I have a simulation setup which has been compiled with intel 2020u4.
problem statement -
The setup has 2 parts - input read, output writing.
scenario a) When i run a single copy of my setup, everything works fine, i.e. input read completes in ~20 minutes and output writing phase in ~60 minutes (overall ~80 minutes)
scenario b) But when i run multiple copies of the same run, then the input phase takes ~50 minutes followed by ~60 minutes of output .
I plan to use intel profiler to find the "slowest part" with the runs. I plan to run the profiler in scenario a) but i plan to profile only initial 50 minutes of simulation, and i dont care about the "output phase" profiling data.
How can i achieve this using aps?
i understand that the mpirun command will be -
mpirun -np X -ppn Y aps --collection-mode=mpi ./application.exe
But are there any additional options to make profiler stop gathering data?
or if i kill the application.exe after 50 minutes, the profiler will be able to show correct data?
Thanks for reaching out to us.
>>"How can i achieve this using aps? are there any additional options to make profiler stop gathering data?"
- Application Performance Snapshot supports the region control with the MPI_Pcontrol() operation.
- This allows you to enable and disable statistics collection for specific application regions in the source code. Region control only affects MPI and OpenMP* metrics, while the other ones will still be collected for the entire application.
For more information refer to the below link:
>>" if i kill the application.exe after 50 minutes, the profiler will be able to show correct data?"
No, It is not recommended to kill an application as it may result in either incorrect data or no data.
Thanks & Regards,
I 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.
Thanks & Regards,