Jump Over Left Menu
Engineering Computing Newsletter: Issue 39
- Visualisation Surgery
- Future Trends and Directions in CFD
- Upwinding in CFD
- CHEST/NISS Update
- New Cray for RAL
- Workshop on Boundary Conditions in CFD
- SDALS: STEP Data Access Interface Specification
- Virtual Shared Memory Seminar
- PACTA '92
- Benchmarking and Evaluation of Parallel Systems
- Advanced Course on Large Scale Scientific Parallel Computing
- The LaTeX Cookbook
- Future Events
At its last meeting on 8 April 1992 the Visualisation Community Club's Steering Group decided to set up a pilot Visualisation Surgery service for the community. The purpose of the service is to offer advice and help in the use of visualisation systems in the context of the user's own application area. The service will provide an opportunity for users to interact directly with visualisation and graphics experts and to see visualisation systems. This service will be run on a regular monthly basis at two centres: Rutherford Appleton Laboratory (RAL) and Manchester University. Initially, the service will be an experimental activity whose results will be reviewed and reported on after a period of three months. Full details of the Visualisation Surgery service available at the two centres are given below.
RAL Centre Surgery
The RAL Visualisation Surgery service is set up to support SERC funded researchers who are using or would like to use visualisation software and techniques in their research. The service is aimed at people ranging from total novice to experienced user.
The service will be a regular point of contact where people can:
- be introduced to visualisation
- be shown various visualisation hardware/software systems
- discuss problems experienced using visualisation systems
- be advised on which existing visualisation systems may be suitable for their application
- be shown how a particular visualisation system can be used for their application
- be advised on other computer graphics questions related to visualisation.
The Surgeries are open to SERC funded researchers free of charge, but places may be limited.
- Stellar GS2000 (4 processor) 32MBs memory, 32 colour planes
- Titan ST750 (2 processor) 32MBS memory 32 colour planes
- Sun SparcStation 2, 32MBs memory, 8 colour planes, GX graphics accelerator
- SG Indigo, 16MBs memory, 8 colour planes
- IBM RS6000 model 530E, 32MBs memory, 8 colour planes, GTO graphics accelerator
- AVS4 (running on Stellar GS2000)
- AVS3 (running on Titan ST750, Sun SparcStation 2 and IBM RS6000)
- Explorer 1.0 (running on SG Indigo)
- Khoros 1
- Visualisation of Scientific and Engineering Data:
- available methods
- visualisation software (as above)
- Graphics Workstations including 3D Workstations (as specified above)
- Novel Computer Input Devices such as Spaceball, DataGlove
- Image Processing
- Window Management: X11
- Window System and X11 Toolkits
- Graphical User Interfaces: Motif, UIMS
- Graphics Programming Interfaces: GKS, PHIGS, PHIGS PLUS
- Exchange of Graphics and Image Data: CGM, PostScript, image formats
- Parallelisation Techniques
- Visualisation of Scientific and Engineering Data:
Manchester Computing Centre Surgery
There will be a number of visualisation open days held towards the end of every month in the Computer Graphics Unit, Manchester Computing Centre. The purpose of these open days is to discuss problems people may be experiencing using visualisation systems and offer possible solutions. These days are open to any UK Academic free of charge but places may be limited.
The Computer Graphics Unit will be using a Hewlett Packard 730 Workstation and A VS4: the day will be geared to using AVS in a Unix environment.
What the day will involve?
The day will include an introduction and overview of AVS with specific examples and a demonstration of AVS will be provided. Throughout the day there will be sufficient time set aside for attendees to ask questions. The first few open days will be operated on an experimental basis but due to time and resources constraints, it will not be the aim of the day to provide complete solutions for each attendee's problem. It will however provide some guidance and assess the feasibility of using AVS to solve their problems.
The service will run from 10:00 to 16:00 on the following days:
- 15 July
- 19 August
- 16 September
- 21 October
- 18 November
- Manchester University
- 16 July
- 16 September
- 21 October
- 18 November
Rajka Popovic, Informatics Dept
A number of changes have recently taken place in the organisation of the Education and Awareness Group. Debbie Thomas, who was the Group Leader, has recently taken 4 months maternity leave to await the arrival of her second child. When she returns to work in November, it will be in a different role in the Computational Fluid Dynamics Group in Informatics Department. We would like to thank her for the hard work she has done on behalf of the EASE E&A Programme and wish her all success in her new post.
The new E&A Group leader is Susan Hilton. Susan has been in the E&A Group for the last three years, and has been involved in such projects as the survey carried out in HEIs throughout the country and the E&A courses and conferences programme.
