Enable Intel MPI

Intel MPI is available on the AWS ParallelCluster AMIs for alinux, alinux2, centos7, centos8, ubuntu1604, and ubuntu1804 values for the base_os setting. To use Intel MPI, you must acknowledge and accept the terms of the Intel simplified software license. By default, Open MPI is placed on the path. To enable Intel MPI instead of Open MPI, you must first load the Intel MPI module. Then, you need to install the latest version by using module load intelmpi. The exact name of the module changes with every update. To see which modules are available, run module avail. The output is as follows.

$ module avail

----------------------------------------- /usr/share/Modules/modulefiles ------------------------------------------
dot                        libfabric-aws/1.8.1amzn1.3 module-info                null                       use.own
module-git                 modules                    openmpi/4.0.2

------------------------------------------------ /etc/modulefiles -------------------------------------------------

--------------------------------- /opt/intel/impi/2019.7.217/intel64/modulefiles ----------------------------------
intelmpi

To load a module, run module load modulename. You can add this to the script used to run mpirun.

$ module load intelmpi

To see which modules are loaded, run module list.

$ module list
Currently Loaded Modulefiles:
  1) intelmpi

To verify that Intel MPI is enabled, run mpirun --version.

$ mpirun --version
Intel(R) MPI Library for Linux* OS, Version 2019 Update 7 Build 20200312 (id: 5dc2dd3e9)
Copyright 2003-2020, Intel Corporation.

After the Intel MPI module has been loaded, multiple paths are changed to use the Intel MPI tools. To run code that was compiled by the Intel MPI tools, load the Intel MPI module first.

Note

Intel MPI isn't compatible with AWS Graviton-based instances.

Note

Before AWS ParallelCluster version 2.5.0, Intel MPI wasn't available on the AWS ParallelCluster AMIs in the China (Beijing) and China (Ningxia) Regions.