Intel® oneAPI Math Kernel Library
Ask questions and share information with other developers who use Intel® Math Kernel Library.

Is there any service function that helps us form the needed matrix storage scheme?

zzzhhh
Beginner
316 Views

Although the various storage schemes are explained in detail in MKL Reference Manual, I believe human users are still willing to present their matrix in conventional full storage form. Is there any service functions in MKL that take in a full-storage matrix and convert it to other storage format like packed, band or RFP? That would reduce programming work a lot.

0 Kudos
4 Replies
mecej4
Honored Contributor III
316 Views

Now deprecated, but still available: https://software.intel.com/en-us/mkl-developer-reference-fortran-mkl-dnscsr

Recommended for use in new code: https://software.intel.com/en-us/mkl-developer-reference-fortran-matrix-manipulation-routines

0 Kudos
zzzhhh
Beginner
316 Views

Hi, mecej4, I don't understand your answer. Maybe you didn't understand my question at all.

0 Kudos
mecej4
Honored Contributor III
316 Views

Please read the MKL documentation using the links that I provided. The routine mkl_sdnscsr(), for example, allows you to provide a dense matrix as input to the routine and returns the arrays IA, JA and ACSR for the compressed sparse row representation of the same matrix. That capability is one of the features that you asked for, is it not?

0 Kudos
zzzhhh
Beginner
316 Views

mecej4 wrote:

Please read the MKL documentation using the links that I provided. The routine mkl_sdnscsr(), for example, allows you to provide a dense matrix as input to the routine and returns the arrays IA, JA and ACSR for the compressed sparse row representation of the same matrix. That capability is one of the features that you asked for, is it not?

Please read my question first before asking me to read something else. Your answer is totally irrelevant to my question because you totally have no idea what I was asking. Always try to UNDERSTAND what people were asking, rather than to ASSUME what people were asking. By the way, could you please not answer in this thread and leave it to someone else? Thanks a lot.

0 Kudos
Reply