Enable the ASYNC progress thread on HPC having infiniband.

[Mohit9638]

Respected sir/madam,

Please help me to enable the async thread for Intel MPI on HPC having interconnect as Infiniband.

Thanks,

Mohit Kumar

[SantoshY_Intel]

Hi,

Thanks for posting in Intel Communities.

Below are the steps to enable asynchronous progress control:

1. To specify the configuration, source the vars.[c]sh script with the -i_mpi_library_kind environment variable.

Example: 

source /opt/intel/oneAPI/2021.4.0.3347/setvars.sh 
source /opt/intel/oneAPI/2021.4.0.3347/mpi/2021.4.0/env/vars.sh -i_mpi_library_kind=release

For more information, refer to the below link:

https://www.intel.com/content/www/us/en/develop/documentation/mpi-developer-guide-linux/top/running-applications/selecting-a-library-configuration.html

(or) You can just export the I_MPI_LIBRARY_KIND environment variable which is equivalent to step 1 mentioned above.

Example:

$ export I_MPI_LIBRARY_KIND=release

For more information, refer to the below link:

https://www.intel.com/content/www/us/en/develop/documentation/mpi-developer-reference-linux/top/environment-variable-reference/other-environment-variables.html

2. Compile the MPI code using the below command:

mpiicc sample.c

3. Now, pass 1 to I_MPI_ASYNC_PROGRESS environment variable to enable asynchronous progress control.

I_MPI_ASYNC_PROGRESS=1 mpirun -n 16 ./a.out

For more information, refer to the below link:

https://www.intel.com/content/www/us/en/develop/documentation/mpi-developer-guide-linux/top/additional-supported-features/asynchronous-progress-control.html

I tried the above steps using the Intel MPI 2021.4 on a Linux machine and sharing with you the complete debug log for each step.

If this resolves your issue, make sure to accept this as a solution. This would help others with similar issues. Thank you!

Best Regards,

Santosh

[Mohit9638]

Hi Santosh,

I have tried your approach but not getting the desired result.

-The time taken by sample code with I_MPI_ASYNC_PROGRESS=0 << time taken by sample code with I_MPI_ASYNC_PROGRESS=1

- I am using HPC having infiniband with Intel Compiler 2020.2.254 version.

Thanks,

Mohit Kumar

[SantoshY_Intel]

Hi,

>>"The time taken by sample code with I_MPI_ASYNC_PROGRESS=0 << time taken by sample code with I_MPI_ASYNC_PROGRESS=1"

We can see that you have raised a similar issue in the Intel community forum. So, please keep track of your issue for updates using the below link:

https://community.intel.com/t5/Intel-oneAPI-HPC-Toolkit/Intel-MPI-Compiler-2018-5-274/m-p/1365509#M9252

>>"I have tried your approach but not getting the desired result."

Since your initial query has been resolved, can we go ahead and close this thread?

Thanks & Regards,

Santosh

[SantoshY_Intel]

Hi,

We haven't heard back from you. Since your initial query has been resolved, can we go ahead and close this thread?

Thanks & Regards,

Santosh

[SantoshY_Intel]

Hi,

Since your initial query is resolved, we are closing this thread. If you need any additional information, please post a new question as this thread will no longer be monitored by Intel.

Thanks & Regards,

Santosh