Modeling giant planets without cores is relevant for extrasolar giant planets as well as for Jupiter. We present three-dimensional numerical simulations of thermal convection in a fully convective (without a solid core), non-magnetic, rotating, density-stratified, spherical fluid body. Discontinuous axially aligned vortices spiral prograde, eastward, momentum away from the axis of rotation as a result of vorticity generated by fluid flowing through the density-stratification. The convergence of this nonlinear Reynolds stress maintains a banded pattern of differential rotation with a strong prograde jet at the equator, without the classical vortex stretching of convective columns.

The main specific features of high performance scientific simulation consist of obtaining optimal performance levels, sustainability of the codes and models and the use of dedicated architectures. The construction of codes and of the physical phenomena to be simulated requires substantial investments both in terms of human resources and of the experiments required to validate the models. The need for increasingly refined modeling leads to optimization of the performance levels of the codes to obtain simulation results within acceptable calculation times. Finally, the necessary recourse to highly specialized hardware resources adds further constraints to the programming of such software, especially when the lifetime of the codes, often close to 20 years, is compared to that of super computers, in the order of just 4 years. The MDA approach (Model Driven Architecture), standardized by the OMG and based on UML 2.0, provides among things a generic method for obtaining a runnable code from a UML model describing a system on several levels of modeling. Its various abstraction mechanisms should make it possible to increase the lifetime of our codes by facilitating the porting operations while at the same time improving performance thanks to the capitalization of good programming practices. MDA technologies have already proved their value in other fields of the computer industry. The sector of real time systems, for example, is based on UML to define a profile specific to their programming constraints (size on the on board code, limited resources, rapidity of upgrades, etc.). By analogy to these other sectors of the computer industry, we have therefore chosen to adopt the definition of a UML profile specific to the constraints for the development of high performance scientific simulation. This approach is explained in the article which addresses in turn the following points: definition of a meta-model for high performance simulation, the use of proven technologies (Topcased, Acceleo) for the automatic transformation of models with in particular the automatic generation of a Fortran code, and finally an overall illustration of an implementation of this profile.

The European high field superconducting dipole magnet, currently under development, will create magnetic fields of up to 12.5 T for performance test of high current superconducting cables (up to 100 kA). To study the behavior of the dipole during a quench (energy stored is 16 MJ at 16.5 kA) a complex simulation model was developed and integrated in the 1-D thermal hydraulic code THEA by CryoSoft.

OPAL is a tool for charged-particle optics in accelerator structures and beam lines. The parallel part of OPAL is based on IPPL (Independent Parallel Particle Layer) a framework which provides parallel particles and fields using a data parallel ansatz. OPAL is built from the ground up as a parallel application exemplifying the fact that HPC (High Performance Computing) is the third leg of science, complementing theory and the experiment. OPAL runs on your laptop as well as on the largest HPC clusters available today. We will show the architecture of OPAL, focusing on parallel issues of the FFT and MG based elliptic solvers and show applications w.r.t . existing and future particle accelerators at PSI.

The Swiss National Grid Association (SwiNG, http://www.swing-grid.ch) is the National Grid Initiative of Switzerland. It counts as its members 19 Swiss academic institutions, including Swiss Federal and Cantonal Universities, Universities of Applied Sciences, several research institutes, CSCS and SWITCH. SwiNG's mission is to

• Ensure competitiveness of Swiss science, education and industry by creating value through resource sharing.
• Establish and coordinate a sustainable Swiss Grid infrastructure as a dynamic network of resources across different locations and administrative domains.
• Provide a platform for interdisciplinary collaboration to leverage the Swiss Grid activities, supporting end-users, researchers, education centers, resource providers and industry.
• Represent the interests of the national Grid community towards other national and international bodies.

The theoretical study of the adsorption of organic molecules on platinum and on other transition metal surfaces is of fundamental importance for the investigation of reaction mechanisms in heterogeneous catalysis [1-3].
Density Functional Theory is the standard tool for such studies, since the systems are typically too large for analysis with post-Hartree-Fock calculations.
We have investigated several methods for the description of a metal surface using (i) periodic slabs and (ii) clusters for the modeling of the metal, and (i) Gaussian-Plane-Wave, (ii) Plane Waves, or (iii) localized basis sets for the expansion of the wave function [4-6]. Physico-chemical properties of the metal as well as the adsorption properties of organics are compared using the different methods. Applications to the study of chiral surface modification of platinum catalysts complements the theoretical investigation [7].

