Managing Prefetchers With Deep Reinforcement Learning

Managing Prefetchers With Deep Reinforcement Learning
复制标题

通过深度强化学习管理预取器

DOI:
10.1109/lca.2022.3210397
复制
发表时间:
2022
影响因子:
2.3
通讯作者:
M. Erez
M. Erez
中科院分区:
计算机科学3区
文献类型:
--
作者:
Majid Jalili;M. Erez

文献摘要

参考文献

被引文献

相似文献

我们的目标是通过多代理强化学习方案减少多个激进预取器对共享资源(例如 LLC 和内存带宽)造成的争用。代理会找到要使用的预取器,并确定它们在执行过程中随时应采取的积极程度。为此,我们利用高度可扩展的动作分支代理,该代理具有共享网络模块,后跟多个网络分支。共享网络跟踪处理器的整体状态,而网络分支中的每个分支专注于一个特定的预取器。我们通过运行 20 个随机混合基准来训练网络,并测量 100 个未见过的混合的性能。我们的实验结果表明,与非常有竞争力的基准相比,我们提出的方法将内存带宽减少了 19%,同时提供相似的性能。
We aim to reduce contention caused by multiple aggressive prefetchers on shared resources (e.g., LLC and memory bandwidth) with a multi-agent reinforcement learning scheme. The agent finds what prefetchers to use and determines how aggressive they should be at any time during the execution. To do so, we utilize a highly scalable action branching agent that features a shared network module followed by several network branches. The shared network tracks the overall state of the processor while each branch in the network branches focuses on one specific prefetcher. We train the network by running 20 randomly mixed benchmarks, and measure the performance for 100 unseen mixes. Our experimental results show that our proposed method reduces the memory bandwidth by 19% while delivering similar performance compared to a very competitive baseline.
DOI: 10.1609/aaai.v32i1.11798
发表时间: 2017-11
期刊: ArXiv
影响因子: --
作者:
Arash Tavakoli;Fabio Pardo;Petar Kormushev
通讯作者: Arash Tavakoli;Fabio Pardo;Petar Kormushev
DOI: 10.1109/isca.2018.00018
发表时间: 2018
期刊: 2018 ACM/IEEE 45th Annual International Symposium on Computer Architecture
影响因子: --
作者:
Kondguli, Sushant;Huang, Michael
通讯作者: Huang, Michael