International journal of advanced smart convergence

The Institute of Internet, Broadcasting and Communication (한국인터넷방송통신학회)
권호 표지
DOI QR코드

DOI QR Code

Comparison of Traditional Workloads and Deep Learning Workloads in Memory Read and Write Operations

  • Jeongha Lee (Department of Computer Engineering, Ewha University) ;
  • Hyokyung Bahn (Department of Computer Engineering, Ewha University)
  • Received : 2023.10.18
  • Accepted : 2023.10.28
  • Published : 2023.12.31
Download PDF
⟨ Previous Next ⟩

Abstract

With the recent advances in AI (artificial intelligence) and HPC (high-performance computing) technologies, deep learning is proliferated in various domains of the 4th industrial revolution. As the workload volume of deep learning increasingly grows, analyzing the memory reference characteristics becomes important. In this article, we analyze the memory reference traces of deep learning workloads in comparison with traditional workloads specially focusing on read and write operations. Based on our analysis, we observe some unique characteristics of deep learning memory references that are quite different from traditional workloads. First, when comparing instruction and data references, instruction reference accounts for a little portion in deep learning workloads. Second, when comparing read and write, write reference accounts for a majority of memory references, which is also different from traditional workloads. Third, although write references are dominant, it exhibits low reference skewness compared to traditional workloads. Specifically, the skew factor of write references is small compared to traditional workloads. We expect that the analysis performed in this article will be helpful in efficiently designing memory management systems for deep learning workloads.

Keywords

Acknowledgement

This work was partly supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (No. 2019R1A2C1009275) and the Institute of Information & communications Technology Planning & Evaluation (IITP) grant funded by the Korea government (MSIT) (No.2021-0-02068, Artificial Intelligence Innovation Hub).

References

  1. S. Dargan, M. Kumar, M.R. Ayyagari and G. Kumar, "A survey of deep learning and its applications: a new paradigm to machine learning," Archives of Computational Methods in Engineering, vol. 27, pp. 1071-1092, 2020. DOI: https://doi.org/10.1007/s11831-019-09344-w
  2. J. Li, N. Mirza, B. Rahat and D. Xiong, "Machine learning and credit ratings prediction in the age of fourth industrial revolution," Technological Forecasting and Social Change, vol. 161, pp. 1-13, 2020. DOI: https://doi.org/j.techfore.2020.120309 https://doi.org/10.1016/j.techfore.2020.120309
  3. S. Idowu, D. Struber and T. Berger, "Asset management in machine learning: state-of-research and state-of-practice," ACM Computing Surveys, vol. 55, no. 7, pp 1-35, 2022. DOI: https://doi.org/10.1145/3543847
  4. H. Fujiyoshi, T. Hirakawa and T. Yamashita, "Deep learning-based image recognition for autonomous driving," IATSS Research, vol. 43, no. 4, pp. 244-252, 2019. DOI: https://doi.org/10.1016/j.iatssr.2019.11.008
  5. J. Xiong, D. Yu, S. Liu, L. Shu, X. Wang et al., "A review of plant phenotypic image recognition technology based on deep learning," Electronics, vol. 10, no. 1, pp. 1-19, 2021. DOI: https://doi.org/10.3390/ electronics10010081
  6. I. H. Sarker, M. M. Hoque, M. K. Uddin and T. Alsanoosy, "Mobile data science and intelligent apps: concepts, AI-based modeling and research directions," Mobile Networks and Applications, vol. 26, pp. 285-303, 2021. DOI: https://doi.org/10.1007/s11036-020-01650-z
  7. E. Lee, H. Kang, H. Bahn and K. G. Shin, "Eliminating periodic flush overhead of file I/O with non-volatile buffer cache," IEEE Transactions on Computers, vol. 65, no. 4, pp. 1145-1157, 2016. DOI: https://doi.org/10.1109/TC.2014.2349525
  8. D. T. Nguyen, H. Kim, H. J. Lee and I. J. Chang, "An approximate memory architecture for a reduction of refresh power consumption in deep learning applications," in Proc. of IEEE International Symposium on Circuits and Systems (ISCAS), Florence, Italy, pp. 1-5, 2018. DOI: https://doi.org/10.1109/ISCAS.2018.8351021
  9. S. Yoo, Y. Jo and H. Bahn, "Integrated scheduling of real-time and interactive tasks for configurable industrial systems," IEEE Transactions on Industrial Informatics, vol. 18, no. 1, pp. 631-641, 2022. DOI: https://doi.org/10.1109/TII.2021.3067714
  10. N. Nethercote and J. Seward, "Valgrind: a framework for heavyweight dynamic binary instrumentation," ACM SIGPLAN Notices, vol. 42, no. 6, pp. 89-100, 2007. DOI: https://doi.org/10.1145/1273442.1250746