Region-level approximate computation reuse for power reduction in multimedia applications
ISLPED '05 Proceedings of the 2005 international symposium on Low power electronics and design
Minimal Multi-threading: Finding and Removing Redundant Instructions in Multi-threaded Processors
MICRO '43 Proceedings of the 2010 43rd Annual IEEE/ACM International Symposium on Microarchitecture
Hi-index | 0.00 |
It has been observed that even highly optimized software programsperform "redundant" computations during their execution,due to the nature (statistics) of the values assumed by input orinternal program variables. For embedded software running onbattery-powered systems, such computations can be viewed asunnecessary energy overheads, and hence represent opportunitiesfor improvement in energy efficiency. We present a systematicmethodology to identify and eliminate redundancies in the computationsperformed by embedded software programs, by exploitingopportunities that dynamically arise for computation reuse.We report the results of experiments on two different embeddedsystems - a detailed simulation model of a Fujitsu SPARClitebased embedded system, and actual current measurements on aCompaq iPAQ PDA. Our results demonstrate that the proposedtechnique can reduce energy by up to 46.9% (average of 21.2%and 13.9% for the SPARClite based system and the iPAQ, respectively)while simultaneously improving performance by upto 45.8% (average of 20.7% and 16.8% for the SPARClite basedsystem and the iPAQ, respectively), compared to well-optimizedprograms that do not employ such a technique.