Can MKL perform matrix-vector multiply or matrix-matrix multiply for sparse matrix in variable block row (vbr) format?
I have a matrix, in sparse, variable block row format, (i.e. vbr) that I need to use the MKL to operate on. Since the off diagonal blocks are not square, Ibelieve that I cannotuse the block sparse row(bsr) that is providedin MKL.
1. Is there a way around this? That is, can Isomehow 'dummy up' the matricesso that I can use MKL?
2. Is my only choice to use the National Institute of Standards, NIST "vbr" subroutines?
3.Are there plans to include the "vbr" format into MKL?