fopenmp

OpenMP is an API for multi-platform shared-memory parallel programming in C, C++, and Fortran. It provides a portable, scalable model based on compiler directives, runtime routines, and environment variables that enable developers to write parallel code for multi-core and multi-processor systems without explicit thread management.

The execution model is fork-join: a program starts with a single thread, and entering a parallel region creates a team of threads that execute the enclosed code in parallel. OpenMP offers work-sharing constructs such as parallel, for, sections, and tasking in newer versions, allowing various granularities of parallelism. In C and C++, directives such as #pragma omp parallel for control the distribution of loop iterations; in Fortran, equivalent directives are used.

OpenMP provides data-sharing attributes and synchronization primitives. Private, shared, firstprivate, and reduction specify variable scoping and data flow, while critical, atomic, barrier, and ordered enforce synchronization. Scheduling options (static, dynamic, guided, runtime) control how iterations are assigned to threads; the OMP_SCHEDULE environment variable and related runtime calls influence behavior. Nested parallelism and tasking extend the model in newer standards.

OpenMP originated in the late 1990s and is maintained by the OpenMP Architecture Review Board. Standards have evolved through OpenMP 4.x and 5.x, adding tasking, SIMD directives, and accelerator support. Implementations are available in GCC, Clang, Intel ICC, IBM XL, Oracle, and others, and OpenMP can be used in hybrid MPI+OpenMP programs.

The API emphasizes portability and ease of use, but achieving high performance requires attention to data locality, synchronization costs, and correct scoping to avoid race conditions. Common practices include choosing appropriate scheduling, minimizing critical sections, and profiling parallel regions.