Search found 10 matches

by abdelfattah83
Thu Jun 11, 2020 4:39 pm
Forum: User discussion
Topic: Max batch size for dgesv ?
Replies: 4
Views: 396

Re: Max batch size for dgesv ?

The y-dimension has the same limit as the z-dimensions. Some kernels do use the X dimension for batching, but most of our code use the x-y dimensions for thread configurations. Changing that would be cumbersome because of the many kernels MAGMA has. BTW, if you problem size is fixed at 3x3, you can ...
by abdelfattah83
Thu Jun 11, 2020 12:37 am
Forum: User discussion
Topic: Max batch size for dgesv ?
Replies: 4
Views: 396

Re: Max batch size for dgesv ?

Most of the batch kernels in MAGMA use the z-dimension of the kernel grid for batching across different problems. The z-dimension has a hardware limit of 65535. That is why the routine fails. I think that for your particular case, the error comes from dgetrs. The batch dgetrf routine does not have t...
by abdelfattah83
Thu Oct 03, 2019 12:54 pm
Forum: User discussion
Topic: Best solution for solving hundreds of small linear systems
Replies: 3
Views: 994

Re: Best solution for solving hundreds of small linear systems

There is a batch routine for solving many small linear systems (magma_Xgesv_batched), where X specifies the precision (s, d, c, z). The routine applies LU factorization with partial pivoting to the input matrices, followed by a row interchanges step and triangular solves. For very small matrices, th...
by abdelfattah83
Sat Jul 20, 2019 8:34 pm
Forum: User discussion
Topic: [Bug?] spotrf_batched fails for batch size > 524280 and dpotrf_batched fails for batch size > 262140
Replies: 1
Views: 1429

Re: [Bug?] spotrf_batched fails for batch size > 524280 and dpotrf_batched fails for batch size > 262140

Most of the MAGMA batched kernels use the z-dimension of the kernel grid for batching across different problems. The maximum value of this dimension is 65535 (a hardware limitation of the GPU itself). Depending on the kernel configuration, a large batch size may lead to exceeding this limit. If you ...
by abdelfattah83
Thu May 23, 2019 11:11 pm
Forum: User discussion
Topic: makefile question when adding a file to magmablas/ directory
Replies: 6
Views: 1833

Re: makefile question when adding a file to magmablas/ directory

You should add your new CUDA file to the Makefile.src file under the magmablas subdirectory.

Ahmad
by abdelfattah83
Thu May 09, 2019 12:58 pm
Forum: User discussion
Topic: [BUG, Fixed] (d/s)potrf_batched doesn't send back errors
Replies: 4
Views: 2051

Re: [BUG] (d/s)potrf_batched doesn't send back errors

Hi Vishwak, Thank you for reporting this. I have committed a fix for reporting the correct info. Please check it out and let me know. One of the reasons was NaN checking, which was missing from the panel factorization kernel. The way the kernel was written made it encounter NaNs when factorizing a c...
by abdelfattah83
Wed Apr 24, 2019 4:24 pm
Forum: User discussion
Topic: [BUG, Fixed] (d/s)potrf_batched doesn't send back errors
Replies: 4
Views: 2051

Re: [BUG] (d/s)potrf_batched doesn't send back errors

Can you please give some details about your specific test case. I have tried some simple tests on my side, which seem to be always detected. Sometimes the value of info is not returned correctly (a non-zero value, but pointing to the wrong column/diagonal) but the routine always catches the error. T...
by abdelfattah83
Mon Dec 17, 2018 10:01 am
Forum: User discussion
Topic: (d/s)getrf_batched doesn't send back errors
Replies: 4
Views: 2077

Re: (d/s)getrf_batched doesn't send back errors

Great, thanks for letting us know.

--Ahmad
by abdelfattah83
Tue Dec 11, 2018 2:20 pm
Forum: User discussion
Topic: (d/s)getrf_batched doesn't send back errors
Replies: 4
Views: 2077

Re: (d/s)getrf_batched doesn't send back errors

This should be fixed now if you update your local copy of the repository. Please test on your side and let us know if the issues are gone.

--Ahmad
by abdelfattah83
Mon Dec 10, 2018 3:43 pm
Forum: User discussion
Topic: (d/s)getri_outofplace_batched can't handle many matrices
Replies: 2
Views: 767

Re: (d/s)getri_outofplace_batched can't handle many matrices

This is a known issue for most of the batch routines, not only getri_outofplace_batched. The explanation might be a little low-level. Most of the MAGMA batch kernels use the grid-z dimension to get a "batch-ID". This dimension has a maximum value of 65535 on all NVIDIA GPUs (hardware limitation), af...