Research On STDS Algorithm Designing Based On Dynamic Task Scheduling

The main way to improve the performance of a traditional single-core processor is to raise it's basic frequency or instruction level parallelism,but the problem such as power consumption increasing and electricity leaking also comes out with basic frequency raising,and the instruction level parallelism is also limited due to data hazard,control hazard and structure hazard in instructions,so it is hard to enhance the performance any more for single-core processor.Now,the mainstream processor is multi-core because it's advantage of higher speed,lower power and higher parallelism,and drawing big users and researchers.As the multiple cores,each core should be in full use to improve whole performance.Task scheduling for multi-core processor have been proved to be a NP complete problem,and that single-core processor's is not appropriate,so the question that how to make the best of multi-core processor by efficient task scheduling has been a hotspot of processor research.The paper do research in dynamic task scheduling that who has higher cost-effective then static for heterogeneous multi-core processor,and propose a dynamic scheduler model named DSMDQ by analyzing the principle and implementation of existing dynamic task scheduling algorithms,and utilizing load balancing strategy.By using centralized scheduling mode,DSMDQ can control the scheduling of every core easily.It sets the maximum and minimum range of core's load according to its capability,to control the load at the same level,and then achieve the effect of load balancing.With the mixed scheduling list of global and native,the scheduler can overcome performance bottleneck of centralized scheduling mode while reducing the scheduling frequency to save time.What's more,STDS,a new dynamic task scheduling algorithm is designed for heterogeneous multi-core processor base on DSMDQ.It is a heuristic algorithm using priority list.When scheduling,waiting time of task,communications between tasks and core's load will be considered together to pick up most appropriate task,and the importance of each element can be adjusted by assigning another value to Real-time factor and load factor,so the STDS algorithm can adapt to various environment.At last,the paper verify the SDTS algorithm by picking up scheduling time,load balancing and waiting time as the evaluation indicator,creating different random task graph as the test case,and observing executive results of algorithm on Simics simulator,and then statistical analysis the experimental data.The analysis results show that the STDS algorithm is more efficient in system with more cores,also keeping a good load balancing maintaining the speed.