This is the old home page of the PARINT
project in the Department of Computer
Science at Western Michigan University
and Washington State University. The
PARINT Project was developed with support from the
National Science Foundation under the award ACR-0000442.
We are working on a new web page and on a new release, ParInt2.0.
ParInt1.2 links are provided under Downloading below. The adaptive
code works as expected, but the MC and QMC codes do not because their dependence on the
random number generator needs modifications.
Announcements
Note: The versions under Software Downloading target 32-bit systems, and
need fixes for dynamic linking on 64-bit systems.
This will be addressed in ParInt 2.0.
PARINT1.2
- beta
The PARINT
Version 1.2 beta includes significant improvements: the use of Quasi-Monte
Carlo techniques over arbitrary hyperrectangular regions, parallel Monte
Carlo and integration over simplex regions.
PARINT1.1
Release
Features of this release include additional integration rules and improved
memory handling methods for reducing memory use when solving very large
problems.
Project Overview
The goals of the PARINT
project include: to investigate new techniques for computing multivariate integrals
in parallel, as well as numerical methods and high precision computations;
to study applications of the package, for example in high energy physics,
Bayesian statistics and computational finance; and to develop a user-friendly
software interface for the package. Research areas include load balancing,
distributed data structures, parallel quasi-Monte Carlo techniques and extrapolation.
Downloading
PARINT1.1
incorporates a set of basic techniques for solving multivariate
integration problems numerically. One of the basic algorithms is parallel global
adaptive integration, with each process storing its own priority queue
of subproblems, and incorporating a scheduler-based local load balancing
technique. Features include:
-
Two techniques for running the software:
-
Running the PARINT executable
from the unix command line
-
Using the PARINT API; allowing
the user to solve integration problems from their own C applications by
using the PARINT functions
-
Support for double and long double accuracy
-
Use of MPI for underlying process communication, providing platform independence
-
Support for user integrand functions written in C, C++ or Fortran
-
Support for vector integrand functions, integrated over hyperrectangular
regions.
-
A full User Manual.
Users can download the parint1.2-beta release
and the documentation.
You can view the website of our HPCS (High Performance Computational Science) Laboratory at http://www.cs.wmich.edu/~hpcs
People
Elise de Doncker and
John Kapenga lead the project at Western Michigan University,
Alan
Genz at Washington State University.
This project was funded in part by the National Science Foundation and Western
Michigan University. If you are interested in contributing
or have any questions, please contact one of the authors.
Click on icon to email us with questions or comments.