Parallel lexical analysis and parsing on the AMT distributed array processor

Abstract

I discuss an implementation of a parallel lexical analyser and parser for the AMT Distributed Array Processor (DAP). I show that by utilising the two dimensional topology of the DAP it is possible to improve the complexity of the Hillis & Steele [5,6] lexing algorithm from logarithmic squared time to logarithmic time in relation to the input length. I also show that the conclusions produced by Barnard & Skillicorn [1] that languages generated by LL(1) grammars can be parsed in logarithmic time on SIMD system is incorrect, and only a subset of the languages generated can be parsed in at least logarithmic time. © 1992.