mvapich2

Version:

2.3.4 (gcc), 2.3.4 (intel), 2.3.4 (pgi), 2.3.4 (psmgcc), 2.3.4 (psmintel)

Category:

mpi

Cluster:

Loki

Author / Distributor

https://mvapich.cse.ohio-state.edu/

Description

MVAPICH2 is an MPI-3 compliant implementation optimized for high-performance interconnects including InfiniBand, Omni-Path, and RoCE. Built on top of MPICH, it is tuned for scalability and performance on modern HPC systems.

This build (psmintel/64/2.3.4) is optimized for Intel-based systems using the PSM (Performance Scaled Messaging) interface.

Documentation

$ mpirun_rsh --help

Usage: mpirun_rsh [-np <num>] [-hostfile <file>] [-env <VAR=VAL>] [program] [args]

MVAPICH2 also provides:
  mpicc     - C compiler wrapper
  mpicxx    - C++ compiler wrapper
  mpif77    - Fortran 77 compiler wrapper
  mpif90    - Fortran 90 compiler wrapper

You can also use:
  mpiexec or mpirun

For full documentation, see:
  https://mvapich.cse.ohio-state.edu/userguide/

Examples/Usage

• Load the MVAPICH2 module:

  $ module load mpi/mvapich2/psmintel/64/2.3.4
  

• Unload the module:

  $ module unload mpi/mvapich2/psmintel/64/2.3.4
  

• Compile and run a C MPI program:

  mpicc -o hello_mpi hello.c
  mpirun -np 4 ./hello_mpi
  

• Use a hostfile for launching:

  mpirun_rsh -hostfile nodes.txt -np 8 ./my_program
  

Installation

Source code is obtained from MVAPICH2