|
|
Glas :Re: [glas] dense and sparse vectors |
From: Richard Vuduc (richie_at_[hidden])
Date: 2005-08-08 01:07:41
Toon Knapen wrote:
> I was mainly planning to benchmark 'sparse axpy' of different libraries
>
>(I'm currently using mtl-v2.1.22, oski-1.0 and ublas). I indeed
>mentioned 'dense' also but mainly because I also plan to benchmark the
>sparse axpy against the dense versions.
>
>
Toon,
Just so you know, OSKI does not provide a sparse axpy (or any
sparse-times-sparse kernels), so this should save you some time. :-)
OSKI only provides variations of sparse matrix-times-dense vector/matrix
kernels (including triangular solve, transpose(A)*A*x, and A^k*x), since
these are the bottlenecks in applications in which we are most
interested currently.
[Aside: Of course, that's not to say we think that sparse-times-sparse
kernels are not important. Obviously a great potential strength of the
generic approach is the ability to express and generate implementations
of this class of kernels easily. An interesting future target for the
GLAS activity might be the frequent multigrid bottleneck kernel, the
sparse matrix triple product (computing B = R*A*transpose(R), where A
and R are sparse).]
--rich