基于模拟退火离散粒子群算法的芯片堆叠热布局优化

Thermal Placement Optimization of Chip Stacking Based on Discrete Particle Swarm Algorithm with Simulated Annealing

针对芯片堆叠的热布局优化问题,使用热叠加模型,结合热传导公式,以所有芯片温度的最高值作为评价指标,确定出三维芯片堆叠热布局优化的适应度函数;然后采用模拟退火离散粒子群算法对芯片热布局进行优化,得到优化后的芯片堆叠布局方案,并用Icepak软件对优化后的布局进行仿真验证。仿真结果表明:采用模拟退火离散粒子群算法对三维芯片堆叠进行热布局优化可以使温度分布更加均匀,最高温度明显降低。

Studies the theory of thermal placement optimization of stacked chips. Combines the thermal superposition model and the discrete particle swarm algorithm, selects the highest temperature of all chips as evaluation index, determines the fitness function of thermal layout optimization, and gets the corresponding thermal layout optimization program. Uses the Icepak software to verify the optimized layout. The verification results show that changing the placement scheme of the stacked chip can reduce the maximum temperature and the average temperature.