The Research Of Deadline Assignment For Real-time End-to-end Tasks In NoC

With the rapid development of chip and semiconductor technology, more and more functions are integrated into a system on a chip (SoC). The traditional bus structure has been unable to meet the growing communication needs. As network on a chip (NoC) uses computer network for reference and it has features such as higher bandwidth, non-blocking concurrent exchange, packet switching of higher utilization, reliable layered protocols, low-power point-to-point transmission and global asynchronous-synchronous scalable architecture, it has been a new communication structure in multi-core field. Meanwhile it provides efficient inter-core communication with linking the routing nodes and multiple cores.Although NoC was put forward in 2002 and its relevant issues had been studied by academia, there are no academics involved in the field of real-time communication. At present numerous research about end-to-end tasks have been carried out in distributed real-time systems, but many issues about real time have not been commenced in multi-core embedded real-time systems. With the rapid development of multimedia services and real-time tasks in the car, the deadline assignment for real-time end-to-end tasks will be an important issue in NoC.This paper will carry out the research from the aspect of deadline assignment for end-to-end tasks in NoC. As BFair was used to schedule the computing tasks and POGen was adopted to schedule the communicating tasks, three constraints were put forward to guarantee the schedulability of end-to-end tasks. This paper will design two heuristic algorithms about deadline assignment for end-to-end tasks to meet the above constraints.This paper carried out a large number of experiments basing on the above heuristic algorithms. At first, this paper analyzed four different impact factors:system resource utilization umax, the value of G in POGen, the scale of NoC and the number of end-to-end tasks. Then this paper compared the receiving rate of two heuristic algorithms to that of traditional PD algorithm. At last, this paper concluded that the receiving rate of two heuristic algorithms is better than that of PD algorithm and these algorithms can effectively solve this problem.