《计算机应用》唯一官方网站 ›› 2025, Vol. 45 ›› Issue (8): 2646-2655.DOI: 10.11772/j.issn.1001-9081.2024081092

• 先进计算 • 上一篇    

基于多层感知机-注意力模型的功耗预测算法

敬超1,2, 全育涛1, 陈艳1,2()   

  1. 1.桂林理工大学 计算机科学与工程学院,广西 桂林 541006
    2.广西嵌入式技术与智能系统重点实验室(桂林理工大学),广西 桂林 541006
  • 收稿日期:2024-08-05 修回日期:2024-10-20 接受日期:2024-10-31 发布日期:2024-11-19 出版日期:2025-08-10
  • 通讯作者: 陈艳
  • 作者简介:敬超(1983—),男,河南长葛人,教授,博士,CCF高级会员,主要研究方向:高性能计算、智能优化算法
    全育涛(2000—),男,湖南衡阳人,硕士研究生,主要研究方向:高性能计算
  • 基金资助:
    国家自然科学基金资助项目(62362018);广西重点研发计划项目(桂科AB23075116)

Improved multi-layer perceptron and attention model-based power consumption prediction algorithm

Chao JING1,2, Yutao QUAN1, Yan CHEN1,2()   

  1. 1.College of Computer Science and Engineering,Guilin University of Technology,Guilin Guangxi 541006,China
    2.Guangxi Key Laboratory of Embedded Technology and Intelligent System (Guilin University of Technology),Guilin Guangxi 541006,China
  • Received:2024-08-05 Revised:2024-10-20 Accepted:2024-10-31 Online:2024-11-19 Published:2025-08-10
  • Contact: Yan CHEN
  • About author:JING Chao, born in 1983, Ph. D., professor. His research interests include high-performance computing, intelligent optimization algorithms.
    QUAN Yutao, born in 2000, M. S. candidate. His research interests include high-performance computing.
  • Supported by:
    National Natural Science Foundation of China(62362018)

摘要:

虽然异构计算系统的应用可以加快神经网络参数的处理,但系统功耗也随之剧增。良好的功耗预测方法是异构系统优化功耗和处理多类型工作负载的基础,基于此,通过改进多层感知机-注意力模型,提出一种面向CPU/GPU异构计算系统多类型工作负载的功耗预测算法。首先,考虑服务器功耗与系统特征,建立一种基于特征的工作负载功耗模型;其次,针对现有的功耗预测算法不能解决系统特征与系统功耗之间的长程依赖的问题,提出一种改进的基于多层感知机-注意力模型的功耗预测算法Prophet,该算法改进多层感知机实现各个时刻的系统特征的提取,并使用注意力机制综合这些特征,从而有效解决系统特征与系统功耗之间的长程依赖问题;最后,在实际系统中开展相关实验,将所提算法分别与MLSTM_PM (Power consumption Model based on Multi-layer Long Short-Term Memory)和ENN_PM (Power consumption Model based on Elman Neural Network)等功耗预测算法对比。实验结果表明,Prophet具有较高的预测精准性,与MLSTM_PM算法相比,在工作负载blk、memtest和busspd上将平均相对误差(MRE)分别降低了1.22、1.01和0.93个百分点,并且具有较低的复杂度,表明了所提算法的有效性及可行性。

关键词: 异构计算系统, 负载特征, 多层感知机, 注意力机制, 功耗预测

Abstract:

Although the use of heterogeneous computing systems can accelerate the processing of neural network parameters, it also increases system power consumption significantly. Good power consumption prediction methods are fundamental for optimizing power consumption in heterogeneous systems and handling multi-type workloads. Based on the above, by improving multi-layer perceptron and attention model, a power consumption prediction algorithm was proposed for CPU/GPU heterogeneous computing systems with multi-type workloads. Firstly, considering server power consumption and system features, a workload power consumption model based on features was established. Then, to address the issue that the existing power consumption prediction algorithms cannot solve long-range dependence between system features and system power consumption, an improved power consumption prediction algorithm based on multi-layer perceptron-attention model was proposed, namely Prophet. In the algorithm, the multi-layer perceptron was modified to extract system features at different moments, and the attention mechanism was employed to synthesize these features, so that the long-range dependency problem between system features and power consumption was solved effectively. Finally, the experiments were conducted on real heterogeneous systems, and the proposed algorithm was compared with the power consumption prediction algorithms such as MLSTM_PM (Power consumption Model based on Multi-layer Long Short-Term Memory) and ENN_PM (Power consumption Model based on Elman Neural Network). Experimental results show that Prophet achieves higher prediction accuracy, reducing the Mean Relative Error (MRE) for workloads blk, memtest, and busspd by 1.22, 1.01, and 0.93 percentage points, respectively, compared to MLSTM_PM, and has low complexity, indicating the proposed algorithm’s effectiveness and feasibility.

Key words: heterogenous computing system, workload feature, Multi-Layer Perceptron (MLP), attention mechanism, power consumption prediction

中图分类号: