Scholarlink

<p>Subnet-based wireless network-on-chip (WNoC) architecture often experiences inefficient core utilization, leading to localized overheating and performance loss. This study introduces a hop-based approach that selects cores for wireless router assignment without forming subnets and schedules jobs with fairness by assigning tasks to the coolest cores. This approach aims to maintain balanced heat distribution across WNoC cores while improving performance. Wireless router cores are chosen so that the maximum hop count from any core to the nearest wireless router remains fixed. Jobs are scheduled using an improved shortest remaining job first (SRJF) algorithm for fairness, and cores are selected based on their lowest Heat Indicator (HI) values. We simulate 49-, 81-, and 144-core WNoC in 1-hop, 2-hop, and 3-hop configurations using benchmark-inspired 184 jobs. Simulation results show that the 2-hop WNoC achieves fair heat distribution and provides a balanced performance across the evaluated metrics, reducing the number of wireless routers compared to hop-1 and lowering latency and power consumption compared to existing subnet-based WNoC systems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A promising approach for better thermal management and performance of WNoC systems

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月26日

查看全文
  • Abu Asaduzzaman,
  • Koteswara R. Pandi,
  • Christian C. Thompson,
  • Md R. Uddin

摘要

Subnet-based wireless network-on-chip (WNoC) architecture often experiences inefficient core utilization, leading to localized overheating and performance loss. This study introduces a hop-based approach that selects cores for wireless router assignment without forming subnets and schedules jobs with fairness by assigning tasks to the coolest cores. This approach aims to maintain balanced heat distribution across WNoC cores while improving performance. Wireless router cores are chosen so that the maximum hop count from any core to the nearest wireless router remains fixed. Jobs are scheduled using an improved shortest remaining job first (SRJF) algorithm for fairness, and cores are selected based on their lowest Heat Indicator (HI) values. We simulate 49-, 81-, and 144-core WNoC in 1-hop, 2-hop, and 3-hop configurations using benchmark-inspired 184 jobs. Simulation results show that the 2-hop WNoC achieves fair heat distribution and provides a balanced performance across the evaluated metrics, reducing the number of wireless routers compared to hop-1 and lowering latency and power consumption compared to existing subnet-based WNoC systems.