رضا زاتي گنزق

With the advent of Reconfigurable Processors and their usual improvement, These processors have become one of the most popular structures that has caused more and more processing systems are being pulled into redefined structures every day. Reconfigurable processors all have a system to modify hardware structure and during the run of the program, continually changing hardware and operating in accordance with the written program. Usually hardware tasks are considered independently, So that when entering a hardware task into the RPU area, it depends on a number of previously assigned hardware tasks and should communicate with them. In this thesis, we first examine the position of reconfigurable processing systems and It also discusses hardware and how to communicate in these systems. Then a new platform that includes multiple slots is introduced for simulation. This thesis presents an algorithm for scheduling tasks related to inputs to the system. The proposed algorithm is an online algorithm and tasks are entered into the system over time. As regards there is data communication between tasks in an input job, so the target of the proposed algorithm is to reduce data communicate overhead between tasks. To achieve the least amount of data communication overhead among tasks, the algorithm tries to assign tasks for a job in a slot and if there is not enough space in a slot, The algorithm uses other slots for scheduling and o‎r the task enters the state of the expected slot to be empty and scheduled. In the simulation process, the proposed algorithm is compared with the ground-state algorithm and evaluated and compared their performance with the base mode algorithm and evaluated and compared their performance with regard to the time of execution of the program and the task rejection ratio based on the increase of the time of execution of tasks, Communication time between tasks and width of task. The results presented in the simulation indicate the superiority of the proposed algorithm in speedup the execution time of the program as well as reducing the task rejection ratio.