Search found 90 matches

by hartwig anzt
Thu Mar 12, 2015 9:40 am
Forum: User discussion
Topic: Running tests fail and visual studio configuration
Replies: 40
Views: 12817

Re: Running tests fail and visual studio configuration

All these are old routines, that should be no longer included in the package. You should just delete them, then you should be fine.

Thanks, Hartwig
by hartwig anzt
Tue Jan 13, 2015 3:39 pm
Forum: User discussion
Topic: floating point exception using magma_dpgmres
Replies: 5
Views: 1352

Re: floating point exception using magma_dpgmres

Noran, unfortunately, I don't think I can help you this time. The problem is, that running several instances of GMRES requires multiple threads - and MAGMA is in the current version not threadsafe. I am sorry, but you may have to solve one system after another... How many systems are you actually so...
by hartwig anzt
Thu Jan 08, 2015 2:38 pm
Forum: User discussion
Topic: Sparse Iterative Solvers Scalability Issue in MAGMA Library
Replies: 1
Views: 562

Re: Sparse Iterative Solvers Scalability Issue in MAGMA Libr

Dear Ayaz ul Hassan Khan, Thank you for your question. I assume you refer with "array dimension" to the size of the linear system, i.e. the number of unknowns. Your graph shows data for up to 33.000 unknowns. This is still very small for a GPU-based solver. I.e. the latency of the kernels launched m...
by hartwig anzt
Thu Jan 08, 2015 12:57 pm
Forum: User discussion
Topic: floating point exception using magma_dpgmres
Replies: 5
Views: 1352

Re: floating point exception using magma_dpgmres

Noran,

could you please provide your complete runfile?

From "opts.solver_par.solver = Magma_GMRES;" I deduce you have a loop starting several solvers at the same time? This may cause problems, but for looking into it, I would need the complete code.

Thanks, Hartwig
by hartwig anzt
Tue Dec 09, 2014 7:16 pm
Forum: User discussion
Topic: Compiling/linking Magma Sparse on Windows 7 using MSVS 2012
Replies: 2
Views: 934

Re: Compiling/linking Magma Sparse on Windows 7 using MSVS 2

Thank you, for picking up on that - you are totally correct, Noran. And indeed, there are some bugs because of the MS C compiler. I apologize for those. I did never work on a MS system to test the functionalities. However, I am now working on fixing these bugs, and they should all be resolved in the...
by hartwig anzt
Wed Nov 26, 2014 9:31 am
Forum: User discussion
Topic: Problem using magma_queue's
Replies: 4
Views: 2514

Re: Problem using magma_queue's

Noran,

very good to hear it's working now. Maybe think about using a preconditioned GMRES to accelerate ( i.e. ILU-GMRES)

Hartwig
by hartwig anzt
Tue Nov 25, 2014 12:16 pm
Forum: User discussion
Topic: Problem using magma_queue's
Replies: 4
Views: 2514

Re: Problem using magma_queue's

Noran, I am not sure I understand your problem, I assume you want to solve AX=B where A is 400x400 and B 400x10. I looked into your code. If GMRES is converging very slow, maybe use a preconditioner? another restart parameter? Or (in case SPD) CG instead of GMRES? I modified your code a bit, as ther...
by hartwig anzt
Mon Nov 24, 2014 10:06 am
Forum: User discussion
Topic: Compiling MAGMA sparse
Replies: 5
Views: 2215

Re: Compiling MAGMA sparse

Noran, yes, there is a precision indicator missing. please add in the "magma_xvpass.cpp" the x before the function definition: magma_zvget magma_cvget magma_dvget magma_svget Sorry for that! However, I use the make.inc.mkl-shared and it works - although the function may not be usable... Please let m...
by hartwig anzt
Wed Nov 19, 2014 10:06 am
Forum: User discussion
Topic: linear solver sgmres very slow
Replies: 3
Views: 1436

Re: linear solver sgmres very slow

Noran, good catch - GMRES may have problems to converge to this accuracy. Also, the sparse-iter package is originally designed for significantly larger systems, typically containing 500K-2M unknowns. You may have a special reason, but systems that small are usually solved with a direct solver - e.g....
by hartwig anzt
Tue Nov 18, 2014 10:03 am
Forum: User discussion
Topic: Bugs in magma_<t>cuspmm (magma 1.6)
Replies: 2
Views: 1353

Re: Bugs in magma_<t>cuspmm (magma 1.6)

Noran,

I apologize again. This happens if you typically deal with squared matrices only... sorry for that! I changed it for the next version.

Great thanks for reporting! Hartwig