EVALUATION OF MASSIVELY PARALLEL PROCESSING ALGORITHMS FOR MATCHED FIELD PROCESSING (MFP) AND CLUSTERED MFP

Authors
  1. Meng, Z.
  2. Racca, R.
Corporate Authors
Defence Research Establishment Pacific, Victoria BC (CAN);JASCO Research Ltd, Sidney BC (CAN)
Abstract
An evaluation of Massively Parallel programming algorithms for Matched Field Processing (MPP) and clustered nearest neighbours MFP was carried out on a DECmpp 12000 parallel computer (from Digital Equipment and MasPar Corporation) with 8192 processors. The performance of MFP on the Maspar machine was compared with that of MFP on a serial VAX9000 equipped with a vector processor. The result showed that the MasPar implementation achieved a factor of 20 speedup relative to the VAX9000. This factor was much higher than the suggested ratio of 7.5 (based on 600 MFLOPS for the MasPar versus 80 MFLOPS for the VAX9000), indicating that the parallel machine is a more efficient platform for MFP than the vectorized serial machine. The successful implementation of parallel clustered MFP algorithms proved the feasibility of pursing this approach on a massively parallel machine and showed definite promise of even greater efficiency.
Report Number
DREP-CRS-93-80 — Contractors Report Series
Date of publication
01 Jan 1994
Number of Pages
23
DSTKIM No
94-04174
CANDIS No
139465
Format(s):
Hardcopy;Originator's fiche received by DSIS

Permanent link

Document 1 of 1

Date modified: