OpenMP

• What is OpenMP?
  • The OpenMP is an Application Program Interface (API) that supports shared-memory parallel programming.
• What is the shared-memory?
  • The memory space that is accessible from multiple CPUs without an operating system call involvement.
• Limitation
  • The number of CPUs on a single compute node bounds the number of threads. (←>Distributed Memory Cluster)

A Programmer’s View of OpenMP

• OpenMP is a portable, threaded, shared-memory programming specification with “light” syntax
  • Exact behavior depends on OpenMP implementation!
  • Requires compiler support (C or Fortran)
• OpenMP will:
  • Allow a programmer to separate a program into serial regions and parallel regions.
• OpenMP will not:
  • Parallelize automatically
  • Guarantee speedup

OpenMP's Execution Model

• OpenMP is built around a shared memory space and generates concurrent threads – this is how the parallelism is facilitated.
• CPUs that share memory are called “symmetric multi-processors”, or SMP for short.
• Each thread is typically run on its own processor, though it is becoming common for each CPU or “core” to handle more than one thread “simultaneously”; this is called “hyper-threading”.
• Thread (process) communication is implicit, and uses variables pointing to shared memory locations; this is in contrast with MPI, which uses explicit messages passed among processes.
• These days, most major compilers do support OpenMP directives for most platforms (IBM's XL suite, Intel, even GCC ≥ 4.2).
• A fork is when a single thread is made into multiple, concurrently executing threads.
• A join is when the concurrently executing threads synchronize back into a single thread.
• During the overall execution, threads may communicate with one another through shared variables.
• OpenMP programs essentially consist of a series of forks and joins.

Brief Tutorials for Programming Languages

• C
• C++
• Fortran