OpenMP Mini-Lab

OpenMP is a shared-memory parallel programming model designed to exploit multiple CPU cores on a single node. It enables developers to parallelize existing C/C++ or Fortran code using compiler directives, often with minimal changes to the original program structure.

In HPC–Quantum workflows, OpenMP is commonly used to accelerate local classical workloads, such as numerical kernels, simulations, and pre- or post-processing steps that run on a single compute node.

What This Mini-Lab Demonstrates¶

This mini-lab focuses on the core ideas behind OpenMP:

• Thread-based parallelism within a single node

• How work is shared among threads

• How performance changes as the number of threads increases

• The relationship between cores, threads, and shared memory

The emphasis is on building intuition rather than achieving peak performance.

Running the OpenMP Example¶

A minimal compile-and-run workflow looks like this:

gcc -fopenmp hello_openmp.c -o hello_openmp
export OMP_NUM_THREADS=4
./hello_openmp

• -fopenmp enables OpenMP support in the compiler

• OMP_NUM_THREADS controls how many threads are launched

• The program runs on a single node using shared memory

You can experiment by changing the number of threads and observing how the output or runtime changes.

Full Lab and Exercises¶

The complete OpenMP mini-lab, including explanations, variations, and additional exercises, is maintained in the following repository:

👉 https://github.com/friedsam/hpc-qc-mini-labs

This page is intentionally lightweight; the GitHub repository is the source of truth for the full lab material.

Notes on Containers¶

The OpenMP examples can also be run inside a containerized environment. Containers are useful for:

• Ensuring consistent compiler and library versions

• Running the same workflow locally and on HPC systems

• Simplifying setup for new users

See the mini-labs repository for details on building and running the provided container image.