next up previous contents
Next: Management Review Up: No Title Previous: Contents

This document presents the first annual report of the European Fourth Framework Project PINEAPL (Parallel Industrial NumErical Applications and Portable Libraries). The PINEAPL project is a coordinated effort to produce a general purpose library of parallel numerical software suitable for a wide range of computationally intensive industrial applications. The initial coverage of the library is primarily determined by the requirements of the end-users' applications within the PINEAPL consortium.

The end-users in the consortium are taking advantage of the opportunity provided by the project to introduce new mathematical models and techniques into existing applications. Reduced wall-clock execution times allow users to experiment with more and larger configurations in their simulations, thereby achieving better designs and gaining new physical insights into their problems. Larger problems may be solved or greater accuracy in the solution may be achieved with these new techniques. In order to demonstrate the power and the efficiency of the resulting Parallel Library, several application codes from the industrial partners in the PINEAPL project (see Table 1) are being ported onto parallel and distributed computer systems by replacing sequential code sections with calls to appropriate parallel library routines.

A tool called PRECISE is also being further developed within PINEAPL which allows the stability of the numerical algorithms and the sensitivity of the application problems to be investigated.

The PINEAPL Project has a duration of 3 years and involves 10 organizations (see Table 1). The work is organized into 9 Workpackages, including Project Management and Exploitation, as described in the following sections.


Project Coordinator: NAG (UK)
Industrial Partners Related Parallelization and
Software Experts
British Aerospace (UK) University of Manchester (UK)
Piaggio(Italy) IBM SEMEA (Italy), CPS (Italy)
Thomson LCR (France) CERFACS (France), CPS (Italy)
Danish Hydraulic Institute (Denmark) Math-Tech ApS (Denmark)
Table 1: PINEAPL Consortium

One of the main goals of the project is to increase the suitability of the NAG Parallel Library for dealing with a wider range of computationally intensive industrial applications by appropriately extending the range of library routines emerging from the project.

The major result from the project will be portable library of high performance mathematical software, suitable for internal end-users' present (and possibly future) application codes as well as for those industrial end-users outside the consortium, that meets the following requirements:

  1. the problems addressed by the library should be relevant to a wide range of industries,
  2. the library should be portable and efficient across a wide range of parallel machines,
  3. the numerical algorithms must be accurate, stable and robust,
  4. existing software should be used as far as possible,
  5. the software and documentation should be produced to commercial quality and should meet recognized standards and pass strict quality control processing.
These objectives are met in the following ways:

The end-users' application codes have been chosen to represent a varied, if not comprehensive, cross-section of industrial problems ranging through electromagnetics, fluid dynamics, chemical reaction, oil reservoir simulation, beam propagation and thermal applications. The numerical software required to solve these problems includes, dense, banded and sparse linear algebra, partial differential equations, discrete Fourier transforms, load-balancing and optimization. This choice of application codes ensures a substantial degree of broadness of applicability for the numerical library.

Each of the end-user applications has been identified as being important to their businesses. In each case the end-user has stated performance expectations from work carried out in the PINEAPL Project. The intention is to measure the performance improvements achieved by the use of numerical software developed in the PINEAPL Project against the end-users' expectations, to gauge the success of the library software.

The numerical software is designed through close collaboration between the end-users and numerical experts in the consortium; the applicability of the numerical software is demonstrated by incorporating library routines into the end-users' seven industrial applications. Contact with external collaborators and other related EC funded projects (EUROPORT) and user groups such as the NAG User's Association gives further input to the software design and specification to ensure that the needs of a larger industrial audience are met. The software specifications are published (electronically on a project bulletin-board) for public comment and every opportunity is taken to publicize and promote the availability of PINEAPL material.

Most of the numerical software included in the library is tested for accuracy and stability using an existing tool, PRECISE. PRECISE is based on well-understood theory but is only in prototype form; development work in the project brings PRECISE to a mature, commercial quality, also providing another exploitation route for project results.

All of the library software meets recognized national, international or de facto standards to aid portability. Tools used within NAG are used to verify the adherence to language standards of the library software. In addition, the library software is also be subject to the same rigorous testing and quality assurance procedures as all NAG's products.

Technical work in the first year has concentrated on

  • determining a definitive contents list for the numerical library;
  • producing a set of specifications for the library software identified;
  • implementing and documenting the software specified;
  • transforming PRECISE into a mature, commercial quality product;
  • testing the accuracy and stability properties of application codes;
  • performing initial implementations of the application codes;
  • preparing benchmark test problems for the application codes; and
  • performing initial parallel implementations of the application codes.

next up previous contents
Next: Management Review Up: No Title Previous: Contents
Mishi Derakhshan
Wed Apr 23 12:54:57 BST 1997