Uwe Küster

Performance improvement of sparse matrix vector product on vector machines

Many applications based on finite element and finite difference methods include the solution of large sparse linear systems using preconditioned iterative methods. Matrix vector multiplication is one of the key operations that has a significant impact on the performance of any iterative solver. In this paper, recent developments in sparse storage formats on vector machines are reviewed. Then, several improvements to memory access in the sparse matrix vector product are suggested. Particularly, algorithms based on dense blocks are discussed and reasons for their superior performance are explained. Finally, the performance gain by the presented modifications is demonstrated.

Teraflops Sustained Performance With Real World Applications

This paper provides a comprehensive performance evaluation of the NEC SX-8 system at the High Performance Computing Center Stuttgart which has been in operation since July 2005. It provides a description of the installed hardware together with its performance for some synthetic benchmarks and five real world applications. All the applications achieved sustained Tflop/s performance. Additionally, the measurements presented show the ability of the system to solve not only large problems with a very high performance, but also medium sized problems with high efficiency using a large number of processors.

Industrial Turbulence Simulations at Large Scale

The most important aspect for simulations in industrial design processes is the time to solution. To obtain highly detailed results nevertheless, massive computational resources have to be deployed. Feasibility and applicability of HPC systems to this purpose is the main focus of this paper. Two different numerical approaches, implemented with parallelism in mind, are investigated with respect to quality as well as turn around times on large super computing systems. The one approach compares the efficiency of high order schemes on coarser meshes to lower order schemes on finer meshes. The second approach employs a zonal coupling of LES and RANS to limit the computational effort by using solution adapted models. Three industrial use-cases evaluate the performance and quality of these approaches. General optimizations are presented as well as solutions for load-balancing.

HPC Processor Technologies and Their Impact on Simulation

Moore’s law has come to an end with respect to the clock speed of the single processor. Clock rates are no longer increasing. Parallelism carries the day and accelerators are making the most of this. What is the future of processors for HPC going to look like? This talk will give a short overview and discuss some potential solutions.

Future Challenges in Supercomputing

ABSTRACT Supercomputing – or High Performance Computing (HPC) – has become to be dominated by standard components. A quick look at the Top500 list shows that clusters built from such standard components have become the architecture of choice. The fraction of clusters in the list has increased from about 2% in 2000 to about 73% in 2006. The key driving factor is the availability of competitive processor technology in the mass market on the one hand and a growing awareness of this potential in the user community on the other hand. These trends have motivated small companies to seek business opportunities in building large systems from standard components. Such commodity based systems rely mainly on massive parallelism in order to compete with traditional approaches. At the same time other approaches have been brought into the supercomputing arena which promise to allow for another leap in peak performance. This paper sets out to describe the current situation in supercomputing from a point of view of architectures and to describe futures scenarios and challenges that arise from the current trends in hardware and software development.

The Role of Supercomputing in Industrial Combustion Modeling

Currently, numerical simulation using automated parameter studies is already a key tool in discovering functional optima in complex systems such as biochemical drug design and car crash analysis. In the future, such studies of complex systems will be extremely important for the purpose of steering simulations. One such example is the optimum design and steering of computation equipment for power plants. The performance of today’s high performance computers enables simulation studies with results that are comparable to those obtained from physical experimentation. Recently, Grid technology has supported this development by providing uniform and secure access to computing resources over wide area networks (WANs), making it possible for industries to investigate large numbers of parameter sets using sophisticated optimization simulations. However, the large scale of such studies requires organized support for the submission, monitoring, and termination of jobs, as well as mechanisms for the collection of results, and the dynamic generation of new parameter sets in order to intelligently approach an optimum. In this paper, we describe a solution to these problems which we call Science Experimental Grid Laboratory (SEGL). The system defines complex workflows which can be executed in the Grid environment, and supports the dynamic generation of parameter sets.

Influence of Fortran 90 Features on Performance on Cray Vector Computer Systems

To formulate data structures suited to numerical problems like selfadapting multilevel algorithms several data constructs, conditions and attributes are necessary. These constructs are offered by languages like C++ or Fortran 90.