intel mpi

Version:

2019.5.281 (x32), 2019.5.281 (x64)

Category:

mpi

Cluster:

Loki

Author / Distributor

https://www.intel.com/content/www/us/en/developer/tools/oneapi/mpi-library.html

Description

Intel MPI Library is a high-performance implementation of the Message Passing Interface (MPI) standard. It is optimized for Intel architectures but also runs efficiently on non-Intel hardware.

Key features:

  • Compliant with MPI-3.1 standard

  • Optimized for performance and scalability

  • Supports hybrid parallelism (MPI + OpenMP)

  • Works with OFI, TCP, and shared memory fabrics

  • Includes wrappers: mpiicc, mpiicpc, mpiifort

Documentation

$ mpirun -help

Intel(R) MPI Library for Linux* OS, Version 2019 Update 5 Build 20190815
Copyright 2003-2019, Intel Corporation

Usage: mpirun [options] [program] [args]

Common options:
  -n, -np <num>           Number of processes
  -ppn <num>              Number of processes per node
  -hostfile <file>        File with list of hosts
  -hosts <list>           Comma-separated list of hosts
  -genv <var> <value>     Set environment variable
  -envall                 Export all environment variables
  -config <file>          Read arguments from config file
  -launcher ssh|rsh|pbs   Launch method
  -help                   Show help message

For more, see: https://www.intel.com/mpi

Examples/Usage

  • Load the Intel MPI module:

    $ module load intel/mpi/64/2019/5.281

  • Unload the module:

    $ module unload intel/mpi/64/2019/5.281

  • Compile and run a parallel job with 4 processes:

    mpiicc -o myprog myprog.c
mpirun -n 4 ./myprog

  • Run with environment variable propagation:

    mpirun -n 4 -genv OMP_NUM_THREADS 2 ./myprog

Installation

Source code and binaries are obtained from Intel MPI Library