SpotLight: A Hotspot-Greedy, Light-Weighted, and Automated Thermal Modeling Framework for Early Smartphone Design

Chin-Wei Wu1, Yu-Min Lee2, Pei-Yu Huang2, Bo-Jiun Yang1, Tai-Yu Chen1, Ting-Chang Huang1, Yen-Lin Lee1
1MediaTek Inc., 2National Yang Ming Chiao Tung University


Abstract

This paper develops a hotspot-greedy, light-weighted, and automated framework, SpotLight, to achieve fast thermal analysis of smartphones at the pre-silicon stage. Besides, we apply superposed models and a designed least penalty problem to improve the accuracy. Experimental results show that the SpotLight can accurately simulate the system-level hotspot temperatures of smartphones with average runtime of 0.56 ms.