topic Pardiso format Vector Manipulation Library in MKL in Intel® oneAPI Math Kernel Library

Pardiso format Vector Manipulation Library in MKL

Ngu_Soon_Hui — Fri, 29 Jan 2010 04:38:21 GMT

This is because for band matrix, lapack has readily available matrix/vector manipulation library (see for example, sgbtrf and sgbmv), which comes in handy when one needs to do ARPACK eigen solvers ( see an example here). However, a lot of times we don't use band matrix storage format and instead, we opt for pardiso format ( is it CSR or CSC or a variant of any of the two? I'm not very sure).

A similar library for Pardiso would be great aid for everyone.

Pardiso format Vector Manipulation Library in MKL

ArturGuzik — Fri, 29 Jan 2010 05:24:49 GMT

I'm not entirely sure what you have in mind, however, you may take a look at the Sparse BLAS level 2 and 3 routines (in MKL), which provide (most probably) what you want, including all sort of multiplications, conversions etc.

Pardiso format Vector Manipulation Library in MKL

Gennady_F_Intel — Sat, 30 Jan 2010 08:12:10 GMT

Quoting Ngu Soon Hui

I just wonder whether there is any vector/matrix manipulation library in MKL if my data is stored in pardiso format? The library shall include, but not limited to factorization, linear system solver, matrix vector product.

This is because for band matrix, lapack has readily available matrix/vector manipulation library (see for example, sgbtrf and sgbmv), which comes in handy when one needs to do ARPACK eigen solvers ( see an example here). However, a lot of times we don't use band matrix storage format and instead, we opt for pardiso format ( is it CSR or CSC or a variant of any of the two? I'm not very sure).

A similar library for Pardiso would be great aid for everyone.

one more comment on this - Pardiso supports only one CSR format ( 1-based indexes).

--Gennady