sparse blas matrix-matrix product

[lanzors]

Hi,

We are interested in a commercail product with the sparse blas 2 and 3 functionnalities.
By reading the doc and I was told that the sprase blas 2 and 3 funcions are only available in the double precision case. I want to know whether and when can we have a single precision version.

Can anyone tell me if there are some benchmark datas of the MKL sparse functions. I've searched in internet and nothing interested found yet.

Thx for any informations.

[Gennady_F_Intel · Accepted solution]

1. The latest MKL v.10.1 is available right now and the functionality you are looking for is available

as for sparse blas routine 2 and 3.

You can find these declarations in mkl_spblas.fi mkl_spblas.h.

For example:

mkl_scsrmv&transa, &m, &k, α, matdescra, val, indx, pntrb, pntre, x, β, y);

mkl_ccsrmv(&transa, &m, &k, α, matdescra, val, indx, pntrb, pntre, x, β, y);

2. There are no some benchmark data available right now.

--Gennady