand what tabulate_sparsity_pattern is supposed to do.
> Should it tabulate the full matrix sparsity pattern, or is it some
> kind of "equation sparsity pattern" tabulating which components of a
> system meet which components?
>
> --
> Anders

I mean tabulating which element tensor entries are
always zero independent of the cell and coefficients.
An "equation sparsity pattern", which may be different
for each *_integral class.

It will then be the task of DOLFIN to build a better global sparsity
pattern based on these local element sparsity patterns.

--
Martin