topic Parallel Direct Sparse Solver for Clusters and iparm[30] in Intel® oneAPI Math Kernel Library

Parallel Direct Sparse Solver for Clusters and iparm[30]

Marius_B_ — Sun, 03 May 2020 06:55:42 GMT

Hi,

Does Parallel Direct Sparse Solver for Clusters support iparm[30] for Partial solve and computing selected components of the solution vectors?

It is mentioned in the reference manual , but it also says "iparm[30] - iparm[32] Reserved. Set to zero."

Non-MPI MKL PARDISO works for iparm[30] 1,2, and 3 but the cluster edition doesn't ? Is this a bug or is iparm[30] simply not supposed to work with the cluster edition?

best,

marius

Hello marius,

Kirill_V_Intel — Mon, 04 May 2020 00:32:15 GMT

Hello marius,

It's a bug in the documentation, it should be "iparm[31]-iparm[33]" for "Reserved. Set to zero". And there is a description for iparm[30] which enables partial solve for the cluster sparse solver. Thanks for noticing it!

Let us know if you encounter any issue when using partial solve with the cluster sparse solver.

Best,
Kirill

Hi Kirill

Marius_B_ — Mon, 04 May 2020 06:08:55 GMT

Hi Kirill

Thanks for the swift reply. I tried the partial solve with the cluster sparse solver but it does not work. It only works for 1 MPI process but for 2 or more it does give the full solution.

I attached you my sample program, it is actually just he cl_solver_complex_unsym_f.f example with the corresponding options set.

Best,

Marius

Hello Marius,

Kirill_V_Intel — Mon, 04 May 2020 17:47:57 GMT

Hello Marius,

I'm afraid I misled you. Actually, it turned out we don't support partial solve in the Cluster Sparse Solver, so we'll need to fix the documentation (not in the way I originally mentioned). Well, it's never late to learn:)

So, if you need this feature, you can submit a feature request to our online customer center (I hope someone will provide a link there since I don't have it).

Best,
Kirill

Thanks for letting me, know.

Marius_B_ — Tue, 12 May 2020 23:59:29 GMT

Thanks for letting me, know. Too bad as this is a really useful feature. I will try to submit a feature request.