IEEE Transactions on Signal Processing, vol. 48, no. 3, pp. 921-926, Mar. 2000

Real-Time Sonar Beamforming on Workstations Using Process Networks and POSIX Threads

Gregory E. Allen (1) and Brian L. Evans (2)

(1) Applied Research Laboratories, The University of Texas at Austin, P.O. Box 8029, Austin, TX 78713-8029 USA
gallen@arlut.utexas.edu

(2) Department of Electrical and Computer Engineering, Engineering Science Building, The University of Texas at Austin, Austin, TX 78712-1084 USA
bevans@ece.utexas.edu

Software Release

Process Networks Research at UT Austin

Abstract

In this paper, we present a formal framework for developing scalable software implementations of high-performance signal processing systems, such as sonar beamformers. The framework models the concurrency and parallelism in these systems using the bounded memory Process Network model. We implement the framework in C++ using POSIX threads to obtain low overhead, high performance, and scalability. We evaluate the framework by prototyping a 4-GFLOP three-dimensional digital interpolation beamformer on a Sun Ultra Enterprise workstation. For this application, we achieve near-linear speedup over 1 to 12 processors, and exceed real-time throughput constraints using 12 UltraSPARC-II processors running at 336-MHz. The C++ implementation of this framework will be available http://bevo.duckdns.org/PNSourceCode/.

Greg Allen's Beamforming Web Page.

Last Updated 05/21/14.