topic here is the link to the Intel in Intel® oneAPI Math Kernel Library

Cannot redefine mkl_progress with tbb and static linking on MacOS

Sadri__Bardia — Wed, 27 May 2020 19:35:08 GMT

We've spend several days trying to override "mkl_progress" function. It is absolutely crucial for us to be able to abort the calculation (our software appears hanged otherwise if the input matrix grows too large and user cannot stop the computation). Unfortunately, despite all the efforts, there appears to be no way to take over the "mkl_progress". Even when I call "mkl_progress" directly, my redefined version is not called. I'm linking statically with MKL on MacOS and we use tbb. Is there any information you could gives to overcome this? The "mkl_set_progress" function also does nothing. The documentations on the matter are also depressingly insufficient.

Hi Bardia,

Aaron_J_Intel2 — Wed, 27 May 2020 23:03:03 GMT

Hi Bardia,

I see a note in the documentation about the mkl_progress feature and TBB.

"WARNING: The mkl_progress function only supports OpenMP* threading and sequential execution."

https://software.intel.com/content/www/us/en/develop/documentation/mkl-developer-reference-c/top/support-functions/miscellaneous/mkl-progress.html

Please submit mkl_progress support for TBB as a feature request.

here is the link to the Intel

Gennady_F_Intel — Thu, 28 May 2020 01:51:28 GMT

here is the link to the Intel Online Service Center - https://supporttickets.intel.com/?lang=en-US

>> "...The documentations on

Gennady_F_Intel — Thu, 28 May 2020 05:56:38 GMT

>> "...The documentations on the matter are also depressingly insufficient. "

What do you like to see from the mkl documentation else regarding this part beyond warnings?

"WARNING: The mkl_progress function only supports OpenMP* threading and sequential execution."

https://software.intel.com/content/www/us/en/develop/documentation/mkl-d.