Search found 90 matches

by hartwig anzt
Thu Nov 15, 2018 12:08 pm
Forum: User discussion
Topic: Matrix overflow when using PARILU
Replies: 5
Views: 895

Re: Matrix overflow when using PARILU

Salmah, the ParILU preconditioner can result in overflow. This is due to the limited floating point range. I.e. dividing by a very small number (resp. 0) can result in overflow. In theory, this does not happen, but with limited precision it can. For very ill-conditioned matrices. We are working on p...
by hartwig anzt
Thu Nov 08, 2018 4:29 pm
Forum: User discussion
Topic: Matrix overflow when using PARILU
Replies: 5
Views: 895

Re: Matrix overflow when using PARILU

Dear Salmah, Sorry for the delay, but I finally found time to look into this. It turns out that you chose a very tough problem. Even though the condition number is about 10^5, iterative methods have a hard time solving this. Also, this is not a positive definite matrix, so you won't be able to use C...
by hartwig anzt
Thu Nov 08, 2018 3:16 pm
Forum: User discussion
Topic: Null pointer exception when solving
Replies: 2
Views: 425

Re: Null pointer exception when solving

Dario, from your codelet, I can not see how the data of your matrix looks like. But I would assume all the data is on the host? Then you may have to copy it to the GPU before invoking the solve... > magma_dmtransfer(A, dA, Magma_CPU, Magma_DEV, queue); Same for the vector. Let me know whether this h...
by hartwig anzt
Thu Nov 01, 2018 1:34 pm
Forum: User discussion
Topic: testing magma_dlsqr
Replies: 2
Views: 435

Re: testing magma_dlsqr

Dario,
check out magma/example/example_sparse.c
Hartwig
by hartwig anzt
Thu Oct 25, 2018 9:23 am
Forum: User discussion
Topic: Matrix overflow when using PARILU
Replies: 5
Views: 895

Re: Matrix overflow when using PARILU

Dear Salmah,

thanks for reaching out! I am happy to look into this. Would you mind sharing your problem (matrix) and maybe also the complete test routine in which you are calling the routine?

Just send it to hanzt@icl.utk.edu.

Thanks, Hartwig
by hartwig anzt
Tue Jul 31, 2018 1:19 pm
Forum: User discussion
Topic: Why is Magma skipping calculation when using SELL-P format?
Replies: 3
Views: 430

Re: Why is Magma skipping calculation when using SELL-P form

Dear Mohammed,

would you mind sharing the complete code along with the test matrices? Or is this difficult?

@hanzt@icl.utk.edu

Thanks, Hartwig
by hartwig anzt
Tue Jul 24, 2018 5:17 am
Forum: User discussion
Topic: Options for a fast PCG-solve? (currently slower than CPU)
Replies: 1
Views: 425

Re: Options for a fast PCG-solve? (currently slower than CPU

Dear Martin, this is a fairly small matrix. MAGMA assigns one thread to each row, this is 4120 threads, 32 thread blocks of size 128. In the end, only a fraction of the GPU can be used, and the overhead of transferring the data to the GPU can not be compensated. MAGMA-sparse is designed for large-sc...
by hartwig anzt
Fri Jul 13, 2018 2:42 am
Forum: User discussion
Topic: PARILUT preconditioner causes SEGFAULT
Replies: 1
Views: 382

Re: PARILUT preconditioner causes SEGFAULT

This is not a surprise. ParILUT is a threshold-ILU preconditioner. The ParILUT considers the level-fill as potential fill in candidates. But if you have a diagonal matrix, there is no level-fill, and there are no locations that are reasonable to consider as fill-in. Depending on your other test matr...
by hartwig anzt
Thu Jul 12, 2018 3:34 am
Forum: User discussion
Topic: Magma Example gives wrong results
Replies: 3
Views: 509

Re: Magma Example gives wrong results

Dear Martin, Thanks for reaching out! I can reproduce the problem, however, I do not have a solution yet. As far as I can see, it is a memory-ownership issue. I am working on it. In the meanwhile, I want to point you to a different collaborative project that is aiming for a sustainable C++-based NLA...
by hartwig anzt
Sat Feb 10, 2018 7:27 pm
Forum: User discussion
Topic: MAGMA 2.3.0 released
Replies: 2
Views: 3659

Re: MAGMA 2.3.0 released

Thank you very much, I updated the repository.