pardiso factorization speed .......... outdated pardiso version in mkl

[may_ka]

Dear Intel team,

several months ago I started this thread about problems with the pardiso version currently used by MKL when trying to factorize very large matrices. I also pointed out that the problems are that severe that MKL-pardiso is rendered useless. I addition I provided a fully operational example including data.

Unfortunately I have not heard anything from Intel since March. Do you have any update on the issue???

Further I could obtain trial version of the latest pardiso library and found that all the problems I had pointed out are solved in that version. Therefore I am wondering whether Intel has any plans for updating the current MKL-pardiso to maintain MKL's cutting edge performance??

It would be appreciated if Intel could clarify it's position to both question as it would help developers to decide whether MKL is(will be again) a viable option for operations on very large sparse matrices.

Thanks

[Khang_N_Intel]

Hi Karl,

I just replied to you in the other thread.

You mentioned that you have obtained a trial version of oneMKL and that trial version solved all the issue that you mentioned in the other threads.

Can you tell me the version number of that oneMKL?

As usual, at Intel, our mkl team is constantly finding ways to improve the performance as well the stability of our libraries.

Best regards,

Khang

[may_ka]

Hi,

I did not write that I obtained onemkl test version, I wrote that I obtained a pardiso test version. That test version was 7.2.

Best

[Khang_N_Intel · Accepted solution]

Thanks Karl!

As I have mentioned in the previous message, our mkl team is constantly looking for ways make our library better and better.

For more information about oneMKL new features or performance improvement, please look at oneMKL release notes:

https://www.intel.com/content/www/us/en/developer/articles/release-notes/onemkl-release-notes.html

Since the questions have been answered, there will be no further communication on this thread. Should you have any oneMKL-related questions, please open new threads.

Best regards,

Khang