topic Re:Issues with MKL PARDISO in Intel® oneAPI Math Kernel Library

Issues with MKL PARDISO

SergP — Thu, 08 Jul 2021 18:19:42 GMT

Hi,

We have found several issues with MKL PARDISO for improved two-level factorization algorithm (iparm[23]=10), namely:

1. CNR functionality works incorrectly, solutions could be incorrect;

2. pardiso_getdiag - segmentation faulted in some cases;

3. user permutation vector (iparm[4]) works incorrectly in multi-threading mode.

MKL information:

Linux

Major version: 2021

Minor version: 0

Update version: 1

Build: 20201104

We have prepared reproducers with specific matrices, how can we share them for confirmation of the issues?

Regards,

Sergey

Re: Issues with MKL PARDISO

Gennady_F_Intel — Thu, 08 Jul 2021 19:07:06 GMT

Sergey,

You could attach this case to this thread or you may send us this reproducer via the private communication.

Re: Issues with MKL PARDISO

SergP — Mon, 12 Jul 2021 17:02:52 GMT

Gennady,

I'd prefer to use private communications. Could you please send me instructions on how to do it?

Sergey

Re:Issues with MKL PARDISO

Gennady_F_Intel — Tue, 13 Jul 2021 14:36:56 GMT

Sergey,

All of the issues you reported were reproduced with MKL 2021.3 and will be escalated against the Sparse Solver Team. We will keep this thread informed of the status of these issues.

Thanks,

Gennady

Re: Re:Issues with MKL PARDISO

SergP — Sat, 23 Oct 2021 05:55:49 GMT

Hi Gennady,

Am I right that

MKLD-11665

Fixed the incorrect result produced by MKL PARDISO when a permutation vectore is used

is not related to issues, reported by me?

Thanks,

Sergey

Re:Issues with MKL PARDISO

Gennady_F_Intel — Sun, 24 Oct 2021 05:25:22 GMT

Hello Sergey,

it is very similar but not exactly. You might check if this update somehow fixes this particular problem.

-Gennady

Re: Issues with MKL PARDISO

SergP — Thu, 16 Dec 2021 15:42:19 GMT

Hi Gennady,

We have checked this update but our issues in it are still reproduced.

Do you have any estimation of the time when they will be fixed?

Regards,

Sergey

Re:Issues with MKL PARDISO

Chao_Y_Intel — Fri, 28 Mar 2025 07:17:22 GMT

The issue was resolved from the MKL 2024.0 release. Thank you for bringing it to our attention. In the latest MKL release, we have made several improvements to the performance and stability of the Pardiso APIs. Please feel free to contact us if you have any questions.