A large context multithreaded architecture

Abstract

Multithreaded architectures synthesize von Neumann computation model with data-driven evaluation to take advantage of both computation models. In this paper we propose the design and performance evaluation of a Large Context Multithreaded (LCM) architecture. The salient features of the proposed architecture are: (i) a layered approach to synchronization and scheduling of the three levels of program hierarchy, (ii) support for a large resident context to improve locality, and (Hi) a novel high-speed buffer organization to ensures 100% data availability. Initial simulation results indicate that the proposed architecture is capable of sustaining very high processor utilization.