| Due to multi-thread parallel processing, the significant rise of performance and efficiency, multi-processor instead of single-processor rose as the main stream processor. However, high temperature and power consumption threat regarding reliability and lifetime of processors, has become the bottleneck of multi-processor development.Dynamic Voltage and Frequency Scaling (DVFS) and clock gating technology are used to solve this problem in hardware. However these methods cause serious expense of processor performance, and cannot have wide range of applications. To lower power consumption, the higher level of the methods, the effect is more pronounced. Today, more and more people pay attention to software level to control temperature and power consumption.This paper has proposed thermal task scheduler in the operating system level, which is implemented through Round Robin (RR) scheduling algorithm assigning time slices to each process for equal portions and circular order without priority. The key implementation of the RR Algorithm is to balance temperature among a group of processors to avoid the peak temperature at each interruption. Experiments showed that the temperature of microprocessors will significantly fall with a modest performance loss. At the same time, the paper extends our scheduling algorithm to many-core, which has a constructive significance on the development of many-core architecture. |