TY - GEN
T1 - A heterogeneous microprocessor for energy-scalable sensor inference using genetic programming
AU - Jia, Hongyang
AU - Lu, Jie
AU - Jha, Niraj K.
AU - Yerma, Naveen
N1 - Publisher Copyright:
© 2017 JSAP.
PY - 2017/8/10
Y1 - 2017/8/10
N2 - We present a heterogeneous microprocessor for IoE sensor-inference applications, which achieves programmability required for feature extraction strictly using application data. Acceleration, though key for energy efficiency, poses substantial programmability challenges. These are overcome by exploiting genetic programming (GP) for automatic program synthesis. GP yields highly structured models of computation, enabling: (1) high degree of specialization; (2) systematic mapping of programs to the accelerator; and (3) energy scalability via user-controllable approximation. The microprocessor (130nm) achieves 325×/156× energy reduction, and farther 20x/9x energy scalability, for programmable feature extraction in two medical-sensor applications (seizure/arrhythmia-detection) vs. GP-model execution on CPU. The energy efficiency is 220 GOPS/W, near that of fixed-function accelerators, exceeding typical programmable accelerators.
AB - We present a heterogeneous microprocessor for IoE sensor-inference applications, which achieves programmability required for feature extraction strictly using application data. Acceleration, though key for energy efficiency, poses substantial programmability challenges. These are overcome by exploiting genetic programming (GP) for automatic program synthesis. GP yields highly structured models of computation, enabling: (1) high degree of specialization; (2) systematic mapping of programs to the accelerator; and (3) energy scalability via user-controllable approximation. The microprocessor (130nm) achieves 325×/156× energy reduction, and farther 20x/9x energy scalability, for programmable feature extraction in two medical-sensor applications (seizure/arrhythmia-detection) vs. GP-model execution on CPU. The energy efficiency is 220 GOPS/W, near that of fixed-function accelerators, exceeding typical programmable accelerators.
UR - http://www.scopus.com/inward/record.url?scp=85034044499&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85034044499&partnerID=8YFLogxK
U2 - 10.23919/VLSIC.2017.8008535
DO - 10.23919/VLSIC.2017.8008535
M3 - Conference contribution
AN - SCOPUS:85034044499
T3 - IEEE Symposium on VLSI Circuits, Digest of Technical Papers
SP - C28-C29
BT - 2017 Symposium on VLSI Circuits, VLSI Circuits 2017
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 31st Symposium on VLSI Circuits, VLSI Circuits 2017
Y2 - 5 June 2017 through 8 June 2017
ER -