Intel® oneAPI Math Kernel Library
Ask questions and share information with other developers who use Intel® Math Kernel Library.
7080 Discussions

mkl_verbose_output_file to same output as main code

JeffS
Novice
807 Views

My code generates an output file as many do...  There are a number of MKL calls throughout the code and I want to MKL_VERBOSE output to go to the same output file that my code is writing to.

 

I tried to implement this by using mkl_verbose_output_file(filePath)

  • If I set filePath to my output file... the return code is 0 (Success), but there is no verbose output in the file.
  • If I set filePath to another file name, return code is 0, and I do get output to the path I set

 

Can the MKL_VERBOSE output be directed to a file that is shared with the main code?

 

If everything goes to stdout (program output and MKL_VERBOSE), then I can capture by piping to a file.  However this code is run using MPI master/slave so I cannot just dump to stdout.

 

 

Side note:  mkl_verbose_output_file seems to require calling with trim(filePath).  I was getting a return status of 1 until I used trim().  I would expect it to work like open() where it handles trimming inherently.

0 Kudos
4 Replies
AryanK_Intel
Employee
756 Views

Hi,

 

Thanks for posting in Intel Communities.

 

>>If I set filePath to another file name, return code is 0, and I do get output to the path I set

Could you please kindly clarify if the mentioned output refers to verbose information, the actual code output, or a combination of both?

 

Thanks and Regards,

Aryan.

 

0 Kudos
JeffS
Novice
745 Views

The output is only the MKL_VERBOSE output.

 

I want to achieve combined output to a single file.  Without using std out.

0 Kudos
AryanK_Intel
Employee
675 Views

Hi,


We are looking into your issue internally and we will update you soon.


Thanks and Regards,

Aryan


0 Kudos
AryanK_Intel
Employee
617 Views

Hi,


Thanks for sharing your requirement. 

This feature is currently not available. We will update this thread once we implement this feature. 


Thanks & Regards,

Aryan.


0 Kudos
Reply