Parallelization of Stochastic Evolution for Cell Placement

(2004) Parallelization of Stochastic Evolution for Cell Placement. KFUPM. (Unpublished)

[img]
Preview
PDF
proposal.pdf

Download (156kB) | Preview

Abstract

VLSI physical design and the problems related to it such as placement, channel routing, etc, carry inherent complexities that are best dealt with iterative heuristics. However the major drawback of these iterative heuristics has been the large runtime involved in reaching acceptable solutions especially when optimizing for multiple objectives. Among the acceleration techniques proposed, parallelization is one promising method. Distributed memory multiprocessor systems and shared memory multiprocessor systems have gained considerable attention in recent years of research. This idea of parallel computing has attracted both the researchers and manufacturers who are targeting to reduce the time to market. Our objective is to exploit the benefits of parallel computing for a time consuming placement problem in VLSI. Finding the best solution for the placement of n modules is a hard problem. Thus the enumerative search techniques, specially those which employ the brute force, are unaccepted for the circuits in which n (number of modules) is large. Constructive and Iterative heuristics play the key role in this scenario and hence are frequently used. We will use Stochastic Evolution for finding the optimal solution to the above mentioned placement problem where the major task in our objective will be the parallelization of Stochastic Evolution using different parallelization techniques and the comparison between these different parallelized versions based on the results achieved. The parallelization will be carried out using MPI (Message Passing Interface) on a distributed memory multiprocessor system and conclusion will be based on the results achieved that are expected to show speedup nearly equal to linear speedup when run over increasing number of processors.

Item Type: Other
Subjects: Computer
Department: College of Computing and Mathematics > Computer Engineering
Depositing User: KHAWAR SAEED KHAN
Date Deposited: 10 Mar 2008 11:33
Last Modified: 01 Nov 2019 13:23
URI: http://eprints.kfupm.edu.sa/id/eprint/278