Enable use of MKL for sparse matrices

[jalvesz]

Motivation

Using optimized BLAS/LAPACK libaries through stdlib for dense linear algebra is now very easy. Sparse matrices do not count with such a level of API standardization. stdlib counts now with a minimal support for sparse matrices but it would be useful to extended in several ways one of such would be to enable using MKL as an optimized backend.

A first step would be to enable switching between the internal spmv and mkl spmv kernels. This would require additional specifications at the level of the preprocessor macros because the sparse matrix API of MKL has a specific signature and the current STDLIB_EXTERNAL_BLAS/LAPACK macros will not be enough.

Prior Art

No response

Additional Information

No response

[demoncoder-crypto]

I would love to work on this issue. is this open?

[jalvesz]

Yes it is! Please go ahead and take a look 👍