The University of California at Davis and Arizona State University jointly will study a new approach to computer architecture, based on the use of demand-driven, asynchronous computation, the use of the tagged data to obviate the need for 'state' information, and the use of a certain general-purpose language as the machine language. Their approach is to design an experimental eduction engine, i.e., a highly parallel machine that uses tagged data with the demand-driven approach to asynchronous, distributed computation. A feature of the machine that distinguishes it from other current machine designs is that it will have as its machine language an existing high-level language (Lucid), that is very well suited to asynchronous computation in general, and to eduction in particular. The machine should have decidedly superior performance, since it will be designed to utilize efficiently a large and easily varied number of processors. Analysis of the behavior of the machine, and the behavior of Lucid programs, will be aimed at realizing a truly scalable machine, with the possibility of expansion of the machine to a true supercomputer. This would be accomplished for a machine that is not restricted to special applications, or to special types of algorithms, as are vector machines, for example. The design will be complete down to the level of actual hardware components, and will include a feasibility study of, and cost analysis for, actually building an experimental machine. The first two years of this research project were conducted at SRI International. Professor Ashcroft has joined the faculty at the University of California at Davis.
