A logarithmic approach to energy-efficient GPU arithmetic for mobile devices

Author

Lastras, Miguel ; Parhami, Behrooz

Author_Institution

Dept. of Electr. & Comput. Eng., Univ. of California, Santa Barbara, Santa Barbara, CA, USA

fYear

2013

fDate

3-6 Nov. 2013

Firstpage

2177

Lastpage

2180

Abstract

Graphic processing units (GPUs) have emerged as useful components in the realization of high-performance and cost-effective digital systems for numerically intensive applications, from simple personal devices to large-scale supercomputers. In this paper, we consider GPUs that incorporate energy-efficient logarithmic arithmetic units. After analyzing numerical errors arising from such an implementation scheme, we present a simple fine-tuning of the design to reduce the worst-case error.

Keywords

digital arithmetic; graphics processing units; power aware computing; cost-effective digital systems; energy-efficient GPU arithmetic; energy-efficient logarithmic arithmetic units; graphic processing units; high-performance digital systems; large-scale supercomputers; logarithmic approach; mobile devices; numerical error analysis; personal devices; worst-case error reduction; Computers; Energy efficiency; Graphics; Graphics processing units; Hardware; Mobile handsets; Arithmetic/logic unit; Error compensation; GPU-based computing; Logarithmic arithmetic unit; Low-power design;

fLanguage

English

Publisher

ieee

Conference_Titel

Signals, Systems and Computers, 2013 Asilomar Conference on

Conference_Location

Pacific Grove, CA

Print_ISBN

978-1-4799-2388-5

Type

conf

DOI

10.1109/ACSSC.2013.6810695

Filename

6810695