National and European Projects

Design of an efficient infrastructure for computations over a grid :
We use the tools developed within the framework of the GRID-ASP project by LIP-ReMaP -- partner of the project --, LORIA-Résédas, and LIFC-SRDP with an additional high-level administrator interface for the definition, the deployment, and the exploitation of services over a grid (Weaver module).

Design of a Web interface for computation over a grid :
This interface -- called Websolve -- allows a standard navigator to submit computational requests to a grid by using third-party software such as the MATLAB computational environment, OCTAVE, Scilab, Scilab//, Netsolve, DIET, ...

The main developments for the expert site consist of :
Writing the procedures for the expertise, inclusion of the sparse matrix software, building a database including a bibliography on sparse matrix software and collections of sparse matrices (Rutherford-Boeing and PARASOL data sets).

Local contact : Christophe Hamerling (hamerling@cerfacs.fr)

Objective
The objective of EUROPORT is to demonstrate the benefits of parallel computing technology for industry. This is done by porting 38 large serial commercial and industrial codes to parallel computers. One of these codes is N3S, a CFD package developed by EDF (Electricité De France) and INRIA-Simulog (France) for the simulation of complex industrial flows. It is widely used in different industries, including the automotive, aerospace, defence, energy, oil, turbomachinery, environment, and medical.

Consortium
Within the HPCN3S consortium, coordinated by Simulog, the complete software (compressible and incompressible implicit time schemes for unstructured meshes made of tetrahedra) is parallelised with the help and expertise of CERFACS and EDF. The end-users of this project are IFP (Institut Français du Pétrole), CISE (Centro Informazioni Studi Esperienze, Italy), and VKI (Von Karman Institute, Belgium).

Local contact: Luc Giraud (giraud@cerfacs.fr)

Consortium
NAG is the leader of the consortium which includes British Aerospace, CERFACS, LCR Thomson-CSF, CPS (Napoli, Italy), the Danish Hydraulic Institute (Denmark), IBM SEMEA, the University of Manchester (UK), Math-Tech, and Piaggio (Italy).

Local contact: Valérie Frayssé (fraysse@cerfacs.fr)

Goals
In conjunction with the Human Capital and Mobility program, we are carrying out the study, development and analysis of a variety of iterative methods and preconditioners applied to challenging problems of significant importance in both industrial and academic worlds. As part of the HCM project, we have established a network of recognized leaders in the area of scientific computing. The network consists of experts in direct solvers, iterative solvers, preconditioning methods, finite element techniques, and computational fluid dynamics. The primary goal of the project is the development of new iterative methods and preconditioners for highly nonsymmetric and/or highly indefinite problems.

Results
Important results have been achieved concerning the error analysis of Krylov subspace methods (including a detailed study of stopping criteria), the development of new efficient hybrid iterative solvers for indefinite and nonsymmetric problems, the analysis and implementation of parallel block iterative schemes for heterogeneous computing environments, and the application of such innovative techniques to several difficult problems, especially in CFD. These activities have been documented in numerous published papers and research reports. The project started in January 1994 and is scheduled to end in September 1996.

Consortium
Partners in this project are CEA (France), CERFACS, Dassault Aviation (France), IAN-CNR Pavia (Italy), Utrecht University (the Netherlands) and CRS4 (Italy).

Local contact: Michele Benzi (benzi@cerfacs.fr)

An industrial requirement in the areas of computational Fluid Dynamics and Structural Analysis applications is to be able to solve large sparse linear equation systems (>1M equations). Many existing codes use either direct solvers or iterative solvers with preconditioning - both have to solve large sparse matrices. Simple iterative methods such as Conjugate Gradient or SOR do not work for stability reasons.

Solving linear equation systems of such sizes in a reasonable time requires the use of powerful MPP systems. Therefore, message-passing versions of direct sparse solvers and various iterative methods are needed. Up to now, however, no generally available library of scalable parallel sparse matrix solvers or preconditioners exists.

Objectives
The PARASOL project has the following main objectives:

• To develop sparse matrix algorithms according to the needs of industry.
• To implement portable prototypes of these solvers and to demonstrate them working on different HPC systems.
• To specify an open library interface for the PARASOL library.
• To integrate the solvers and additional utilities into the PARASOL library and provide an interface to the state-of-the-art programming environment TRAPPER.
• To establish continuous library support.
• To evaluate the PARASOL library and utilities by industrial application developers.

• leading European research organizations with a well-known experience and track-record in the development of parallel solvers (CERFACS, ENSEEIHT, GMD-SCAI, ONERA, RAL, Univ. of Bergen).
• industrial code developers who define the requirements for PARASOL and will use its results (INPRO, MacNeal-Schwendler, Det Norske Veritas (DNV), Polyflow, Apex Technologies).
• the two leading European HPC software companies who will exploit the project results and will provide programming development tools (GENIAS, PALLAS).

• specification of an open library interface for parallel solvers (which will be disseminated to other solver developers outside the project)
• solver prototypes based on MPI, such as the parallel distributed sparse direct solver MUMPS (MUltifrontal Massively Parallel sparse direct Solver.)
• a prototype of the PARASOL library
• assessment and evaluation reports by industrial software developers
• technology transfer and commercial exploitation of the project results.

Local contacts: Iain S. Duff (duff@cerfacs.fr) and Jean-Yves L'Excellent (excelle@cerfacs.fr)