by Uwe Harms
Munich, GERMANY — This year Professor Arndt Bode and his team from the Informatics Department at Technical University Munich, Lehrstuhl f¸r Rechnertechnik und Rechnerorganisation (LRR), successfully organised the Euro-Par 2000 Conference. Some years ago several independent conferences with the theme parallel processing decided to form only one bigger conference. This year it was the 6th edition of the Euro-Par Conference. It attracted more than 400 participants from all over the world. Additionally to the six keynote and invited talks, 167 talks have been accepted from a total of 326 papers. Additionally other interest groups connected to the conference, SCI Europe 2000 (Scalable Coherent Interface), proceedings are available, and APART 2000, Automatic Performance Analysis.
The 167 talks can be grouped in 20 topics:
– Support Tools and Environments
– Performance Evaluation and Prediction
– Scheduling and Load Balalancing
– Compilers and High Performance
– Parallel and Distributed Databases and Applications
– Complexity Theory and Algorithms
– Applications on High-Performance Computers
– Parallel Computer Architectures
– Distributed Systems and Algorithms
– Parallel Programming: Models, Methods and Programming Languages
– Numerical Algorithms for Linear and Nonlinear Algebra
– European Projects
– Routing and Communication in Interconnection Networks
– Instruction-Level Parallelism and Architecture
– Object Oriented Architectures, Tools and Applications
– Architectures and Algorithms for Multimedia Applications
– Cluster Computing
– Parallel I/O and Storage Technology
– Problem Solving Environments
Six invited talks highlighted the Performance of Parallel Simulations based on Partial Differential Equations, E2K Technology and Implementation (Elbrus International), Grid-based Computing, Communication Modes for Parallel Computing, Moore’s Law and the Top500 project. David Keyes talk, Old Dominion Univ. Norfolk, Law. Livermore Lab and Nasa, discussed “Four Horizons for Enhancing the Performance of Parallel Simulations Based on Partial Differential Equations (PDE)”. He proposed four possible ways:
1. Expand the number of processors
2. More efficient use of faster processors
3. More architecture-friendly algorithms
4. Algorithms delivering more “Science per Flop”.
For 2. he noted that with highly parallel machines he only gets between 8% and 27% of the peak performance for a highly L1- and register-optimised unstructured grid Euler flow code. In 3. he gained a performance improvement of a factor 7 down to nearly 3, when spatial reordering tooks place.
Gregor von Laszewski, Argonne Nationa Lab., discussed grid computing and the thread migration system MOBA (Mobile Agents). He presented the components and its migration into Java Virtual Machines.
The proceedings, more than 1360 pages, are published with Springer-Verlag in Lecture Notes in Computer Science, Vol. 1900.
In the registration area some vendors presented their products in an exhibition for example Fujitsu Siemens Computers, AEA Technology, Dolphin, RasDaMan and KONWIHR in the reguistration area.
In parallel the 3rd International Conference on SCI-based Technology and Research took place. It was a combination of users talks and for examle vendor presentations from Dolphin, SCALI abd ICL. Fourteen talks covered the topics:
– Message Passing Programming
– Middleware for SCI based clusters
– SCI Monitoring
– SCI Hardware Developments
– SCI-based Data Acquisition in High-Energy Physics
– SCI Switch Evaluation and Design.
The Proceedings are available from Dr. Wolfgang Karl (LRR), email: [email protected] .