中文摘要：

为了更高效地实现多核片上系统（MPSoC）温敏布图设计和实时功耗温度管理，采用自下而上的建模方法，提出MPSoC结构级热分析方法．首先采用HotSpot热分析软件提取功能模块之间的相关热阻参数；然后基于这些参数，提出模块级方法BloTAM、核级方法CorTAM和考虑本核内模块相互影响的改良核级方法BiCorTAM3种具有不同复杂度与精度的热分析方法，它们均具有简单、高效、与现有简化模型兼容、易于扩展、能够解决考虑温度对漏电流的影响等优点．实验结果表明，对MPSoC进行热分析时，相对于HotSpot热分析软件，BloTAM和BiCorTAM方法都可以在保证精度的前提下大幅度提高热分析的速度，其中局部热点的温度增量平均误差可以控制在3％以下，热分析的速度实现了50倍以上的分析加速；两者均可作为理想的结构级热分析方法．

英文摘要：

In order to efficiently implement temperature-aware floorplan design and dynamic power and temperature management （DPTM） for multi processor system-on-chip （MPSoC）, this paper adopts bottom-up modeling method so as to propose architecture-level MPSoC thermal analysis methods. First equivalent thermal resistances between functional modules were extracted with HotSpot temperature analysis software; then, based on these parameters, we studied three analysis methods with different accuracy and algorithm complexity： block-level temperature analysis method （BloTAM）, core-level temperature analysis method （CorTAM） and block improved core temperature analysis method （13iCorTAM）. Experimental results show that, comparing to HotSpot, both 131oTAM and BiCorTAM substantially reduce the total consumed time for MPSoC thermal analysis with guarantee of accuracy： speedup higher than 50 times is achieved with average temperature delta error as low as 3%.