As you know, Terry Mawby recently took over the editorship of this Newsletter and he now also has the additional responsibility of the EASE E&A events programme.
We are delighted that two new members have joined the Group. Rachel Miles, who does a great job in formatting the ECN and Lynne Oldfield who runs the successful E&A events.
I have every confidence that this new team will continue the excellent work of the E&A Group under its new leader.
Visualisation in CFD
As part of its on-going programme of technical meetings, the SERC CFD Community Club held a one day meeting on Visualisation in Computational Fluid Dynamics on 9 March 1992.
The meeting was attended by 90 people who made the following recommendations:
- a wider survey of visualisation systems should be undertaken, including those available in the public domain and on PCs
- links with the USA and Europe should be clarified and possible cooperation with NASA on the virtual wind tunnel investigated
- closer links should be established between the Visualisation and the CFD Community Clubs in order to improve the information flow
A copy of the presentations at the seminar can be obtained by contacting the authors.
C Greenough and C Fitzsimons, Informatics Dept
Future Trends and Directions in Computational Fluid Dynamics
This seminar at RAL on 6 April marked the end of the second successful year of operation of the SERC CFD Community Club. It was chaired by Prof Geoff Hammond (Bath), the chairman of the Community Club Steering Group, who announced that the club was making commercial CFD software available on the Cray at RAL for use by the academic community.
A European dimension was added to the meeting by Prof Maurice Meneguzzi (CERFACS France) who gave the keynote address and spoke about current work in direct numerical and large eddy simulation in France. He also discussed techniques in visualisation which aid the interpretation of the results and the storage of the collected data.
Prof Michael Leshziner (UMIST) surveyed turbulence models for complex flows in combustion and external aerodynamics, while Dr Bassam Younis (City) addressed environmental flows.
This was followed by a survey of the wide range of numerical techniques for external aerodynamics by Prof Ken Morgan (Swansea). Prof Derek Bradley (Leeds) discussed the present status of combustion modelling, stressing the variety of disciplines involved in the work and the requirement for developing the role of CFD in total design.
From the discussion period the view emerged that a lot more work was required on direct numerical and large eddy simulations but that, at present, there were insufficient UK computing resources for the level of work required.
The meeting recommended to the Steering Group that:
- the feasibility of acquiring some or all, of the NASA CFD test data should be investigated
- access to test data generated under SERC grants should be improved
A copy of the presentations at the seminar can be obtained by contacting the authors.
C Greenough and C Fitzsimons, Informatics Dept, RAL
Upwinding in CFD - Community Club Seminar
This Seminar was held at RAL on 26 May jointly chaired by Professor P L Roe (Michigan) and Dr P K Sweby (Reading). Dr Sweby began the morning session with an introduction to flux-limiter schemes for conservation flows. He reviewed the design criteria for high resolution Total Variation Diminishing (TVD) schemes, i.e. high order accuracy, entropy-satisfying as well as TVD, before going on to discuss Flux Corrected Transport (FCT) and flux limiter schemes. Their application both to scalar equations and to systems of conservation laws (with reference to Roe's Approximate Riemann solver) was described.
Dr R Hillier (ICSTM) presented the use of a high resolution Gudonov-type scheme for a range of practical problems in steady and unsteady aerodynamics. After a brief description of a generalised Riemann problem for the calculation of the fluxes was given, the presentation went on to concentrate on the applications of the method for: an unsteady shock wave diffracting at a sharp convex edge; an unsteady shock propagation through convergent-divergent nozzles; and an application to steady hypersonic viscous flows.
Dr E F Toro (Cranfield Institute of Technology) gave a presentation on how to use upwind schemes economically through a Riemann solver adaptive procedure approach. He proposed the adaptive use of a cheap linearised solver (based on primitive variables) for smooth parts of the flow and a robust solver (for example, exact Riemann solver, the Harten-Lax-van Leer approach, curve fitting) for use elsewhere, with a switching criterion. The procedure can be used directly in the first-order Gudonov method, and higher-order extensions implemented with reference to a weighted average flux method. Applications to shallow water flows and steady supersonic flows were given.
Professor P L Roe gave a presentation on the current status of his research on multi-dimensional upwind methods. He mentioned that in current research, the higher-order Gudonov methods were fairly satisfactory for shock-capturing, and sheer wave-capturing was fine for flows that are aligned with the mesh. But for separated flow, neither central nor current upwind methods seem to offer a sound basis for economical Navier-Stokes codes, with ID upwinding inserting misleading results. The requirements for a truly multidimensional method and application to viscous flows with maximal time-step were given.
