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

PARDISO-CSR format

Reza2021
Beginner
384 Views

Hi Everyone,

I wanted to use the PARDISO solver for my finite element Fortran code. I found the sample code (pardiso_sym_f90) from MKL library examples, and I was able to run the sample code. However, I could not find the routine to convert my matrix (unsymmetric banded matrix) to CSR format to use in PARDISO. Would you please advise me about that? Thank you.  Reza

0 Kudos
1 Solution
Kirill_V_Intel
Employee
363 Views

Hello Reza!

To answer directly your question: AFAIK, we don't have such a routine in oneMKL (but one can write some relatively simple code to do that).

I'd like to ask a question though and I think you should have a good answer before proceeding with PARDISO:
If you already have your matrix in the banded format, why don't you use a linear solver which is specialized for banded matrices?
E.g. you can find one here in LAPACK:
https://software.intel.com/content/www/us/en/develop/tools/oneapi/components/onemkl/function-finding...
if you choose general + band storage matrix type.

The power of PARDISO is that it is a direct sparse linear solver which works for sparse matrices with general sparsity patterns. While it can still perform quite well on matrices with simple regular patterns (like banded), there are specialized solvers for banded matrices which should work better (because they are specialized for this kind of matrices). 

Best,
Kirill

 

 

View solution in original post

4 Replies
Kirill_V_Intel
Employee
364 Views

Hello Reza!

To answer directly your question: AFAIK, we don't have such a routine in oneMKL (but one can write some relatively simple code to do that).

I'd like to ask a question though and I think you should have a good answer before proceeding with PARDISO:
If you already have your matrix in the banded format, why don't you use a linear solver which is specialized for banded matrices?
E.g. you can find one here in LAPACK:
https://software.intel.com/content/www/us/en/develop/tools/oneapi/components/onemkl/function-finding...
if you choose general + band storage matrix type.

The power of PARDISO is that it is a direct sparse linear solver which works for sparse matrices with general sparsity patterns. While it can still perform quite well on matrices with simple regular patterns (like banded), there are specialized solvers for banded matrices which should work better (because they are specialized for this kind of matrices). 

Best,
Kirill

 

 

Reza2021
Beginner
357 Views

Hi Kirill,

Thank you for getting back to me and for your advice. 

That’s right. I am using LAPACK for my code. I just wanted to try PARDISO if I can reduce the computational time on my program.

Best Regards,

Reza

VidyalathaB_Intel
Moderator
331 Views

Hi Reza,

As your query has been answered.

If you don't have any other issues let us know if we can close the thread.

Regards,

Vidya


VidyalathaB_Intel
Moderator
310 Views

Hi Reza,

Thanks for the confirmation!

As this issue has been resolved, we will no longer respond to this thread. 

If you require any additional assistance from Intel, please start a new thread. 

Any further interaction in this thread will be considered community only. 

Have a Good day.

Regards,

Vidya


Reply