The primary objective of both the EUROPORT-1 (8421) and EUROPORT-2 (8586) projects is to build confidence and generate awareness in the effectiveness of parallel high-performance computing (HPC) platforms by demonstrating their industrial potential. In this context, EUROPORT-2 focuses on the areas of:
oil reservoir engineering
To achieve the project aims, the following specific objectives must be met:
performance improvements should be demonstrated by porting codes to parallel architectures
the codes are to be portable between different parallel architectures
the results are to be scalable to larger machines
the codes are to be supportable in a practical working environment
the results of the work are to be widely publicised and effectively exploited.
Clustering several porting activities into a single project has a number of advantages, but particularly helps with:
monitoring of individual ports
planning and conducting review
identifying synergies between ports
broking the availability of services
defining project wide standards for all Porting Consortia
defining and auditing benchmarks
managing financial and contractual aspects of the project. liaison with the CEC
promoting and disseminating the outcome of EUROPORT-2 preparing an overall project report.
The service providers will provide the skills and other resources which are required by the individual Porting Consortia but which are not available within the consortium. These include:
benchmarking services: these will include hardware platforms together with skilled staff to support them
expertise-based services: these might include parallelisation expertise, platform-specific advice, arbitration advice etc.
product-based services: these might include ensuring access to tools, the requirement for which had not been anticipated
dissemination services: these might include conference organisers and press agencies.
These facilities are likely to be offered by a range of organisations including specialist HPC centres within universities and government laboratories and hardware vendors.
The EUROPORT-2 Manager reports directly to the CEC and will carry out many of the duties on the project’s behalf.
Reporting to the Manager there are 10 individual Porting Consortia, which will carry out the 15 code porting activities themselves. Each of the 10 Porting Consortia is summarised below:
A portable, scalable parallel version of a widely used application, Eclipse 100, will be developed. This is used in the petroleum industry for black oil reservoir simulation. The use of parallel processing will enable larger simulations to be conducted (improving accuracy) together with reductions in execution time (reducing costs) compared with the vector parallel systems currently used.
Three computational chemistry codes, GAMES-UK, ADF and VAMP, will be ported. Parallel processing will permit many molecular modelling computations in the chemistry and the pharmaceutical industry to become interactive rather than batch thereby reducing development costs.
Three computational chemistry codes, GROMOS, MNDO and TURBOMOLE will be ported. The use of parallel processing will permit many molecular modelling computations in the chemistry and the pharmaceutical industry to become interactive rather than batch thereby reducing development costs.
The aim is to convert the computational electromagnetic codes, TOSCA and ELEKTRA, so that they will run efficiently on a range of parallel processing systems from networked workstations to shared memory multiprocessors. the use of parallel processing will enable larger simulations to be conducted (improving accuracy) together with reductions in execution time (reducing costs) compared with the systems currently used.
The general aim is to build a high performance system for determining the best treatment plans in radiotherapy. This will involve the migration of the Monte Carlo method code EGS4 onto a parallel platform and the realisation of software packages to build the target of the simulation process, namely a 3D representation of human bodies deriving data from computer topography, and to permit the visualisation of the resulting effects of the radiation on the patient. The use of parallel processing will enable visualisation to be carried out cost-effectively. Only vector parallel systems are currently capable of such calculations within an acceptable time limit and these systems are too expensive to be widely available to hospitals.
An electromagnetic problem solving environment which consists of the Field Analysis Modeller FAM and a three dimensional finite difference time domain solver of Maxwell equations, THREDE, will be parallelised. The use of parallel processing will address the main areas where the lack of computational power seriously limits the size of problems that can currently be modelled on vector parallel systems.
This aims to produce a scalable parallel and highly portable version of the satellite radar processing and image understanding programmes developed by two of the partners. The proposed work falls into two parts, the analysis of radar data and the visualisation of the results. The computational requirements of the algorithms are heavy and current processors, typically workstation-based, provide inadequate response to deal with the high volumes of data involved. The advent of fully commercial highly parallel computing systems has, however, made it possible to provide adequate power for real-time SAR processing at economic cost.
It is proposed to migrate the ADABAS database to the European Declarative System (EDS) parallel platform designed specifically for the large scale commercial corporate market. The technical issues which would need to be resolved concern scalable performance, reliability and the provision of an integrated single image of the system. The use of parallel processing will enable transaction per second (tps) rates to be achieved that are significantly higher than may be achieved on mainframe systems. In addition, the relative cost of each tps will be much lower than on mainframe systems.
The aim is to implement, optimise and benchmark a protein database search software running on the most advanced parallel computers. This will be achieved by porting the multiple sequence alignment programme MaxHom. The goal of this project is a significant performance improvement such that scanning a complete database using the most sensitive search technique can be done interactively.
It is proposed to port the cartoon animation ANIMO onto a parallel environment thereby demonstrating the advantages of parallel implementations to the entire visualisation and graphics industries. This will be achieved using end-user supplied benchmarks, like interactive painting operations and off-line rendering the animation to film, HDTV or videotape.
20120 San Donato Milanese
PR4 1AX Preston
WA4 4AD Warrington
CB2 1SJ Cambridge
NG1 4BT Nottingham
CB4 5AF Cambridge
53757 Sankt Augustin
SW1P 3JF London
OX14 1DZ Abingdon
SW15 1SW London
CF1 6RW Cardiff
L3 5TF Liverpool
5656 AA Eindhoven
OX4 4GA Oxford
BA4 5QE Shepton Mallet
9700 AB Groningen
SO17 1BJ Southampton
L62 3JD Wirral
BD7 1DP Bradford
L69 3BX Liverpool
OX1 3QD Oxford
1081 HV Amsterdam
OX5 1JE Kidlington
W1Y 6LN London