I found "mkl_csrmultcsr" but as I understand both A and B must include full storage of data (upper + lower diagonal terms) whereas I have A & B and only upper diagonals stored.
I do not see the function for such computation. I can create a feature in the internal database, so our engineer team can check it in the future release.
That would be great if you include it the list of developments.
For a typical finite element application, we usually deal with symmetric matrices which saves us a lot of memory and computation time. In such an application (Eigen sub-space solution), I need to perform matrix multiplication of 2 sym. sparse matrices.
For this feature request, we have some questions:
"By computing the product of symmetric sparse matrices and getting SYMMETRIC SPARSE result"
In many cases, such a product is non-symmetric, and dense. Does you some mean some specific matrices to be multiplied?