Professor D M Causon (Manchester Polytechnic) discussed the use of TVD schemes applied to steady and unsteady flows, after reviewing some of the classical shock capturing schemes and their drawbacks. He illustrated a method of constructing a high-order accurate flux limited scheme with the development of a 6th order accurate monotonicity preserving TVD scheme.
Professor K Morgan (Swansea) presented the application of the methods discussed earlier to more general meshes. He discussed the problems of computational efficiency and the need for storage reduction for unstructured meshes. The methods were applied to a number of standard cases, including the NACA 0012 aerofoil.
Dr Sweby then brought an informative and interesting meeting to a close. The next meeting of the CFD Community Club is a Workshop with the MOD AGCFM on Boundary Conditions on 22 September 1992 at the University of Surrey.
A copy of the presentations can be obtained by contacting Conor Fitzsimons.
Some of the latest centrally funded agreements from CHEST include: FIGARO+ from Liant Ltd - a PHIGS system for Unix and VMS systems which has been recommended by the Advisory Group On Computer Graphics. Further details about this offer will shortly be mailed to the community. X-Desktop and Motif from IXI Ltd a desktop and GUI for a wide range of systems (and selected primarily to allow a standard look and feel on a range of Unix platforms). These products have been recommended by the IUSC Workstations Working Party. CHEST Site Contacts have further details about this agreement.
An agreement has also been signed for Embase, a medical bibliographic dataset likely to be of wide interest. It has much of the same journal coverage as Medline, but has a stronger emphasis on pharmaceutical issues. Discussions are underway to determine how the service will be provided and details about the agreement will subsequently be mailed to the community.
An initial agreement has been reached for a central purchase of Monotype Fonts - a large range of fonts that can be used with Postscript - and details should be finalised in the near future.
New Cray for RAL
At its Council meeting on 10 June, SERC agreed the procurement of a Cray Y-MP8I/8128 supercomputer to be installed at the Atlas Centre of SERC's Rutherford Appleton Laboratory.
The new supercomputer will replace a five year old Cray X-MP/416 and will have about three times the power of its predecessor. The machine will come into service in the autumn and will be used for the most demanding computations in a broad range of science and engineering projects sponsored by the UK's Research Councils.
The new supercomputer is being provided as the first stage in a renewed programme of investment in national high performance computing facilities, overseen by the Advisory Board for the Research Councils and managed by SERC.
In the immediate future, the new Cray will enable important research work, whose progress has been slowed recently by the intense competition for the present supercomputing resources, to proceed more quickly.
These include major projects in oceanographic and atmospheric science which will help our understanding of climatic behaviour, studies of the structures of new materials and processes that are important scientifically, and the realistic simulation of complicated structural and fluid flow problems in engineering applications.
The new facility is likely to attract completely new types of project, and it may also enable scientists to tackle in this country some of the large projects which could only be tackled on facilities in other countries.
The Cray Y-MP8I/8128 is an 8-processor supercomputer with 128 million 64-bit words of memory and, in this instance, 100 Gbytes of disk storage. Its peak performance is 2.7 Gflops (2.7 thousand million floating point operations per second). At the Atlas Centre the supercomputer will be connected to the Joint Academic Network (JANET) through which access can be provided to higher education institutes. The supercomputer will run Cray Research's UNICOS operating system and it will have access, together with other computers at Atlas, to very large data storage facilities.
Dr Brian Davies, Rutherford Appleton Laboratory
CFD Community Club Workshop on Boundary Conditions in CFD - 22 September 1992
The CFD Community Club and MOD AGCFM will hold a one day workshop on Boundary Conditions in CFD on 22 September at the University of Surrey, under the chairmanship of Dr Brian Williams (DRA, Aerospace). The workshop will form the second day of the AGCFM's annual meeting. The meeting will comprise presentations in the morning session on solid surface and far-field boundary conditions for Euler and Navier-Stokes equations and for zonal models, e.g. matching Navier-Stokes and Euler calculations and matching near wall models with Reynolds stress models. Invited speakers include C M Albone (DRA), G P Hammond (Bath), N May (ARA Ltd), and K W Morton (Oxford). There will be a poster session before lunch. Those interested in taking part in this session should send Dr Williams a short description of their work before the end of July; he will issue acceptances by mid-August.
The afternoon session will be devoted to parallel discussion sessions on zonal models, Euler and Navier-Stokes models, and time-varying flows. There will then be reports to the meeting from each discussion group and an opportunity to discuss the points raised.
