PROJECT TITLE :

Toward Energy-Efficient Stochastic Circuits Using Parallel Sobol Sequences - 2018

ABSTRACT:

Stochastic computing (SC) typically needs long stochastic sequences and, so, a protracted latency to achieve accurate computation. The long latency leads to an inferior performance and low energy efficiency compared with most standard binary designs. During this paper, a sort of low-discrepancy sequences, the Sobol sequence, is considered for use in SC. Compared to the utilization of pseudorandom sequences generated by linear feedback shift registers (LFSRs), the use of Sobol sequences improves the accuracy of stochastic computation with a reduced sequence length. The inherent feature in Sobol sequence generators permits the parallel implementation of random range generators with an improved performance and hardware potency. In particular, the underlying theory is formulated and circuit style is proposed for an arbitrary level of parallelization during a power of two. As well, totally different methods are implemented for parallelizing combinational and sequential stochastic circuits. The hardware efficiency of the parallel stochastic circuits is measured by energy per operation (EPO), throughput per area (TPA), and runtime. At the same accuracy, the 8× parallel stochastic circuits using Sobol sequences consume approximately onep.c of the EPO of the standard LFSR-based mostly nonparallelized circuits. Meanwhile, a median of 70 (up to 89) times enhancements in TPA and less than 1p.c runtime are achieved. A sorting network is implemented for a median filter (MF) as an application. For the same Image Processing quality, a higher energy potency is obtained for an 8× parallelized stochastic MF compared with its binary counterpart.