What is OpenMP
OpenMP is an industry standard API of C/C++ and Fortran for shared memory parallel programming. OpenMP Architecture Review Board (ARB) consists of major compiler vendors and many research institutions. Common architectures include shared memory architecture (multiple CPUs shared global memory, uniform memory access (UMA), with typical shared memory programming model of OpenMP, Pthreads), distributed memory architecture (each CPU has own memory, non-uniform memory access ((NUMA), with typical message passing programming model of MPI), and hybrid architecture (UMA within one node or socket, NUMA across nodes or sockets, with typical hybrid programming model of hybrid MPI/OpenMP). Current architecture trend needs a hybrid programming model with three levels of parallelism: MPI between nodes or sockets, shared memory (such as OpenMP) on the nodes/sockets, and increased vectorization for lower level loop structures.
OpenMP has three components: Compiler directives and clauses, runtime libraries and environment variables. The compiler directives are only interpreted when OpenMP compiler option is turned on. OpenMP uses the "fork and join" execution model: Master thread forks new threads at the beginning of parallel regions; Multiple threads share work in parallel; And threads join at the end of parallel regions.
In OpenMP, all threads have access to the same shared global memory. Each thread has access to its private local memory. Threads synchronize implicitly by reading and writing shared variables. No explicit communication is needed between threads.
Major features in OpenMP 3.1 include:
- Thread creation with shared and private memory
- Loop parallelism and work sharing constructs
- Dynamic work scheduling
- Explicit and implicit synchronizations
- Simple reductions
- Nested parallelism
- OpenMP tasking
New features in OpenMP 4.0 (released in July 2013) include:
- Device constructs for accelerators
- SIMD constructs for vectorization
- Task groups and dependencies
- Thread affinity control
- User defined reductions
- Cancellation construct
- Initial support for Fortran 2003
- OMP_DISPLAY_ENV for all internal variables
Relevant NERSC Trainings on OpenMP:
- OpenMP Basics and MPI/OpenMP Scaling. Helen He. LBNL Computational Sciences Postdocs Training, Mar 2015.
- Intel OpenMP Training at NERSC (part 1, part 2, part 3, part 4). Jeongnim Kim, Intel. March 2015.
- Explore MPI/OpenMP Scaling on NERSC Systems. Helen He, NERSC Training, October 2014.
- Hybrid MPI/OpenMP Programming. Helen He, NERSC User Group Training, Feb 2013.
- Introduction to OpenMP. Matt Cordery, NERSC User Group Training, Feb 2013.
- Using Hybrid MPI/OpenMP, UPC, and CAF at NERSC. Helen He and Woo-Sun Yang, NERSC User Group Training, Feb 2012.
- Introduction to OpenMP. Helen He, NERSC Using the Cray XE6 Training, Feb 2011.
Below are a collection of some useful OpenMP resources and tutorials:
- Official OpenMP Web Site: OpenMP standards, API specifications, tutorials, forums, and a lot more other information and resources.
- Tim Mattson's (Intel) "Introduction to OpenMP" (2013) on YouTube: 27 video segments, 4 hrs total. slides, exercises.
- OpenMP Basics and MPI/OpenMP Scaling
- SC13 Tutorial: Hybrid MPI and OpenMP Parallel Programming.
- LLNL OpenMP Tutorial. Blaise Barney, LLNL
- ANL Training Program on Exascale Scale Computing, 2013
- Using OpenMP for Intranode Parallelism: Tutorial Overview. Bronis de Supinski, LLNL; Paul Petersen, Intel.
- Using OpenMP for Intranode Parallelism: Useful Information. Bronis de Supinski, LLNL; Paul Petersen, Intel.
- Using OpenMP for Intranode Parallelism. OpenMP 4.0 and the Future of OpenMP. Bronis de Supinski, LLNL.
- UC Berkeley ParLab Boot Camp, 2013
- Shared Memory Programming with OpenMP - Basics, Tim Mattson, Intel. Video.
- More about OpenMP - New Feature. Tim Mattson, Intel. Video (start from 1:25:20).
- CSCS Getting the Best Out of Multi-Core Training, Dec 2012