topic Re: oneMKL DPC++ getri GPU in Intel® oneAPI Math Kernel Library

oneMKL DPC++ getri GPU

TestDR — Mon, 27 Feb 2023 22:20:10 GMT

Hi,

Reference:

Intel® oneAPI Math Kernel Library (oneMKL) - Data Parallel C++ Developer Reference, October 18, 2022 Release 2022.2

Section 5.1 Device Support

GPU offloading supported LAPACK routines

Routine	Supported precision (i.e. float, double, std::complex<float>, std::complex<double>, etc.)	Comment
getrf ()	all precisions	Computes the LU factorization of a general m-by-n matrix A.
getri ()	real precisions	Computes the inverse of an LU-factored general matrix determined by getrf.

As per above, getrf supports all precisions on GPU whereas only real precisions are supported for getri on GPUs. These APIs often used in cascade for LU factorisation and matrix inverse calculations respectively. This claim in section 5.1 contradicts with the API declaration for getri in section 11.26 according to which following precisions and devices are supported:

Precision (T)	Devices supported
float	Host, CPU, and GPU
double	Host, CPU, and GPU
std::complex<float>	Host, CPU, and GPU
std::complex<double>	Host, CPU, and GPU

Please confirm whether getri supports all precisions on GPU or only real precisions with Intel oneAPI DPC++ 2023.0 compiler?

Thanks,

Re: oneMKL DPC++ getri GPU

ShanmukhS_Intel — Thu, 16 Mar 2023 08:17:27 GMT

Hi Rajitha,

Thanks for posting on Intel Communities.

We would like to inform you that, getri routine supports float, double, std::complex<float> and std::complex<double> precisions on CPU and GPU devices. Regarding the contradiction, it seems there might be a typo error in the developer reference. We have informed the concerned team regarding the same.

Please refer to the below link for more details regarding getri routine and corresponding precisions/devices supported by the same for Version: 2023.0.

https://www.intel.com/content/www/us/en/develop/documentation/oneapi-mkl-dpcpp-developer-reference/t...

In addition, We would like to recommend you use the latest Intel® oneAPI Math Kernel Library (oneMKL) - Data Parallel C++ Developer Reference (version 2023.0) available for reference purposes.

Intel® oneAPI Math Kernel Library (oneMKL) - Data Parallel C++ Developer Reference

Version: 2023.0
Last Updated: 12/16/2022

Best Regards,

Shanmukh.SS

Re: oneMKL DPC++ getri GPU

ShanmukhS_Intel — Thu, 16 Mar 2023 08:18:22 GMT

Hi Rajitha,

A gentle reminder:

Has the information provided helped? If this resolves your issue, make sure to accept this as a solution. This would help others with a similar issue. Could you please let us know if we you need any other information?

Thank you!

Best Regards,

Shanmukh.SS

Re:oneMKL DPC++ getri GPU

ShanmukhS_Intel — Tue, 14 Mar 2023 05:15:38 GMT

Hi Rajitha,

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.

Have a great day!

Best Regards,

Shanmukh.SS