Novel VLSI Architectures and Micro-Cell Libraries for Subscalar Computations

Abstract

Parallelism is the key to enhancing the throughput of computing structures. However, it is well established that the presence of data-flow dependencies adversely impacts the exploitation of such parallelism. This paper presents a case for a new computing paradigm namely subscalar digital arithmetic which is aimed at mitigating this issue. It proposes to break up atomic data and atomic operations thereon into sub-atomic data fragments and sub-atomic partial operations. Such a break-up exposes hitherto unexploited levels of parallelism by way of allowing overlap of operations even if data-dependent. Surprisingly this enhanced exploitation of latent parallelism comes with a favorable impact on the area-power characteristics of corresponding computing structures which is contrary to common sense. The paper also proposes a novel micro cell library with logic primitives at corresponding subscalar levels. The synthesized circuits for several sequential benchmarks show an order of magnitude improvement in their area-throughput figure-of-merit (FOM).