Intel® MPI Library
Get help with building, analyzing, optimizing, and scaling high-performance computing (HPC) applications.
2153 Discussions

Program froze when using MPICH_ASYNC_PROGRESS

black__edgar
Beginner
1,040 Views

Hi all,

I am having a problem when the MPICH_ASYNC_PROGRESS  (or I_MPI_ASYNC_PROGRESS) environment variable is set on; the program freeze and I have to kill it.

Have anyone experience this problem.

This is the version of the Intel compiler and Intel MPI library producing the problem:

icc (ICC) 2021.1 Beta 20201112
Copyright (C) 1985-2020 Intel Corporation. All rights reserved.

Intel(R) MPI Library for Linux* OS, Version 2021.1 Build 20201112 (id: b9c9d2fc5)
Copyright 2003-2020, Intel Corporation.

I have experienced this problem ever since I updated from:

icc (ICC) 19.1.3.304 20200925
Copyright (C) 1985-2020 Intel Corporation. All rights reserved.

Intel(R) MPI Library for Linux* OS, Version 2019 Update 9 Build 20200923 (id: abd58e492)
Copyright 2003-2020, Intel Corporation.

 

Thanks in advance.

 

 

 

 

0 Kudos
1 Solution
PrasanthD_intel
Moderator
984 Views

Hi Edgar,


Thanks for the confirmation.

As the issue has been resolved, we will no longer respond to this thread. If you require additional assistance from Intel, please start a new thread. Any further interaction in this thread will be considered community only


Regards

Prasanth


View solution in original post

0 Kudos
6 Replies
PrasanthD_intel
Moderator
1,020 Views

Hi Edgar,


The additional features like multiple-endpoints and async threads are supported in release_mt and debug_mt versions only, as mentioned in the developer reference (Asynchronous Progress Control (intel.com)).

To enable those two library configurations, use this command:

source <install-dir>/env/vars.sh release

For more information on selecting library configuration please refer : Selecting a Library Configuration (intel.com)

If you are already enabled those configurations and still facing issues, please provide the log info you will get after setting I_MPI_DEBUG=10.

eg: export I_MPI_DEBUG=10


Regards

Prasanth


0 Kudos
black__edgar
Beginner
1,015 Views

Hi Prasanth,

Thank you for your advice. It fixed the problem.

I was using the wrong configuration script;

Instead of using: /opt/intel/oneapi/compiler/2021.1.1/env/vars.sh

I was using : /opt/intel/oneapi/setvars.sh

I also notice that you do not actually need the "release" parameter. Could this be correct?

Thank you again.

Edgar

 

 

0 Kudos
black__edgar
Beginner
1,008 Views

Hi Prasanth,

Here is an update of my previous post. The problem is not fixed just by changing the configuration script.

I was fooled by cmake mixing the Intel compiler and the gcc mpich library. That is what worked for me before.

After further exploration, I also set the mpi configuration script (/opt/intel/oneapi/mpi/2021.1.1/env/vars.sh) with the release_mt  parameter you suggested before as shown here:

/opt/intel/oneapi/mpi/2021.1.1/env/vars.sh -i_mpi_library_kind=release_mt

if "release" is used instead of "release_mt" the program freeze.

Thanks again,

Edgar

0 Kudos
PrasanthD_intel
Moderator
1,001 Views

Hi Edgar,


In the command i mentioned, I have given release as an example. To use async progress we have either set the library configuration to release_mt or debug_mt.

source <install-dir>/env/vars.sh release_mt

If this doesn't fix your problem, please let us know.


Regards

Prasanth


0 Kudos
black__edgar
Beginner
995 Views

Hi Prasanth,

Yes. The problem is gone.

The problem was produced by a bad setting of the configuration scripts.

Thanks,

Edgar

0 Kudos
PrasanthD_intel
Moderator
985 Views

Hi Edgar,


Thanks for the confirmation.

As the issue has been resolved, we will no longer respond to this thread. If you require additional assistance from Intel, please start a new thread. Any further interaction in this thread will be considered community only


Regards

Prasanth


0 Kudos
Reply