Intel® oneAPI Math Kernel Library
Ask questions and share information with other developers who use Intel® Math Kernel Library.

MKL running in parallel on a server with two processors

bubai_l_
Beginner
313 Views

Hello everyone.

I have one server with two processors. Every processor has 16 cores. I found that MKL routines such as DGEMM can only run on one processor. I want that the MKL can employ all computing resource, i.e. two processors and 32 cores. How can I deal with this problem? 

Thank you very much!

0 Kudos
1 Solution
Ying_H_Intel
Employee
313 Views

Hi Bubai

Could you please tell the processor type of the server  and the matrix size of dgemm? 

According to MKL user guide : https://software.intel.com/en-us/node/528369

By default, Intel MKL uses the number of threads equal to the number of physical cores on the system. If your processor is Hyper Threading, (logical core is on), then it may only use the physical cores.  

Best Regards,

Ying

View solution in original post

0 Kudos
2 Replies
Ying_H_Intel
Employee
314 Views

Hi Bubai

Could you please tell the processor type of the server  and the matrix size of dgemm? 

According to MKL user guide : https://software.intel.com/en-us/node/528369

By default, Intel MKL uses the number of threads equal to the number of physical cores on the system. If your processor is Hyper Threading, (logical core is on), then it may only use the physical cores.  

Best Regards,

Ying

0 Kudos
bubai_l_
Beginner
313 Views

Ying H (Intel) wrote:

Hi Bubai

Could you please tell the processor type of the server  and the matrix size of dgemm? 

According to MKL user guide : https://software.intel.com/en-us/node/528369

By default, Intel MKL uses the number of threads equal to the number of physical cores on the system. If your processor is Hyper Threading, (logical core is on), then it may only use the physical cores.  

Best Regards,

Ying

Hi! Thank you very much! The problem has been solved with your help. Just like what you said, my processor is Hyper Threading and Intel MKL uses the number of threads equal to the number of physical cores. Thank you very much again!

0 Kudos
Reply