System and Design Technology Co-optimization of Chiplet-based AI Accelerator with Machine Learning

System and Design Technology Co-optimization of Chiplet-based AI Accelerator with Machine Learning
复制标题

基于Chiplet的AI加速器与机器学习的系统和设计技术协同优化

DOI:
10.1145/3583781.3590233
复制
发表时间:
2023
期刊:
Proceedings of the Great Lakes Symposium on VLSI
影响因子:
--
通讯作者:
Sadi, Mehdi
Sadi, Mehdi
中科院分区:
--
文献类型:
--
作者:
Mishty, Kaniz;Sadi, Mehdi

文献摘要

参考文献

被引文献

相似文献

随着先进封装技术的出现及其吸引人的特性,基于芯片的架构在芯片设计师中获得了吸引力。设计空间大,缺乏系统级和封装级协同设计方法,使得设计人员难以做出最优的设计选择。在本研究中,考虑到先进封装技术、资源分配和芯片放置的巨大设计空间,我们设计了一个优化器,寻找最大限度地提高功率、性能和面积(PPA)的设计选择,并最大限度地降低基于芯片的人工智能加速器的成本。受黑盒函数优化的贝叶斯方法的启发,我们的优化器引导搜索空间走向全局最大值,而不是随机遍历搜索空间。我们从搜索空间分析合成一个数据集,并训练一个ML模型来预测优化器建议点上我们定义的成本函数的目标值。优化器从指定的搜索空间(≥1M数据点)中以最小的迭代(≤200次迭代)和最小的运行时间找到最优设计选择。
With the availability of advanced packaging technology and its attractive features, the chiplet-based architecture has gained traction among chip designers. The large design space and the lack of system and package-level co-design methods make it difficult for the designers to create the optimum design choices. In this research, considering the colossal design space of advanced packaging technologies, resource allocation, and chiplet placement, we design an optimizer that looks for the design choices that maximize the Power, Performance, and Area (PPA) and minimize the cost of the chiplet-based AI accelerator. Inspired by the Bayesian approach for black-box function optimization, our optimizer guides the search space toward global maxima instead of randomly traversing through the search space. We analytically synthesize a dataset from the search space and train an ML model to predict the target value of our defined cost function at the optimizer-suggested points. The optimizer locates the optimum design choices from the specified search space (≥ 1M data points) with minimal iterations (≤ 200 iterations) and trivial run time.
DOI: 10.1109/isca52012.2021.00083
发表时间: 2021-06
期刊: 2021 ACM/IEEE 48th Annual International Symposium on Computer Architecture (ISCA)
影响因子: --
作者:
Zhanhong Tan;Hongyu Cai;Runpei Dong;Kaisheng Ma
通讯作者: Zhanhong Tan;Hongyu Cai;Runpei Dong;Kaisheng Ma
2.2 适用于高性能服务器和桌面产品的 AMD Chiplet 架构
DOI: 10.1109/isscc19947.2020.9063103
发表时间: 2020
期刊: 2020 IEEE International Solid- State Circuits Conference - (ISSCC)
影响因子: --
作者:
S. Naffziger;Kevin M. Lepak;Milam Paraschou;M. Subramony
通讯作者: M. Subramony
英特尔的 Ponte Vecchio GPU:架构、系统
DOI: 10.1109/hcs55958.2022.9895631
发表时间: 2022
期刊: 2022 IEEE Hot Chips 34 Symposium (HCS)
影响因子: --
作者:
Hong Jiang
通讯作者: Hong Jiang
DOI: 10.1109/vlsitechnology18217.2020.9265085
发表时间: 2020
期刊: 2020 IEEE Symposium on VLSI Technology
影响因子: --
作者:
Frank J. C. Lee;Mei Wong;J. Tzou;J. Yuan;Daniel Chang;S. Rusu
通讯作者: S. Rusu
DOI: 10.1109/micro50266.2020.00058
发表时间: 2020-09
期刊: 2020 53rd Annual IEEE/ACM International Symposium on Microarchitecture (MICRO)
影响因子: --
作者:
Sheng-Chun Kao;Geonhwa Jeong;T. Krishna
通讯作者: Sheng-Chun Kao;Geonhwa Jeong;T. Krishna