Application specific parallel architectures

Abstract

This paper deals with the design of parallel architectures dedicated to specific applications and mainly focuses on the relationships between software and hardware tradeoffs, putting the emphasis on parallel architecture for non-numeric applications. The analysis covers both medium and massive parallelism. Medium level parallelism is seen through the analysis of VLIW architectures where code-generation takes an important role for achieving high performance. In particular we show the main design steps of a parallel architecture for Prolog. High level parallelism in non-numeric applications is introduced through the presentation of a dedicated architecture for the solution of NP-complete problems with the Boltzmann Machine model. Also in this case, we present the problems related to the design of the architecture and the interaction between hardware and software components.<>