Speedup of Type-1 Fuzzy Logic Systems on Graphics Processing Units Using CUDA
Department of Electronics & Communication Engineering, SBS State Technical Campus, Ferozepur-152004, (Punjab) India
International Conference on Communication, Computing and Systems, 2014
@conference{chauhan2014,
title={Speedup of Type-1 Fuzzy Logic Systems on Graphics Processing Units Using CUDA},
author={Chauhan, Durlabh and Singh, Satvir and Singh, Sarabjeet and Banga, Vijay Kumar},
booktitle={International Conference on Communication, Computing and Systems},
address={Ferozepur, Punjab, India},
pages={267–271},
month={August},
year={2014}
}
Parallelcomputing is one of significant components of the High Performance Computing (HPC) and is being used to solve problems, which are large and complex in nature. Fuzzy Logic System (FLS) is a problem that becomes computationally intensive with increase in number of inputs and/or fuzzy rules. Running an FLS is highly parallel in nature, therefore, can be implemented in parallel on GPU using CUDA. In this paper, various fuzzy computations viz. rule firing, implication, aggregation and defuzzification are performed in parallel. Multiple threads are run at a time to fire multiple fuzzy rules, simultaneously, that reduces the overall FLS execution time. It is observed from simulation results that GPU works faster as compared to CPU when either number of inputs is increased or number of fuzzy rules.
August 26, 2014 by hgpu
