Translators for Parallel Computer Systems

Principal Investigator: Ast/P Alfred Heng

Starting Date: 1998-01

Duration: NIL

Overview:

It is widely accepted today that parallel computing is the only plausible way to achieve the computational power required in an increasingly demanding world. Shared memory systems have existed for many years, and their supporting compilers are quite mature. However, scalability of such systems are limited. Alternatives are distributed memory systems and heterogeneous systems, both of which are highly scalable. Unfortunately, programming these systems is a difficult task. many systems support only message-passing programming model, and implemented using a language which is an extension of FORTRAN or C. Programs written for these machines tend to be machine-specific and hence are not portable to other machines. We are currently working on a compiler for MIMD distributed memory systems to support machine-independent programming; at a later stage, we intend to extend this compiler for use in a heterogeneous environment.