Data race in Pardiso Solver?

[danielsue]

Hi All,

I ran into a tricky problem when test data race for my codes. The Intel Inspector XE 2013 shows that there is data race in the following calling:

282    !C.. Factorization.
283          phase = 22 ! only factorization
284          CALL pardiso (pt, maxfct, mnum, mtype, phase, n, a, ia, ja,     &
285           idum, nrhs, iparm, msglvl, ddum, ddum, error)

But it can still solve the problem and the result seems correct. How does this data race come from?

I attached the testing codes and matrix data.

Thanks and regards,

Daniel

[danielsue]

Please find the data race summary and sources in the attached screen snip.

[Noah_C_Intel]

Hi Daniel, thanks for all of this information. I will do some research with your code and screenshots and get back to you ASAP.

[Noah_C_Intel]

Hi Daniel, I discussed this issue with both the MKL engineering and Intel Inspector teams and it appears the data race is benign. This 'false positive' should go away in the next release of Parallel Studio (which includes inspector and all other updates of its components). Please let me know if you get this message in the future after the release, and let me know if you have any other questions.

[danielsue]

Noah Clemons (Intel) wrote:

Hi Daniel, I discussed this issue with both the MKL engineering and Intel Inspector teams and it appears the data race is benign. This 'false positive' should go away in the next release of Parallel Studio (which includes inspector and all other updates of its components). Please let me know if you get this message in the future after the release, and let me know if you have any other questions.

Hi Noah,

Thanks. This data race should be benign as the solver can solve most of my equations very well. But I do have some matrice that the results from pardiso are not correct. I am not sure if this is caused by data race problem before. These matrice are from a flow problem, I can solve these matrice with another solver that has been used over ten years in our model. The result produced by Pardiso is quite different what I guess is not correct.

If you're interested in it, please look into the attached files, where the spase matrice (ia, ja, a, b) and results (x, x_pardiso_ws209) are provided. The number in the first line indicates the number of values included. The file x_pardiso_ws209.txt includes the results from pardiso and another solver (named ws209).

Thanks and regards,

Daniel