WE2IOpk  —  Trends in High Performance Computing II (plenary)   (02-Sep-09   10:00—11:40)

Paper Title Page
WE2IOPK01 Hard- and Software-based Acceleration Techniques for Field Computation 93
 
  • M. Schauer, P. Thoma
    CST, Darmstadt
 
 

Due to high demand in more realistic graphics rendering for computer games and professional applications, commercial, off-the-shelf graphics processing units (GPU) increased their functionality over time. Recently special application programming interfaces (API) allow programming these devices for general purpose computing. This talk will discuss the advantages of this hardware platform for time domain simulations using the Finite-Integration-Technique (FIT). Examples will demonstrate typical accelerations over conventional central processing units (CPU). Next to this hardware-based accelerations for simulations also software-based accelerations are discussed. A distributed computing scheme can be used to accelerate multiple independent simulation runs. For memory intense simulations the established Message Passing Interface (MPI) protocol enables distribution of one simulation to a compute cluster with distributed memory access. Finally, the FIT framework also allows special algorithmic improvements for the treatment of curved shapes using the perfect boundary approximation (PBA), which speeds up simulations.

 
WE2IOPK03 Graphical Processing Unit-Based Particle-In-Cell Simulations 96
 
  • V.K. Decyk, S.A. Friedman, T.V. Singh
    UCLA, Los Angeles, California
 
 

New emerging multi-core technologies can achieve high performance, but algorithms often need to be redesigned to make effective use of these processors. We will describe a new approach to Particle-in-Cell (PIC) codes and discuss its application to Graphical Processing Units (GPUs). We will conclude with lessons learned that can be applied to other problems. Some of these lessons will be familiar to those who have programmed vector processors in the past, others will be new.

 
WE2IOPK05 VizSchema - A Standard Approach for Visualization of Computational Accelerator Physics Data 101
 
  • S.G. Shasharina, J.R. Cary, M.A. Durant, S.E. Kruger, S.A. Veitzer
    Tech-X, Boulder, Colorado
 
 

Even if common, self-described data formats are used, data organization (e.g. the structure and names of groups, datasets and attributes) differs between applications. This makes development of uniform visualization tools problematic and comparison of simulation results difficult. VizSchema is an effort to standardize metadata of HDF5 format so that the entities needed to visualize the data can be identified and interpreted by visualization tools. This approach allowed us to develop a standard powerful visualization tool, based on VisIt, for visualization of large data of various kinds (fields, particles, meshes) allowing 3D visualization of large-scale data from the COMPASS suite (VORPAL and Synergia) for SRF cavities and laser-plasma acceleration.

 

slides icon

Slides