topic Re:About threading performance of PARDISO on Xeon CPU in Intel® oneAPI Math Kernel Library

About threading performance of PARDISO on Xeon CPU

DreamlikeCastle — Fri, 03 Nov 2023 01:16:33 GMT

Hi, I am curious about the threading performance of PARDISO on the platform of a lot of CPU cores.

For instance, the PARDISO solver finishes in about 20 seconds to finish on my Xeon 8480 CPU with more than 20 MKL threads, no matter 20, 40, or 50 threads.

Is it limited by RAM bandwidth or other aspects? Could you give any advice on how to continue to improve threading performance? Does the MPI version of the solver behave better?

To use the attached VISUAL STUDIO example, please unzip all 4 files and put the files in one folder.

Thanks in advance.

Re: About threading performance of PARDISO on Xeon CPU

VarshaS_Intel — Tue, 07 Nov 2023 16:17:59 GMT

Hi,

Thanks for posting in Intel Communities.

Could you please let us know the OS, hardware, CPU details, and Intel MKL and Fortran compiler you are using?

Could you please provide us with the complete output you getting while running the code?

>>Could you give any advice on how to continue to improve threading performance?

Please find the below link for improving the threading performance:

https://www.intel.com/content/www/us/en/docs/onemkl/developer-guide-macos/2023-2/improving-performance-with-threading.html

Thanks & Regards,

Varsha

Re:About threading performance of PARDISO on Xeon CPU

VarshaS_Intel — Wed, 15 Nov 2023 13:06:39 GMT

Hi,

We have not heard back from you. Could you please provide us with an update on your issue?

Thanks & Regards,

Varsha

Re:About threading performance of PARDISO on Xeon CPU

VarshaS_Intel — Wed, 22 Nov 2023 10:57:10 GMT

Hi,

We have not heard back from you. Could you please provide us with an update on your issue?

Thanks & Regards,

Varsha