Granular Lucid (or GLU) is a coarse-grain dataflow language for programming conventional parallel computers. It is based on Lucid (circa 1994) which is an implicitly parallel, multidimensional dataflow language. A GLU program is a Lucid program with imperativelydefined data functions and data types. In this paper, we briefly describe a system for coarse-grain parallel programming based on GLU. We discuss the expressiveness of GLU in composing different kinds of parallel programs. We also discuss the efficiency with which parallelism in GLU programs can be exploited on conventional parallel computers.

Recent developments in shared-memory multiprocessor systems advocate using off-the-shelf hardware to provide basic communication mechanisms and using software to implement cache coherence policies. The exposure of communication mechanisms to software opens many opportunities for enhancing application performance. In this paper we propose a set of communication primitives that are absent from pure cache coherent schemes. The communication primitives, implemented on a communication co-processor, introduce a flavor of message passing and permit protocol optimization, without sacrificing the simplicity of the shared memory systems. To assess the overhead of the software implementation of the primitives and protocols, we compare, via simulation, the execution of three programs from the SPLASH-2 suite on four environments: a PRAM model, a hardware cache coherence scheme, a software scheme implementing only the basic cache coherence protocol, and an optimized software solution supporting the ...

We describe our experiences with a very high-level parallel composition language (called GLU) that enables rapid construction of parallel applications using sequential building blocks (extracted from existing sequential applications) and their execution on diverse parallel computer systems. GLU is sufficiently rich to succinctly express different forms of parallelism --- from function parallelism to data parallelism and from pipeline parallelism to tree parallelism. We show by example how a typical sequential application can be converted to a parallel one in GLU and executed on different parallel systems. We also show how GLU has been used to convert two widely used, sequentially written, inherently parallel workstation applications --- the make utility and a raytracing system --- to parallel equivalents that can then be run much faster on a network of workstations. 1 Introduction The idea of composing new applications from existing ones is commonplace in various Unix 1 shells in w...