OpenMP — a directive-based API for shared-memory parallel programming in C/C++/Fortran, enabling parallelization with minimal code changes.

Basic Usage

#pragma omp parallel for
for (int i = 0; i < N; i++) {
    result[i] = compute(data[i]);
}

One line added → loop runs on all available cores.

Key Directives

• #pragma omp parallel — create a team of threads
• #pragma omp for — distribute loop iterations among threads
• #pragma omp critical — mutual exclusion for a code block
• #pragma omp atomic — atomic update of a single variable
• #pragma omp barrier — synchronization point
• #pragma omp task — create a task for dynamic parallelism

Data Sharing

• shared(var) — all threads see the same variable (default for most)
• private(var) — each thread gets its own copy
• reduction(+:sum) — each thread has private copy, combined at end
• firstprivate / lastprivate — control initialization and final value

Scheduling

• schedule(static) — divide iterations equally upfront
• schedule(dynamic) — threads grab chunks from a queue
• schedule(guided) — decreasing chunk sizes (good for imbalanced workloads)

OpenMP is the easiest way to parallelize existing serial code — 80% of the benefit with 20% of the effort compared to manual threading.