Search found 264 matches

by Stan Tomov
Fri Dec 28, 2018 1:34 pm
Forum: User discussion
Topic: (d/s)potrf_batched has some kind of memory leak
Replies: 2
Views: 674

Re: (d/s)potrf_batched has some kind of memory leak

Thank you for reporting it.
The leak has been fixed you can update from bitbucket.
by Stan Tomov
Tue Oct 16, 2018 3:15 pm
Forum: User discussion
Topic: Create distributed matrix on gpus with no cpu to gpu copy
Replies: 9
Views: 1793

Re: Create distributed matrix on gpus with no cpu to gpu cop

The routine described is actually in MAGMA, called dgegqr_gpu version 4, along with a few other versions also described there. You can test them and see how they are called, e.g., with ./testing_dgegqr_gpu --version 4 -N 10000,64 -c The implementation itself is very simple using the magma building b...
by Stan Tomov
Sun Oct 14, 2018 11:53 pm
Forum: User discussion
Topic: Create distributed matrix on gpus with no cpu to gpu copy
Replies: 9
Views: 1793

Re: Create distributed matrix on gpus with no cpu to gpu cop

The magma_dpotrf_mgpu function requires that the matrix is distributed among the GPUs in 1D block cyclic way, where nb is obtained by magma_get_dpotrf_nb(n). The magma_dsetmatrix_1D_col_bcyclic function is just one example on how one can get to this distribution starting from CPU memory. If you alre...
by Stan Tomov
Fri Oct 12, 2018 5:46 pm
Forum: User discussion
Topic: Matrix formats for lobpcg
Replies: 4
Views: 1216

Re: Matrix formats for lobpcg

We can easily enable SpMM for any sparse matrix, e.g., just do SpMV for each vector. However, special optimizations for SpMM can do much better than SpMV. We have done this only for the Magma_SELLP format, and have disable the others - just as a reminder to add them (but we can also enable the slow ...
by Stan Tomov
Wed Sep 19, 2018 10:09 am
Forum: User discussion
Topic: Batched dpotri?
Replies: 2
Views: 1929

Re: Batched dpotri?

Hi Rene, This sounds very good! We are interested in any improvement in these kernels. I will contact you with the procedure how to contribute it - we will have some software engineering requirements and we have to see if the code can be generalized (and tuned easily for other sizes and precisions)....
by Stan Tomov
Thu Jul 12, 2018 7:09 pm
Forum: User discussion
Topic: Magma Example gives wrong results
Replies: 3
Views: 926

Re: Magma Example gives wrong results

Dear Martin,

Very good! So the problem was not in MAGMA?

Yes, we recommend using MAGMA. We support multiple right hand sides for CG
and have very highly optimized SpMM product that is used in this case! Tell us how it goes.

Best regards,
Stan
by Stan Tomov
Mon Sep 18, 2017 1:11 pm
Forum: User discussion
Topic: magma_dgeqp3 parameter 2 incorrect
Replies: 4
Views: 986

Re: magma_dgeqp3 parameter 2 incorrect

You can just recompile - remove testing_dgeqp3 and testing_dgeqp3.o, and do make testing_dgeqp3 to see how it is compiled and linked through the magma testers. The way you have it is a little different but still worked on the system that I tested. Compiling and linking the same way indeed sounds lik...
by Stan Tomov
Fri Sep 15, 2017 10:39 pm
Forum: User discussion
Topic: magma_dgeqp3 parameter 2 incorrect
Replies: 4
Views: 986

Re: magma_dgeqp3 parameter 2 incorrect

This is interesting. I took your example and the way you compile and link it,
and managed to run the code without problem. I would have guessed there is
problem with MKL, but you say the testing_dgeqp3 in magma-2.2.0 runs fine.
Stan
by Stan Tomov
Mon May 09, 2016 2:57 pm
Forum: User discussion
Topic: problem with magma_mic_server
Replies: 1
Views: 760

Re: problem with magma_mic_server

Hi, This procedure is correct but for the older versions of the MAGMA MIC library (see file INSTRUCTIONS_to_run.txt). The current version is 1.4, and here you don't have to login directly to the MIC and start the server. The MAGMA MIC initialization routine is doing this (copying files to the MIC an...
by Stan Tomov
Mon Oct 19, 2015 1:07 pm
Forum: User discussion
Topic: LOBPCG method for Eigen Value Computation.
Replies: 4
Views: 2158

Re: LOBPCG method for Eigen Value Computation.

Nithish,
We haven't provided explicitly an option for that but if you compute the eigenvalues for -A, and multiply the result by -1 at the end, you will get the set of the largest eigenvalues for A.
Best regards,
Stan