dolfin team mailing list archive

[A Navaei <axnavaei@xxxxxxxxxxxxxx>]

2009/3/6 Anders Logg <logg@xxxxxxxxx>:
> On Fri, Mar 06, 2009 at 03:40:46PM +0000, A Navaei wrote:
>> Since the backend support in dolfin comes with vectors and matrices,
>> corresponding to tensors of rank 1 and 2, I was wondering if there is
>> a standard efficient way of creating higher ranked tensors based on
>> this.
>>
>> For instance, if we are interested in creating sparse tensors of rank
>> 4, at least in MTL style, should we create a 2D matrix of 2D matrix
>> type? That is:
>>
>> mtl::Matrix< mtl::Matrix< double > >
>>
>> I don't expect this to be as efficient as a 4D matrix as this cannot
>> be converted to a 1D array pointer. Am I missing something here?
>
> There is no standard way to create higher ranked tensors, but the
> GenericTensor interface (and the assembler) supports it so if you
> should happen to have a data structure for rank 10 tensors, you can
> assemble forms of arity 10 into it.

It seems the interface:

virtual void GenericTensor::resize(uint rank, const uint* dims)

has not been implemented in dolfin for ranks higher than 2, is that
right? Are there any 3D matrices used in dolfin?


-Ali

>
> --
> Anders
>
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.9 (GNU/Linux)
>
> iEYEARECAAYFAkmxV+UACgkQTuwUCDsYZdET0ACdF8tT77wVI9pPcQFMnbCOMvTz
> xcsAmgM9Blpph070qlSoaBMXtRjFLFZg
> =UkD3
> -----END PGP SIGNATURE-----
>
> _______________________________________________
> DOLFIN-dev mailing list
> DOLFIN-dev@xxxxxxxxxx
> http://www.fenics.org/mailman/listinfo/dolfin-dev
>
>