Chapter 5

Distributed memory programming

Abstract

What makes distributed memory programming relevant to multicore platforms, is scalability: stand-alone, shared memory, multicore systems just do not have to potential to scale beyond the performance envelope of their memory bus. Attaining even higher performance is thus, relegated to collections of systems that communicate over a network.

The Message Passing Interface (MPI) covered in this chapter, can be considered as the “glue” that makes such distributed memory systems operate as cohesive platforms. MPI facilitates both point-to-point communication, and efficient collective communication operations that improve performance and reduce program clutter. In its latest incarnation, MPI-3 also ...

Get Multicore and GPU Programming now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.