Contents Issue 12 June 1999
|
|
| Should we all take German lessons: a commentary on Supercomputer'99, the 14th Mannheim Conference
|
|
"Should we all take German lessons: a commentary on Supercomputer'99, the 14th Mannheim Conference",... Or, more simply, should the conference be held entirely in English which, not being superior to the "language of Goethe", has however the advantage of being universally accepted as a means of communication in the world of computers?
|
| Pallas to productise Unicore
|
|
|
The German Unicore project is aimed at producing a seamless access interface
for researchers all through
Germany to the large machines at the national
supercomputer centres. First Unicore prototypes
are now running in Munich
and Jülich. At the Supercomputer Seminar in Mannheim, Karl Solchenbach announced that his company, Pallas will bring the Unicore results as a product to the market.
|
| Building a
Petaflop computer from scratch is not that difficult
|
|
|
Building a Petaflop
machine, that is thousand times as powerful as the current fastest supercomputer within ten years, is not that difficult when you do not have the constraints by the day-to-day responsibilities inside a computer vendor company. During the evening event at the Mannheim baroque castle, Giacomo Polosa from RCI Europe introduced Steven J. Wallach, the current expert and adviser to Centerpoint Venture Partners.
As an experienced former Chief Technology
Officer of Convex Computers who has crossed many supercomputing bridges, Mr. Wallach took the audience on an adventurous trip with the task to build a Petaflop machine from tabula rasa in ten years time.
A both instructive and envisioning exercise into
the future of supercomputing.
|
| Supercomputer centres rapidly evolve towards megastores of Teraflops and Petabytes
|
|
|
As the director of the National Energy Research Scientific Computing (NERSC) Division at Lawrence Berkeley National Laboratory in California, Dr. Horst Simon is facing a new fundamental transition in the role of the supercomputing centre after a first earth quake in the mid 1990s.
Having only recently succeeded in the adoption of innovative intellectual services which meet the enhanced operational complexity of the supercomputer centre due to the birth of parallel distributed memory machines built from commodity parts, the management has to integrate the impact of three new technological realities
|
| Experiences with the Siemens/Fujitsu VPP300E with 16 processors at AUDI AG
|
|
|
Karl Huber reported that CAE applications are used along the whole product development process in different technical fields. Today this is named as digital mock-up or virtual car.
The central organised tasks help the development engineer. Central organisations plan the implementation of hard and software, and CAE is intergrated in the product development. Audi uses different software packages,
such as Nastran for structure analysis, which runs on
SGI workstations.
The Fujitsu crash simulation Pamcrash runs on Fujitsu only, and the computational fluid dynamics StarCD is recently tested on the VPP300.
Currently, it runs on SGI. Audi uses 150 CAE workstations and additionally 7 SGI SMPs with 100 R10000 processors as well as 60 software packages.
|
| CTO panel discussion
addresses supercomputing issues of the year 2010
|
|
|
In the HPC panel discussion: Meet the chief technology officers, Steve Wallach was the chairman and he commented the vectors as a reservat of whales in a national protected park. Here is an overview of the short presentations on the following topics: present and next HPC architecture, MPP vs SMP vs PVP vs Clusters, commodity vs proprietary chips, future interconnection architectures, vector vs scalar processing, Unix vs NT, when is the time to introduce 10/100 Tflop systems, MPP applications with sufficient scalable parallelism.
|
| An update on Tera
|
|
|
At the Mannheim Supercomputer'99 Conference, Primeur editor Uwe Harms had an interview with Burton Smith from Tera about the latest developments in the Tera company.
|
|
|
| Cray T3E used in direct simulation of turbulent reacting flows
|
|
|
At the Interdisciplinary Center of Scientific Computing of the University of Heidelberg, a code has been parallelized for the direct numerical simulation (DNS) of turbulent chemically reacting flows.
The application fields for this type of code relate to combustion processes in automotive engines, electrical power generation, and heating. It brought
the team the third prize in the SuParCup.
|
| Solving linear sets of equations originating from implicit finite element applications
|
|
|
Direct solvers for finite element matrices are becoming less attractive when it comes to solving huge problems.
As larger and faster computers increasingly enter the scene, scientists will have to resort to iterative methods again.
Multi-grid constitutes a highly optimal multi-level domain decomposition method.
The SuParCup Carl Benz Award winning team from the University of California in Berkeley has applied classical multi-grid methods in the solution of the large sparse system of equations arising from unstructured finite element issues in 3D elasticity.
|
| Automotive SME industry to benefit from vehicle dynamics simulation on a parallel platform for virtual prototyping
|
|
|
The Student Prize of the SuParCup'99 went to Torsten Butz from the Technical University of Munich for his research paper dedicated to the parameter estimation problem which is associated with vehicle dynamics simulations.
Engineers have to deal with this non-linear least-squares issue which can be solved with the use of mathematical optimization algorithms.
|
| Virtual Laser Plasma Lab code demontrates an excellent performance
|
|
|
Dr. Pukhov from the Max-Planck-Institut für Quantenoptik reveived an honourable mention at the SuParCup'99 Mannheimer event for his paper on the Virtual Laser Plasma Lab (VLPL) code.
|
| SuParCup's First Prize Winner 1999 PRONTO to parallelize crash simulations
|
|
|
A team from Sandia National Laboratories has successfully parallelized the PRONTO code as the only solid dynamics code to run effectively on thousands of processors.
PRONTO is used in car crash simulations and as such, has a similar scope to the familiar
DYNA, PamCrash, and ABAQUS codes.
|
| Tree type numerical algorithms successfully applied in science and engineering
|
|
|
To earn the Second prize in the SuParCup, a team from the Rheinische Friedrich-Wilhelms-Universität in Bonn has developed a cost-effective but very efficient load-balancing scheme for hierarchical tree algorithms based on space-filling curves for partial differential equations (PDEs) or N-body problems.
The parallel codes were performed on the Parnass2 cluster, consisting of Shared Memory Processor (SMP) PCs and a Myrinet network at Gigabit/s speed configured with full bisection bandwidth.
|
