Ravindran, Somasundaram (1993) Aspects of practical implementations of PRAM algorithms. PhD thesis, University of Warwick.

Preview

PDF WRAP_Theses_Ravindran_1993.pdf - Unspecified Version - Requires a PDF viewer. Download (155Kb) | Preview

Request Changes to record.

Abstract

The PRAM is a shared memory model of parallel computation which abstracts away from inessential engineering details. It provides a very simple architecture independent model and provides a good programming environment. Theoreticians of the computer science community have proved that it is possible to emulate the theoretical PRAM model using current technology. Solutions have been found for effectively interconnecting processing elements, for routing data on these networks and for distributing the data among memory modules without hotspots. This thesis reviews this emulation and the possibilities it provides for large scale general purpose parallel computation. The emulation employs a bridging model which acts as an interface between the actual hardware and the PRAM model. We review the evidence that such a scheme can achieve scalable parallel performance and portable parallel software and that PRAM algorithms can be optimally implemented on such practical models. In the course of this review we presented the following new results:
1. Concerning parallel approximation algorithms, we describe an NC algorithm for findings an approximation to a minimum weight perfect matching in a complete weighted graph. The algorithm is conceptually very simple and it is also the first NC-approximation algorithm for the task with a sub-linear performance ratio.

2. Concerning graph embedding, we describe dense edge-disjoint embeddings of the complete binary tree with n leaves in the following n-node communication networks: the hypercube, the dc Bruijn and shuffle-exchange networks and the 2-dimcnsional mesh. In the embeddings the maximum distance from a leaf to the root of the tree is asymptotically optimally short. The embeddings facilitate efficient implementation of many PRAM algorithms on networks employing these graphs as interconnection networks.

3. Concerning bulk synchronous algorithmic, we describe scalable transportable algorithms for the following three commonly required types of computation; balanced tree computations. Fast Fourier Transforms and matrix multiplications.

Item Type:	Thesis or Dissertation (PhD)
Subjects:	Q Science > QA Mathematics > QA76 Electronic computers. Computer science. Computer software
Library of Congress Subject Headings (LCSH):	Parallel programming (Computer science), Computer algorithms, Computer architecture
Official Date:	December 1993
Dates:	Date Event December 1993 Submitted
Institution:	University of Warwick
Theses Department:	Department of Computer Science
Thesis Type:	PhD
Publication Status:	Unpublished
Supervisor(s)/Advisor:	Gibbons, Alan (Alan M.)
Extent:	ix, 154 leaves
Language:	eng

Request changes or add full text files to a record

Repository staff actions (login required)

View Item

Downloads