The first day of the AGCFM meeting, 21 September, will be on Future Requirements Research. Presentations will address the current state of research at DRA, ARA and BAe and consider topics such as flutter and aeroelastics for civil and military aircraft, helicopters, the inclusion of viscous effects and moving boundaries. Invited speakers include: S P Fiddes (Bristol), D G Foster (DRA), D King (BAe), J V Miller (DRA), and P D Smith (DRA).
SDAIS - STEP Data Access Interface Specification
The STEP (Standard for the Exchange of Product Data) project was originally started as a project to design mechanisms for efficient data exchange between dissimilar computer systems. Under those circumstances, there is a sender system which creates an exchange file, and a receiver system which reads this file. One of the strong aspects of the standard is the strict separation between the description of the exchange file and the description of the data to be transferred. This led to the development of a language for information modelling, Express. The data models of STEP are expressed in this Express language. The exchange file format is derived from the Express language by defining for each construct in the Express language how the construct maps onto the physical file. The data model is defined completely independent of any implementation of the data.
It is now realised that, apart from their use for setting up data exchange mechanisms, the data models which are being developed as part of the STEP project are very useful in their own right. It is anticipated that in future, application programs dealing with an area where a STEP model is available will quite often be based on this data model. This creates the interesting possibility of detaching the application program from the data it uses. One way of doing this would be to use an existing database language (like SQL) and defining for all Express constructs how they map onto SQL constructs. The disadvantage of this approach is that it is tied to one particular language, and also that SQL is not particularly suitable for dealing with data models of the type that appear in engineering applications. It also requires the possibility of embedding SQL statements in a programming language.
For these reasons it is felt that a mechanism would be required for dealing with STEP entities which is linked more closely to the Express language. This mechanism is called the STEP Data Access Interface, and its specification the STEP Data Access Interface Specification. It will hide from the application program all detail of how the data is stored. The advantages are twofold. First, the application can be developed without the programmer having to be concerned with details of the data storage; secondly, the access routines and the data repository can be specifically designed to match the Express language in an efficient manner.
In addition, it will be possible for vendors of database systems to extend their database with an implementation of the interface, and so create a product which is STEP compatible.
The anticipated way of working with the Data Access Interface is as follows. The user runs an application program (for instance a CAD system) and gives a command which requires data to be fetched. The application program creates and issues a request to the Data Access Interface. The Interface then retrieves the data from the repository. This repository could be a database, but it need not be. It will be possible for the repository to have a two-level structure, one level (presumably on disk) for long term storage, and a second level (presumably in main memory) for fast response time.
The specification of the Interface consists of a document with details of subroutine actions. Examples of the actions associated with the subroutines are:
- retrieve an entity
- store an attribute of an entity
- give information about the total number of entities of a particular type.
The description of the subroutines will be accompanied by language bindings, which give details of the subroutine call format for several programming languages. It is this specification which will be standardised.
An implementation of the interface consists of actual code of the subroutines in some language, together with a data management system. Most of the code will be implementation specific because it is closely linked with this data management system.
Jan Van Maanen, RAL
Virtual Shared Memory Seminar
Centre for Novel Computing
University of Manchester
Thursday 17 and Friday 18 September 1992
Parallel computers have been available for many years now, but have had many problems. Distributed Memory computers (such as Meiko and iPSC) are difficult to programme, while Shared Memory computers (such as Alliant and Sequent) are not scalable and SIMD computers are limited in the types of problem they can address.
Virtual Shared Memory offers the scalability of Distributed Memory with the ease of programming of Shared Memory computers.
This seminar will discuss all aspects of Virtual Shared Memory computing from hardware through operating systems and languages to algorithms and applications.
The seminar is aimed at all people working in the area of parallel computing. It will provide a detailed introduction to Virtual Shared Memory. It will be of particular interest to application developers, whether of numeric, symbolic or database applications, who need the power of parallel computing, but have been put off in the past by the difficulties of parallel computing. The seminar will also be of interest to systems implementers and architects working on parallel systems.
Speakers will include key workers in each of the areas, drawn from both industry and academia.
Ursula Hayes, Department of Computer Science, University of Manchester
Barcelona seems to be the place to be in 1992 and if you are involved with Parallel Computing, and transputers in particular, then 21-25 September is a very good time to go, for PACTA '92 - PArallel Computing & Transputer Applications 1992 - is going to be the parallel computing Conference and Exhibition of 1992.
Papers addressing practical applications dominate the programme, but with papers presenting state-of-the-art parallel computing theory and practice also figuring prominently, the programme will be of interest whatever type of parallel architecture you use. The 183-paper programme includes keynote presentations and technological presentations by some of the leading figures in parallel computing. The final afternoon session of the Conference is a Panel Session entitled Towards Terraflops-Architectures and Applications. On the last day there is a series of tutorial sessions on specialist subjects e.g. image processing, for those who are new to these areas.
PACTA'92 incorporates the 4th International Conference on Transputer Applications and as with the previous conferences there will be a large exhibition alongside; however for PACTA '92 the leading transputer suppliers are joined by Cray, Fujitsu, Digital, Thinking Machines, IBM, Texas Instruments and Intel to name a few!
And despite the massive rise in Barcelona room prices this year, PACTA has arranged accommodation from only £29 a night!
The advance programme has been issued and copies together with registration details can be obtained from Dr Mike Jane at RAL.
Benchmarking and Evaluation of Parallel Systems
BECAUSE is an ESPRIT project involved with BEnchmarking Concurrent Architectures for Use in Scientific and Engineering applications. The basic aim of the work is to assess the usefulness of presently available distributed memory computers in areas of computational field modelling. Examples of this are computational fluid dynamics (CFD), electromagnetic design and semiconductor device modelling.
The partners in the project include commercial companies such as Bertin & Cie, Vector Fields and PARSYS, along with staff from RAL, INRIA and Athens University of Economics.
The first phase of the project has focussed on the extraction of the critical parts from several real application codes, in the areas of CFD, electromagnetics and device modelling, that are currently in use by the partners. These critical parts are representative of the most time consuming operations that occur in a wide range of design and modelling codes using finite element, finite volume and related techniques. From this analysis, a series of benchmark test problems, called the BECAUSE Benchmark Set (BBS), have been defined. These are based on the extracted critical parts, along with a number of related tests of basic communication and computational performance. Each benchmark consists of documentation specifying the test along with standard Fortran code to perform the operation and timing on a serial machine. The partners are currently implementing these tests on a selection of parallel machines including the Intel iPSC/2 and iPSC/860, a Connection Machine CM2 and a PARSYS SNlOOO Supernode. In addition, the benchmark problems have been made publicly available on an ftp server at INRIA.
Though the critical parts of an application are those that one must spend the most effort on in adapting a code to run on a parallel machine (or any new architecture), they are not the whole story. It may well be that other parts of the code that are insignificant in a serial implementation become critical on a parallel machine. To this end selected whole application codes will also be posted to some of the parallel machines available within the project.
The initial results of this project will be presented during a workshop, organised by the BECAUSE partners, to be held in INRIA, Sophia-Antipolis, France, 13-16 October 1992. In addition to the measurements made by the partners, contributions are welcome from anyone who has run all or part of the BBS on a parallel machine (a call for papers is included :n this issue of ECN). As well as presenting results of the performance levels obtainable on the current generation of parallel machines, the workshop will also consider other aspects such as the ease of use of the systems and the quality of the software development environment that is provided.
Ron Fowler, Chris Greenough, Informatics
Advanced Course on Large Scale Scientific Parallel Computing
20-23 October, 1992
The course is the fourth in a series organised by ERCIM (European Research Consortium for Informatics and Mathematics). Like its predecessors it features speakers from four ERCIM institutes (RAL, CWI, GMD, INRIA) plus guest speakers with an emphasis on activities in the host country(this time the UK).
The programme is designed to teach students, researchers, scientists and engineers how to exploit modern parallel and vector computers. both present and future, in order to produce appropriate and efficient applications on these high performance machines.
The course is fully residential at The Cosener's House, SERC's own Conference Centre beside the River Thames in Abingdon.
Dr C P Wadsworth, Informatics
The EASE Education & Awareness Group publishes a very useful little book called the LaTeX Cookbook.
LaTeX is increasingly coming in to use (for the uninitiated - its a typesetting program for printing technical documents) and this book is ideal for the new user. It provides examples for the reader to experiment with, gently introducing more complicated examples to give hands-on learning. Combined with some very useful tips and an easy-to-read style it will be of value wherever LaTeX is used. What it is not, however, is a reference book.
The same group is also producing on behalf of the Information Technology Transfer Initiative (ITTI), a LaTeX course which will be available from late 1992. A pilot course (11 places only) is being run on September 17th and if you would like to take part please contact Martin Prime at RAL; note each application should be accompanied by a brief case for support - future course presenters will take preference.
- International Conference & Workshop on Neural Networks and Genetic Algorithms, University of Innsbruck, Austria, 13-16 April 1992
- Data in Visualisation Workshop, Manchester 15-16 October 1992