Search found 9 matches

by ichitaro
Thu Mar 12, 2015 1:25 pm
Forum: User discussion
Topic: Symmetric and Hermitian solvers
Replies: 5
Views: 1451

Re: Symmetric and Hermitian solvers

Thank you for the post, again,

Apparently, the previous patch was a quick-fix to address the multiple rhs. Here is another version, which should perform better. Please let us know if you have more issues!!

Best,
Ichi
by ichitaro
Thu Mar 05, 2015 1:23 pm
Forum: User discussion
Topic: Symmetric and Hermitian solvers
Replies: 5
Views: 1451

Re: Symmetric and Hermitian solvers

Hi, Mauro,

Actually, there was a bug in ZHESV for multiple right-hand-sides, which has been fixed in our SVN. I attached the patch.

Please let us know if you have more comments!!
Ichi
by ichitaro
Thu Mar 05, 2015 1:09 pm
Forum: User discussion
Topic: Symmetric and Hermitian solvers
Replies: 5
Views: 1451

Re: Symmetric and Hermitian solvers

Hello, Mauro, 1. I also need to take a look at this numerical issues. Since we use random matrix and vectors, I would think the error should stay about the same for multiple right hand sides (unless we get very unlucky).. I will let you know if I find something.. 2. That is a bug. We use LU flops fo...
by ichitaro
Thu Mar 05, 2015 12:52 pm
Forum: User discussion
Topic: testing_dgetrf_gpu runs, testing_dgetrf_mgpu crashes
Replies: 1
Views: 680

Re: testing_dgetrf_gpu runs, testing_dgetrf_mgpu crashes

Hello, Gilles, The reason the multi-GPU version requires twice GPU memory is that to get good performance of getrf (during pivot), we transpose the matrix and store it in row-major. Since we do not have in-place transposition routine for non-square matrix, we needs the same amount of workspace as th...
by ichitaro
Thu Apr 10, 2014 2:48 pm
Forum: User discussion
Topic: Calling magma_dgeqrf_mgpu and magma_dpotrf_mgpu
Replies: 1
Views: 537

Re: Calling magma_dgeqrf_mgpu and magma_dpotrf_mgpu

MAGMA multi-GPU interface assumes the matrix is distributed using the block size specified by the get_nb routine. If the matrix is distributed using a block size different from get_nb(), then the code will either give a wrong result or will crash. You may be able to this by modifying control/get_nb....
by ichitaro
Thu Sep 26, 2013 3:30 pm
Forum: User discussion
Topic: testing magma_dsytrd_mgpu
Replies: 2
Views: 1166

Re: testing magma_dsytrd_mgpu

Hello, Jeremiah,

I am attaching the testing programs that I used for my development.
Please let me know if you encounter any issues with the codes.

Thank you for your interest,
Ichi
by ichitaro
Tue Jul 16, 2013 8:51 pm
Forum: User discussion
Topic: Support for very large Matrices
Replies: 1
Views: 1952

Re: Support for very large Matrices

Hi, If the matrix is too large to be stored on your GPU, then magma_dgetrf automatically calls its out-of-GPU-memory implementation. So, the size of the matrix that can be factorized using magma_dgetrf is limited by the size of the CPU memory. If you are using MAGMA 1.4.0 beta, please see the bug re...
by ichitaro
Tue Jul 16, 2013 5:13 pm
Forum: User discussion
Topic: magma-1.4.0-beta2: numerical problems [dz]getrf
Replies: 7
Views: 3963

Re: magma-1.4.0-beta2: numerical problems [dz]getrf

Hi, I'm sorry for the problem.. Unfortunately, I don't have an access to the particular GPU and cannot reproduce the error. More information about the error would be helpful (e.g., was an older version of getrf working and is it reproducible)? It is probably not helpful, but I am attaching the lates...
by ichitaro
Fri Jul 12, 2013 11:54 am
Forum: User discussion
Topic: magma-1.4.0-beta2: numerical problems [dz]getrf
Replies: 7
Views: 3963

Re: magma-1.4.0-beta2: numerical problems [dz]getrf

This will be fixed in the next release. For now, please put the attached files under the src directory. Please let us know if this did not fix the problems.

Thank you very much for the bug report,
Ichi