Solving parallel problems by OTMP model

Abstract

Since the early stages of parallel computing, one of the most common solutions to in troduce parallelism has been to extend a sequential language with some sort of parallel version of the for construct, commonly denoted as forall construct. Although similar syntax, these forall loops di er in their semantics and implementations. The High Performance Fortran (HPF) and OpenMP versions are, likely, among the most popular. This paper presents yet another forall loop extension for the C language. In this work, we introduce a parallel computation model: One Thread Multiple Processor Model (OTMP). This model proposes an abstract machine, a programming model and cost model. The programming model de nes another forall loop construct, the theorical machine aims for both homogeneous shared and distributed memory computers, and the cost model allo ws the prediction of the performance of a program. OTMP does not only in tegrates and extends sequential programming, but also includes and expands the message passing programming model. The model allows and exploits any nested levels of parallelism, taking advan tage of situations where there are several small nested loops.