ParSim 2006

This special session is intended to provide a discussion forum for scientists from both the engineering disciplines and computer science in order to foster a closer cooperation between them. In contrast to traditional conferences, emphasis is put on the presentation of up-to-date results with a short turn-around time, which can also include the presentation of work-in-progress with only partial results available at the time of submission. Authors are also encouraged to present new projects and/or ideas.

After its successful introduction in 2002, the session is embedded into the EuroPVM/MPI for the fifth time. The host conference, with its long tradition as one of the European prime events in parallel computation, serves as an ideal surrounding for this special session and will enable the participants to present and discuss their work within the scope of both the session and the EuroPVM/MPI.

Topics for this special session include, but are not limited to:

  • Real-world problems from electrical, mechanical, civil, or medical engineering requiring parallel simulation techniques
  • Scalability of large-scale parallel simulations
  • Software techniques to optimize parallel simulations
  • Problem solving environments for parallel simulation problems
  • Mathematical aspects for parallel simulation approaches

Five contributions have benn selected for the 2006 edition of ParSim.

ParSim 2006Time: 20th Sep 2006, 11:25 – 13:15

Location: S1/S2

An approach for parallel fluid-structure interaction on unstructured meshes

Ulrich Küttler and Wolfgang A. Wall

The simulation of fluid-structure interaction (FSI) problems is a challenge in contemporary science and engineering. This contribution presents an approach to FSI problems with incompressible Newtonian fluids and elastic structures and discusses its realization in a general purpose parallel finite element research code. The resulting algorithm is robust and effcient and scales well on parallel machines. Recent attempts on effciency improvements are discussed and a numerical example is shown.

MPJ Express Meets Gadget: Towards a Java Code for Cosmological Simulations

Mark Baker, Bryan Carpenter, and Aamir Shafi

Gadget-2 is a massively parallel structure formation code for cosmological simulations. In this paper, we present a Java version of Gadget-2. We evaluated the performance of the Java version by running a colliding galaxy simulation and found that it can achieve around 70% of C Gadget-2’s performance.

Optimizing a Conjugate Gradient Solver with Non Blocking Collective Operations

Torsten Hoefler

This paper presents a case study about the applicability and usage of non blocking collective operations. These operations provide the ability to overlap communication with computation and to avoid unnecessary synchronization. We introduce our NBC library, a portable low-overhead implementation of non blocking collectives on top of MPI-1. We demonstrate the easy usage of the NBC library with the optimization of a conjugate gradient solver with only minor changes to the traditional parallel implementation of the program. The optimized solver runs up to 34% faster and is able to overlap most of the communication. We show that there is, due to the overlap, no performance difference between Gigabit Ethernet and InfiniBand for our calculation.

Parallel DSMC gasflow simulation of an in-line coater for reactive sputtering

Andreas Pflug, M. Siemers, B. Szyszka

There is an increasing demand for high precision coatings on large areas via in-line reactive sputtering, which requires advanced process control techniques. Thus, an improved theoretical understanding of the reactive sputtering process kinetics is mandatory for further technical improvement. We present a detailed Direct Simulation Monte Carlo (DSMC) gas flow model of an in-line sputtering coater for large area architectural glazing. With this model, the pressure fluctuations caused by a moving substrate are calculated in comparison with the experiment. The model reveals a significant phase shift in the pressure fluctuations between the areas above the center and the edges of the substrate. This is a geometric effect and is e. g. independent of the substrate travelling direction. Consequently, a long sputtering source will observe pressure fluctuations at its center and edges, which are out of phase. For a heuristic model of the reactive sputtering process, we show that in certain cases a two-dimensional model treatment is sufficient for predicting the film thickness distribution on the moving substrate. In other cases, a strong phase shift between averaged pressure fluctuations and reactive sputtering process response is observed indicating that a three-dimensional model treatment is required for a realistic simulation of the
in-line deposition process.

Parallel simulation of T-M processes in underground repository of spent nuclear waste

Jiri Stary, Ondrej Jakl, Roman Kohut

The contribution deals with mathematical (finite element) simulation of the KBS prototype nuclear waste repository in a simplified form, i.e. as a thermo-elasticity problem. It describes the solvers developed for such kind of problems and principles and benefits of their parallelization, both in MPI and OpenMP.