Coherent parallel C

Abstract

Coherent Parallel C (CPC) is an extension of C for parallelism. The extensions are not simply parallel for loops; instead, a data parallel programming model is adopted. This means that one has an entire process for each data object. An example of an “object” is one mesh point in a finite element solver. How the processes are actually distributed on a parallel machine is transparent—the user is to imagine that an entire processor in a distributed-memory environment is dedicated to each process. This simplifies programming tremendously: complex if statements associated with domain boundaries disappear; problems which do not exactly match the machine size and irregular boundaries are all handled transparently. The usual communication calls are not seen at all at the user level. Variables of other processes (which may or may not be on another processor) are merely accessed (global memory). The first pass of the CPC compiler schedules the necessary communications in an efficient, coherent manner. Processes in CPC are insulated from one another and interact in a deterministic manner. This allows tractable debugging. Standard C I/O is provided, with simple extensions for parallelism. We currently have a CPC runtime system implemented on an NCUBE and have started implementing a true compiler for the language. CPC is not specific to distributed memory machines. Implementation of this language on other architectures is natural—for example, there seems to be no fundamental problem with CPC on shared-memory parallel computers.