I'm working with a very non-equilibrated sparse linear system, which in MATLAB I treat with a function probably similar to dgeequ. Question: is there anything like this for MKL Sparse BLAS? Or something doing something similar dealing with preconditioners?
In case dgeequ or similar is not available, I have a second question: to implement dgeequ I need to find the minimum/maximum element of a row and/or column of a sparse matrix. Is it possible to do it in MKL working directly with the matrix handle? Otherwise, my best solution would be to export in the three-array format the matrix and work on the vectors...
Currently, we don't support such functionality in IE SpBLAS. So I think the easiest way to do it is to export data and work on raw pointers. You can submit a feature request for us to add the functionality like dgeequ.
Khang, could you give an appropriate link to a place where feature requests are created for MKL?
Thank you for providing the answer to this question.
Why don't you sent me an email at email@example.com describing the requested feature in detail along with the OS, data types and data sizes preferences?