Abstract:Chiplet technology has emerged as a promising solution to address the increasing demand for high-performance computing in light of the slowdown of Moore's law. While chiplet-based multicore systems offer higher performance through heterogeneous integration, they also pose challenges for power delivery system (PDS) design. The integration of additional vertical and inter-chiplet connections, along with higher power density, impose stringent requirements on power delivery. Moreover, PDS efficiency is affected by workload variations at runtime, necessitating the need to design and manage PDSs and processors as a whole to improve system energy efficiency while balancing performance. In this article, we propose an offline-online co-design optimization methodology that combines offline PDS design optimization with online power management. To address the power consumption and delivery mismatch, we introduce a centralized deep Q-network (DQN)-based online control scheme for power co-management in chiplet-based multicore systems. By carefully designing the state space and reward functions, our approach achieves workload-aware adaptive control to reduce the energy-delay-product (EDP) while maintaining PDS efficiency under a given performance target (PT). We conduct evaluations on realistic applications to validate the effectiveness of our approach. For 64-core systems, our method achieves an average EDP reduction of 67% while meeting a 90% PT, surpassing state-of-the-art modular Q-learning (MQL)-based and heuristic-based approaches by up to 4% and 16%, respectively. Additionally, our approach demonstrates wiser action selection policies, higher control stability, and lower implementation overhead compared to the MQL-based approach.

Heuristic on a Novel Power Management System Cooperating with Compiler

Power Reduction Through Measurement and Modeling of Users and CPUs: Summary

Multi-core Chip Dynamic Power Management Framework Based on Reinforcement Learning br

Coordinating System Software for Power Savings

Optimizing the Energy Efficiency of Power Supply in Heterogeneous Multicore Chips with Integrated Switched-Capacitor Converters

Power Profiling for Low Power Optimization Using New Computing Model

Using dynamic and Static Approach with Compiler-Assisted in Power management for DVS-Enabled Embedded System

Performance-Aware Power Management in Embedded Controllers with Multiple-Voltage Processors

A power-aware system combining static compiling and DVFS

A User-Configured Low Power Optimizing Mechanism on Embedded System

Power-Aware Run-Time Scheduler for Mixed-Criticality Systems on Multi-Core Platform

Exploiting Power-Area Tradeoffs In High-Level Synthesis Through Dynamic Functional Unit Allocation

Dynamic Power Management and Dynamic Voltage Scaling in Real-Time Cmp Systems

An Overview of Power Optimization in High Level Synthesis

A Novel Distributed Control and Its Tolerance Analysis for Microprocessor Power Management

Deep Reinforcement Learning-Based Power Management for Chiplet-Based Multicore Systems

Efficient Method of Energy Saving on Variable Voltage Multi-core Processor

Energy Efficiency Optimization Scheme for Multi-core ProcessorUnder Power Consumption Constraint

Energy Saving Compiler Framework On Embedded Systems

Practical Strategies for Power-Efficient Computing Technologies

Using Nonvolatile Processors to Reduce Leakage in Power Management Approaches