Please use this identifier to cite or link to this item:
|Title:||MULTIPROCESSING OF APG's ON HYPERCUBE AND De BRUIJN NETWORKS|
|Authors:||Gupta, Vikas Kumar|
|Keywords:||ELECTRONICS AND COMPUTER ENGINEERING;MULTIPROCESSING;HYPERCUBE;DE BRUIJN NETWORKS|
|Abstract:||Exploiting parallelism is an increasingly common approach to improve the performance of computer systems. In terms of hardware this typically means providing multiple, simultaneously active processors. In terms of software, this typically means structuring a program as a set of largely independent sub tasks. The structure of programs is usually represented by acyclic precedence graphs (APG's). The nodes of the graph denote the subtasks of the program and the directed arcs denote the precedence relations among the subtasks. The work, reported in this dissertation, evaluates the performance of two popular multiprocessor networks for graph structured programs, through simulation. Performance measures like speedup and efficiency have been studied by varying problem size, number of processors, granularity of task graphs, and degree of task nodes. The simulation system, developed •in this work, generates APG's and simulates the Hypercube and De Bruijn networks. A fast, static scheduling algorithm has been developed to schedule the generated APG on to the target processor networks. Performance measures are then computed for the obtained schedule. A comparative study of the performance of multiprocessor networks is then carried out.|
|Research Supervisor/ Guide:||Kumar, Padam|
|Appears in Collections:||MASTERS' DISSERTATIONS (E & C)|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.