PRELUDE: A System for Portable Parallel Software

In this paper we describe Prelude, a programming language and accompanying system support for writing portable MIMD parallel programs. Prelude supports a methodology for designing and organizing parallel programs that makes them easier to tune for particular architectures and to port to new architectures. It builds on earlier work on Emerald, Amber, and various Fortran extensions to allow the programmer to divide programs into architecture-dependent and architecture-independent parts, and then to change the architecture-dependent parts to port the program to a new machine or to tune its performance on a single machine. The architecture-dependent parts of a program are speciied by annotations that describe the mapping of a program onto a machine. Prelude provides a variety of mapping mechanisms similar to those in other systems, including remote procedure call, object migration , and data replication and partitioning. In addition, Prelude includes novel migration mechanisms for computations based on a form of continuation passing. The implementation of object migration in Prelude uses a novel approach based on xup blocks that is more eecient than previous approaches, and amortizes the cost of each migration so that the cost per migration drops as the frequency of migrations increases. The current implementation of Prelude is built on top of Proteus, a conngurable simulator that provides both fast and accurate simulations of a wide range of MIMD architectures. Proteus itself is a useful tool for developing parallel applications, since it provides powerful non-intrusive debugging and performance monitoring capabilities that are diicult or impossible to obtain on a real machine. Much of the testing, debugging, and initial testing of an application can be accomplished using Proteus, typically with less eeort than would be required on a real machine. In addition, Proteus allows the programmer to test the scalability and portability of a program, including on a range of machine sizes and architectures not supported by available machines. We are using Proteus to develop our initial prototype of Prelude, and plan to port the implementation of Prelude to commercial and research multiprocessors in the near future.