09 Nov Machine Learning for Optimized Computer Architecture Designs: Cache design
As technologiesare scaled, computer architecture designs face serious challenges including scalingperformance, minimizing power, providing reliability, and guaranteeing Quality of Service(QoS) within the power, area, and cost constraints. The critical objective of moderncomputer architecture designs is to tackle these challenges at the same time. This ischallenging since the trade-offs are everywhere. For example, using error checking unitsto ensure reliability can result in the increasing of power and latency, and low powerdesigns can slow down the system. Manually designing rules for optimizing suchcompeting trade-offs is nearly impossible, given the combinatorial size of the decisionspace and the complexity of the interactions between these techniques. Machine learning(ML), on the other hand, can be used to automatically infer complex decisions in order tooptimize designs.Can you find some applications where you can obtain higher efficiency by using one of themachine learning algorithms? Can you think of a way of applying ML to design not onlyfor performance but also for power efficiency at the same time? How will you implementthe architecture? What are the design issues of building such an architecture?
[1] Wang, Ke, Ahmed Louri, Avinash Karanth, and Razvan Bunescu. “IntelliNoC: A HolisticDesign Framework for Energy-efficient and Reliable On-chip Communication for Manycores.”In ACM/IEEE 46th Annual International Symposium on Computer Architecture (ISCA), pp. 1-12.2019.
[2] Zheng, Hao, and Ahmed Louri. “An Energy-efficient Network-on-Chip Design usingReinforcement Learning.” In Design Automation Conference (DAC), pp. 1-6. 2019.
[3] Jieming Yin, Subhash Sethumurugan, Yasuko Eckert, Chintan Patel, Alan Smith, EricMorton, Mark Oskin, Natalie Enright Jerger, and Gabriel H Loh. Experiences with ML-drivenDesign: A NoC Case Study. In IEEE International Symposium on High-Performance ComputerArchitecture (HPCA), 2020.
[4] Wu, Nan, and Yuan Xie. “A Survey of Machine Learning for Computer Architecture andSystems.” arXiv preprint arXiv:2102.07952 (2021).
