Compute LU triangles directly for sparse (compressed row storage) vectors

[EJRicketts]

Hi,

I have a sparse/CRS vector which I would like to compute the LU decomposition of for the lower triangle. Is there a way to output this directly?

Thanks

[ShanmukhS_Intel]

Hi,

Thank you for posting on Intel Communities.

Could you please be more specific regarding the format in which you would like the operation to be performed, so that we could assist you in appropriate solvers to be used.

In addition, Please refer to below link for the routines which would be useful to choose based on your desired operation to be performed and let us know if the use case matches your requirement.

https://www.intel.com/content/www/us/en/develop/documentation/onemkl-developer-reference-c/top/sparse-solver-routines/onemkl-pardiso-parallel-direct-sparse-solver-iface.html#onemkl-pardiso-parallel-direct-sparse-solver-iface

Best Regards,

Shanmukh.SS

[EJRicketts]

Hi,

Thank you for your reply!

I would be looking at using pardiso with the three array variation of CSR format:

https://www.intel.com/content/www/us/en/develop/documentation/onemkl-developer-reference-c/top/appendix-a-linear-solvers-basics/sparse-matrix-storage-formats/sparse-blas-csr-matrix-storage-format.html

Best,

Evan

[ShanmukhS_Intel]

Hi,

Could you please let us know if you need L and U factors directly from Pardiso?

Best Regards,

Shanmukh.SS

[EJRicketts]

Hi,

Yes, I would like the lower triangle of a sparse matrix.

Thanks,

Evan

[ShanmukhS_Intel]

Hi,

As of now, there are no functions available for PARDISO which can return LU factors.

We would like to request you to raise a Feature Request if you have it as a requirement.

Best Regards,

Shanmukh.SS

[ShanmukhS_Intel]

Hi,

A gentle reminder:

Has the information provided helped? Kindly let us know so that we could close this thread at our end.

Best Regards,

Shanmukh.SS

[EJRicketts]

Hi,

Just a last question. Is there a routine in the MKL which could perform a Cholesky decomposition of a matrix in CRS format? I know that ?potrf exists, but I think that this is for full arrays only.

Thanks,

Evan

[ShanmukhS_Intel]

Hi,

p?potrf computes the Cholesky factorization of a symmetric (Hermitian) positive-definite distributed matrix.

It computes the Cholesky factorization of a real symmetric or complex Hermitian positive-definite distributed n-by-n matrix A(ia:ia+n-1, ja:ja+n-1).

Please find the below link for more information regarding the computational routine. Kindly get back to us with more specific details in case of any help is needed.

https://www.intel.com/content/www/us/en/develop/documentation/onemkl-developer-reference-c/top/scalapack-routines/scalapack-computational-routines/matrix-factorization-scalapack-computation/p-potrf.html

Best Regards,

Shanmukh.SS

[ShanmukhS_Intel]

Hi,

A gentle reminder:

Has the information provided helped? Kindly let us know so that we could close this thread at our end.

Best Regards,

Shanmukh.SS

[ShanmukhS_Intel]

Hi,

We assume that your issue is resolved. If you need any additional information, please post a new question as this thread will no longer be monitored by Intel.

Best Regards,

Shanmukh.SS