Computational Process Networks

This is the download page for CPN releases circa 2010-2011, until it got moved to bitbucket.
After bitbucket did its reorganization, the code was offline for awhile. Now it's hosted on github.

Computational Process Networks is a model and a framework library for real-time high-throughput signal and image processing. The model combines:

Process Networks [1], a formal, concurrent model for functional parallelism

Bounded Scheduling [2] for realization in finite memory

Firing Thresholds from Computation Graphs [3]

Computational Process Networks model algorithms on overlapping continuous streams of data (such as digital filters and overlap-and-save fast Fourier transforms), and provides liveness and determinate execution. Our C++ implementation provides a portable, high-performance, scalable framework using POSIX Threads.

The Computational Process Network source code is publicly available, under the GNU GPL.

Release 1.0.0 (October 2011) and all future releases have moved to bitbucket.org/gallen/cpn

Bitbucket is a widely used code hosting site. From there we can easily make small bugfix releases, have tagged release versions (including tarballs), and easily accept contributions. This is a minor release from before, mostly an update to the build system. CPN now uses GNU autotools instead of including a set of makefiles for a limited number of platforms.

Release 20110601 - This is our fifth major release. Specific new features include:

user API changes and updates for clarity
a tutorial document with sample code
improved remote queue performance
updates and improvements to case studies (as used in dissertation results)
addition of several types of utility nodes in node library
bug fixes to D4R, the CPN Kernel, PthreadLib, and more
stability and robustness as demonstrated by multiple 72 hour tests

This release represents the version used for dissertation case studies [5]. It contains about 30K lines of code in about 330 source files (as estimated by cloc). This software for POSIX systems includes makefiles for Linux on i686 and x86_64, and MacOS X on i386.

Release 20101207 - This is our fourth major release. Specific new features include:

D4R - the distributed dynamic deadlock detection and resolution algorithm
beamforming case study (used in Asilomar 2010 paper)
updated prime sieve case study
construction of CPN programs from XML and JSON descriptions
significant changes to the remote queue implementation
reorganization of layout to have a node library
Pseudo nodes - I/O endpoints to a CPN network embedded in another application
Functional nodes - templated nodes from functions or functional objects
improvements to the pthread library
general cleanup
bug fixes

Release 20100212 - This is our third major release, and our first release in nearly 4 years. It is our first release after a major rewrite of portions of the CPN framework. Specific new features include:

a distributed implementation targeting clusters of workstations
a CPN kernel for dynamic construction of CPN programs
extensive built-in unit testing
case studies to exercise this new framework

Release 20061110 - This release branch never left pre-alpha state. It was our first release to include the D4R algorithm.

Release 20000329 - The original release of CPN worked only on symmetric multiprocessing workstations. It was used in a March 2000 paper [4] which prototypes a 3-D sonar beamfomer. The beamfomer sustains a real-time 4 GFLOPS on a Sun Ultra Enterprise 4000 with 12 336 MHz UltraSPARC-II processors.

G. Kahn, "The Semantics of a Simple Language for Parallel Programming," Information Processing, pp. 471-475, Stockholm, Aug. 1974.
T. M. Parks, "Bounded Scheduling of Process Networks," Technical Report UCB/ERL-95-105, Ph.D. Dissertation, EECS Department, University of California, Berkeley, CA 94720-1770, Dec. 1995.
R. M. Karp and R. E. Miller, "Properties of a Model for Parallel Computations: Determinacy, Termination, Queueing," SIAM Journal, vol. 14, pp. 1390-1411, Nov. 1966.
G. E. Allen and B. L. Evans, "Real-Time Sonar Beamforming on a Unix Workstation Using Process Networks and POSIX Threads", IEEE Transactions on Signal Processing, pp. 921-926, March 2000.
G. E. Allen, "Computational Process Networks: A Model and Framework for High-Throughput Signal Processing," doctoral dissertation, Dept. of Electrical and Computer Engineering, The University of Texas at Austin, Austin, Texas, 2011.

Software Developers: Gregory E. Allen and John F. Bridgman
Research Guidance: Brian L. Evans

This work is supported by the Independent Research and Development program at
Applied Research Laboratories: The University of Texas at Austin.

For more information contact: Greg Allen <gallen AT utexas DOT edu>