Architectural directions for microsequencers

Abstract

Current microsequencers, although very impressive compared to early microprogrammable control units, lack vital features. This is largely due to the increased understanding of concurrent computer architectures (Cray, Star, MPP, systolic arrays, array processors, novel digital signal processing devices, and the attendant parallel and pipelining possibilities of such rich architectures. In this paper we explore the requirements of future microsequencers for controlling such organizations without incurring overhead device costs of peripheral controller logic (encoders, decoders, bit steerers). A number of new attributes is proposed for these new controller chips, along with justifications for each property. We conclude that current VLSI technologies can already support the fabrication of these new microsequencers.