MAGMA Forum

Does anyone know the answer to my question?

Thanks,
Jeremiah

Hello!

Are there plans for including a routine in the testing directory to run magma_dsytrd_mgpu?

I have tried to run magma_dsytrd_mgpu from my own code, and I keep getting an error "MKL INTERNAL ERROR: Condition 2 detected in function magma_dsytrd_mgpu."

Thanks,
Jeremiah

There was a src/zgetrf_ooc.cc in magma_1.2.1, but I don't see it in magma_1.3.0. Was the zgetrf_ooc.cc in magma_1.2.1 faulty?

Thanks,
Jeremiah

Thanks for the info.

...Or maybe they will appear in a later magma version? I am eager to use this functionality.

Hello! I see in the magma_1.3 presentation pdf mentions of multi-gpu dsytrd routines. Where can I find the routines that were used to produce those results? Are they referenced in the testing directory?

Thanks,
Jeremiah

Right - my matrix size was 14307, which I padded up to size 14336 to avoid allocating the large amount of memory. Can I ask why the transpose at the beginning is performed at all?

The magma code zgetrf_gpu allocates a large space in gpu memory to hold a copy of the input matrix, padded to be of a size that is a multiple of 32. My gpu memory requirements are rather tight, so I modified zgetrf_gpu to never allocate space on the gpu, and I always make sure that my input matrix i...

I see. Well, the reason I was even looking at that was b/c my code uses the zgesv routine, and it kept hanging in the zlaswp2 kernel.

Hello! In the magmablas/zpermute-v2.cu, I see a couple of structs defined: typedef struct { cuDoubleComplex *A; int n, lda, j0; short ipiv[BLOCK_SIZE]; } zlaswp_params_t; typedef struct { cuDoubleComplex *A; int n, lda, j0, npivots; short ipiv[BLOCK_SIZE]; } zlaswp_params_t2; In the subroutine magma...