Skip to content

What is OpenMPI?

OpenMPI is an open-source implementation of MPI (Message Passing Interface) — a standard used to develop parallel programs that can run across multiple computers or CPU cores in a cluster. It enables efficient communication between processes, regardless of whether they're running on different machines.

Key Uses of OpenMPI

  • High-Performance Computing (HPC)
  • Scientific Simulations (e.g., physics, chemistry, climate models)
  • Machine Learning and Data Processing at scale
  • Any application that requires faster computation by distributing work across processors.

How OpenMPI Works

Conceptually, OpenMPI operates by running multiple copies of the same program, where each copy is called a process. These processes communicate by sending and receiving messages. Here's how it works:

  1. You write one program.
  2. OpenMPI runs multiple instances of that program.
  3. Each process communicates with others using MPI messages.

Example: - Process 0 handles task A. - Process 1 handles task B. - They exchange results via MPI.

Key Features of OpenMPI

  • Runs on laptops, servers, or large clusters.
  • Supports languages such as C, C++, Fortran, and Python (via bindings like mpi4py).
  • Very fast and scalable.
  • Works over different networks: Ethernet, InfiniBand, shared memory, etc.

Installing OpenMPI from Source Code

Step 1: Download the Source Code

Download the source code from the official OpenMPI website.

Step 2: Extract the Source

After downloading the .tar.gz file, extract it using:

tar -xvf package.tar.gz
cd package

Step 3: Check Extracted Files

Ensure the following files are present:

  • configure
  • configure.ac
  • autogen.sh

Step 4: Generate the Configure Script (if missing)

If the configure script is missing, run:

./autogen.sh

Step 5: Configure the Build

Run the following command to configure the build:

./configure --prefix=/opt/openmpi/5.0.1

Configuring OpenMPI

Step 6: Build the Software

Use the make command to build OpenMPI. You can specify the number of cores for parallel compilation:

make -j <number_of_cores>

For example:

make -j 16  # Using 16 cores
make -j     # Use all available cores

Configuring OpenMPI Build

Step 7: Run Optional Checks

You can run a test to check the build:

make check

Step 8: Install OpenMPI

Once the build is successful, install OpenMPI with:

make install

Configuring OpenMPI Installation

Post-Installation Steps

After installation, perform the following steps:

  1. Create a module file for OpenMPI and the compiler.
  2. Add correct PATH and LD_LIBRARY_PATH to environment variables.
  3. Test the compiler and MPI setup.

Final Verification

To verify the installation, run the following commands:

module load compiler/version
module load openmpi/5.0.1
mpicc --version  # Check if MPI compiler is working
mpirun -np 2 hostname  # Run a simple MPI job

Final Verification

Best Practices and Notes

  • Use module files for HPC environments to manage versions.
  • Keep only one version per directory.
  • Document prefix paths for future reference.
  • Test with small MPI jobs to verify setup.

Document prepared for reference purposes only.