Paragon is a collection of data-parallel program constructs
implemented in C++.  Paragon is intended to be a research vehicle for
investigating issues for parallel execution of large-scale scientific
programs.  We use it to develop and evaluate parallel programming
constructs and techniques.  We also use it in the Parallel Processing
course at Cornell to demonstrate the data-parallel programming
paradigm. 

What does it run on?
There are two versions of the Paragon library.  The uniprocessor
version should run on any machine with a working C++ compiler.  It is
known to work on Sun workstations with either AT&T's Cfront 2.0
compiler or GNU's g++ 1.39.0  The uniprocessor version has also been
tested on, the IBM RS 6000 and 3090 computers.

The multiprocessor version should run on the Intel iPSC/2 and iPSC/860
hypercube.  The multiprocessor version has also been ported to the
Parallel Virtual Machine (PVM).  This release of Paragon supports PVM
to provide parallel execution on heterogeneous networks of
workstations. 

How do I get it?
The paragon release is provided in three tar files.  The complete
distribution is available on tesla.ee.cornell.edu in:
pub/paragon/paragon-1.1.tar.Z

If you're just interested in the serial version, get
pub/paragon/uniproc.tar.Z

If you're just interested in some reading material, the LaTeX source
and raw postscript for the Paragon User's Guide is available in

pub/paragon/uguide.tar.Z.  

The Paragon paper presented at the 1991 International Conference on
Parallel Processing is available as 
pub/paragon/ICPP-91.ps.Z 

and an updated version of this paper is included as
pub/paragon/ICPP-91_revised.ps.Z

Please send all comments, suggestions, bug reports, and complaints to
paragon@ee.cornell.edu