The Community for Technology Leaders
Green Image
Issue No. 03 - May-June (2013 vol. 33)
ISSN: 0272-1732
pp: 16-27
Hadi Esmaeilzadeh , University of Washington
Adrian Sampson , University of Washington
Luis Ceze , University of Washington
Doug Burger , Microsoft Research
ABSTRACT
This work proposes an approximate algorithmic transformation and a new class of accelerators, called neural processing units (NPUs). NPUs leverage the approximate algorithmic transformation that converts regions of code from a Von Neumann model to a neural model. NPUs achieve an average 2.3× speedup and 3.0× energy savings for general-purpose approximate programs. This new class of accelerators shows that significant performance and efficiency gains are possible when the abstraction of full accuracy is relaxed in general-purpose computing.
INDEX TERMS
Computer architecture, Neural networks, Approximation algorithms, Algorithm design and analysis, Accelerators, NPUs, approximate computing, accelerators, neural networks, Parrot algorithmic transformation, neural processing units
CITATION
Hadi Esmaeilzadeh, Adrian Sampson, Luis Ceze, Doug Burger, "Neural Acceleration for General-Purpose Approximate Programs", IEEE Micro, vol. 33, no. , pp. 16-27, May-June 2013, doi:10.1109/MM.2013.28
80 ms
(Ver 3.3 (11022016))