10409

9604

كارشناسي ارشد

معماري

اصفهان: دانشگاه صنعتي اصفهان، دانشكده برق و كامپيوتر

1394

سيزده، 110ص.: مصور جدول، نمودار

علي فانيان

1394/06/08

شادرخ سماوي

1396/10/04

كتابنامه

برق و كامپيوتر

مهندسي برق و كامپيوتر

ID9604

به فارسي و انگليسي

Task Mapping of a Hard Real Time Application on Heterogeneous NOC using Genetic Algorithm to Decrease Power Consumption Nafiseh Hejabdoost n hejabdoost@ec iut ac ir Date of Submission 2015 05 17 Department of Electrical and Computer Engineering Isfahan University of Technology Isfahan 84156 83111 Iran Degree M Sc Language FarsiSupervisor Ali Fanian a fanian@cc iut ac irAbstract Today with advances in semiconductor technology the number of processing elements in a system on chip SOC is increased Communication architecture of such systems is based on the bus Hence byincreasing the number of processing components and due to the lack of bus performance and expandability the network on chip or NOC concept as an efficient and scalable inter chip communication plan toovercome the buses problems has been proposed One of the major challenges in the NOC research is theproblem of mapping tasks of an application on the homogeneous or even heterogeneous processing coresconnected to the network routers On the other hand one of the most versatile applications is embeddedapplications with real time requirements In many previous works the problem of mapping has beeninvestigated for homogeneous processing cores In other words although heterogeneous cores are closer tothe real application most of the proposed schemes have ignored this property In addition the thecharacteristic of real time application wasn t the main focus of the previous research works One of theother challenges in the network on chip is the power consumption in the NOC In this thesis at first asurvey of the work done in the last one decade in the domain of application mapping is discussed then anew application mapping for hard real time application for heterogeneous core based on multi objectivegenetic algorithm is proposed Since optimal solution is a NP hard problem we use genetic algorithm toachieve semi optimal solutions In addition the proposed method prevents infeasible solutions beingproduced in new generations This strategy cause that the proposed scheme converge to the pareto optimalsolution faster than other schemes Experimental results are presented and evaluated using several well known metrics as well as a new metric This shows the effectiveness of the proposed method compared toother approaches Keywords Network on chip hard real time application mapping multi objective genetic algorithm

علي فانيان

شادرخ سماوي