Real-time Task Scheduling Strategies For Multimedia Stream Computing On Heterogeneous Multicore Systems

With the increasingly rich and popularization of embedded mobile multimediaapplications, keeping real-time of the intelligent terminal is greatly prominent. To ensurethe processing performance of intelligent terminal, the rapid development of processorsmake real-time task scheduling on multi-core processors also become hot research in thefield of computer technology. The focus of the current study is that, while a large numberof parallel tasks running, you need to balance the real-time requirements of the task andefficiency of the system, and decrease the executing time of the whole task as well asensure the entire system load balancing. Designing efficient multimedia stream computingtask scheduling strategies is the most direct and effective way to improve performance ofheterogeneous multicore processors system, which has important research significance.Task allocation and scheduling problems on heterogeneous multi-core systems havebeen proved to be NP-complete problem. Generally, using heuristic algorithm to solve thiskind of problems can obtain a better solution. Accordingly, for the video decoding, whichis a typical multimedia stream computing application, a new scheduling algorithm basedon ant colony algorithm is proposed to solve the real-time task scheduling problem inheterogeneous multi-core processors systems. Firstly, to realize the algorithm, a systemmodel is created in order to simulate actual heterogeneous multi-core environment. Thensome algorithm formulas are designed, including the formula of task selection, the formulaof processor selection, as well as task pheromone and processor pheromone updateformulas. At last, the implementation of algorithm is elaborated, including the selection oftask and the selection of processor. In order to meet the real-time requirements of the taskson heterogeneous multi-core systems, the task arrival time, estimation computation andcut-off time etc. is considered. To simulate heterogeneous multi-core environment, thedifference of execution speed among kernels as well as the communication bandwidth between the kernels are considered. The algorithm can better meet constraint relations ofexecution precedence among tasks as well as the requirements of the deadline, and obtaina better total length of the task scheduling.Use Matlab7.0to achieve this algorithm, in order to assess the performance of theabove real-time task scheduling algorithm based on ant colony in heterogeneousmulti-core systems, and discuss the values of the parameters. The experiments show thatthe algorithm can be used to solve the problem of real-time task scheduling inheterogeneous multi-core systems, and gets good convergence.