@article{Tao2018, 
author = {Jin-Hua Tao and Zi-Dong Du and Qi Guo and Hui-Ying Lan and Lei Zhang and Sheng-Yuan Zhou and Ling-Jie Xu and Cong Liu and Hai-Feng Liu and Shan Tang and Allen Rush and Willian Chen and Shao-Li Liu and Yun-Ji Chen and Tian-Shi Chen},
title = {BENCHIP: Benchmarking Intelligence Processors},
year = {2018},
journal = {Journal of Computer Science and Technology},
volume = {33},
number = {1},
pages = {1-23},
keywords = {deep learning, benchmark, intelligence processor},
url = {https://www.sciopen.com/article/10.1007/s11390-018-1805-8},
doi = {10.1007/s11390-018-1805-8},
abstract = {The increasing attention on deep learning has tremendously spurred the design of intelligence processing hardware. The variety of emerging intelligence processors requires standard benchmarks for fair comparison and system optimization (in both software and hardware). However, existing benchmarks are unsuitable for benchmarking intelligence processors due to their non-diversity and nonrepresentativeness. Also, the lack of a standard benchmarking methodology further exacerbates this problem. In this paper, we propose BENCHIP, a benchmark suite and benchmarking methodology for intelligence processors. The benchmark suite in BENCHIP consists of two sets of benchmarks: microbenchmarks and macrobenchmarks. The microbenchmarks consist of single-layer networks. They are mainly designed for bottleneck analysis and system optimization. The macrobenchmarks contain state-of-the-art industrial networks, so as to offer a realistic comparison of different platforms. We also propose a standard benchmarking methodology built upon an industrial software stack and evaluation metrics that comprehensively reflect various characteristics of the evaluated intelligence processors. BENCHIP is utilized for evaluating various hardware platforms, including CPUs, GPUs, and accelerators. BENCHIP will be open-sourced soon.}
}