• [1] J.K. Norskov, M. Scheffler, H. Toulhoat, MRS Bull., 2006, 31, 669.
• [2] R.A. van Santen, M. Neurock, Handbook of Heterogeneous Catalysis (Eds. G. Ertl, H. Knözinger, J. Weitkamp), Wiley-VCH, Weinheim, 1997, Vol. 3, p. 942.
• [3] A. Gross, Theoretical Surface Science, Springler-Verlag, Berlin Heidelberg New York, 2003.
• [4] N. Bonalumi, A. Vargas, D. Ferri, A. Baiker, J. Phys. Chem. B, 2006, 110, 9965.
• [5] G. Santarossa, A. Vargas, M. Iannuzzi, A. Baiker, ChemPhysChem, 2008, 9, 401.
• [6] G. Santarossa, C. Pignedoli, M. Iannuzzi, D. Passerone, A. Vargas, A. Baiker (In preparation).
• [7] A. Vargas, G. Santarossa, M. Iannuzzi, A. Baiker, J. Phys. Chem. C, 2008 (In press).

Ongoing biomedical research often requires imaging of large and thick specimen. Such specimens do not fit into the field of view (FOV) of a standard confocal microscope. Moreover the variations in the fluorescence intensities (due to absorption effects) can not be resolved with a single recording. To overcome these limitations a recombination of multiple recordings can be done. While software for such a recombination exists, we have often found it to be limited in applicability. Furthermore recent objectives provide a distortion-free image even at the border of the field of view (FOV). Such datasets allow the use of new fast, efficient and robust techniques.

Hyperthermia treatment is a promising option in oncology. By heating the tumor using electro-magnetic energy, it is made more susceptible to an accompanying radio or chemo therapy. The problem addressed in this project is a large-scale optimal control problem for finding the therapeutical optimal antenna parameters given the patient geometry. The goal is optimization in real-time using HPC hardware (Cell BE, GPUs) so that the physician could react to patient feedback during treatment. We are using primal-dual interior point methods as the most efficient methods for solving these nonlinear nonconvex programming problems. These resulting optimization problems are computationally demanding and require special algorithmic solution schemes that are addressed in the research project.

The Parallel ROOT Facility, PROOF, enables the inter- active analysis of very large distributed data sets in a transparent way. It exploits the inherent parallelism in data sets of uncorrelated events via a multi-tier architecture that optimizes I/O and CPU utilization in heterogeneous clusters with distributed storage. On a grid, PROOF can use the available services to find out the location of data to analyze and the resources to use. Dedicated PROOF-enabled clusters are now being deployed for analyzing the (now imminent) data of the LHC experiments. Being part of the ROOT framework PROOF inherits the benefits of a high performance object-oriented storage system and a wealth of statistical and visualization tools. The basic ideas and architecture underlying the PROOF project are described, together with the status of the project, focusing mostly on the issues of data access optimization, scheduling of user sessions and user interface.

S. Chikatamarla(1), C. Frouzakis(1), I. Karlin(1), A. Tomboulides(2), K. Boulouchos(1)

1. LAV, Institute of Energy Technology, ETH Zurich, Switzerland
2. Department of Engineering and Management of Energy Resources, University of Western Macedonia, Greece

Fluid turbulence remains a key problem for science and engineering, and has been a grand challenge for both theoretical and computational techniques. Direct numerical simulation of the Navier-Stokes equations using higher order methods, like spectral methods; have long been accepted as best techniques for these problems. However, their complexity (especially with increasing geometric complexity) and restricted scaling as the number of compute cores increase has kept them out of reach of many computational fluid dynamics (CFD) practitioners. In the last decade, simple and efficient kinetic-theory-based methods like lattice Boltzmann (LB) methods are emerging as an alternative to the continuum Navier-Stokes equations for fluid mechanics.
In LBM, one does not attempt a direct discretization of the Navier-Stokes equations; instead, a kinetic equation of Boltzmann type (with a small number of discrete velocities) is solved numerically on a regular grid. These methods are blessed with many advantages (over traditional methods) such as, easy and fast implementation for the most complex flow geometries, straightforward inclusion of complex physics with a guarantee of physical and mathematical well-posedness and realizability. Also, extremely good scalability of these methods on various modern computer architectures makes them attractive to fluid mechanists and computer science engineers. While traditional methods are limited to parallel efficiencies of 10% or less on ultra-parallel computers, in LBM, efficiencies of 30-50% or more are regularly achieved; also, in contrast to traditional approaches, LBM can work well with 32 bit arithmetic[1].
Our approach here is based on a new entropic version of the lattice Boltzmann method, which promises strong stability and thermodynamic consistency of simulations[2,3]. Here we shall establish by direct computational validations, both the speed and accuracy of entropic LB methods on large parallel machines. Apart from problems of engineering interest, high resolution simulations of classical benchmark turbulence problems like Kida-vortex flow and Taylor-Green vortex flow are studied in detail. Along with optimization strategies, parallel performance of the LB methods on various architectures will be presented.

1. Chikatamarla S., Bhaskar G., Babu V., Strenski D., Int. J. of High Performance Computing Applications, Vol 20 (4), pp 557-570 (2006).
2. Chikatamarla, S., Karlin, I., Phys. Rev. Lett., Vol 97 (19): No. 190601 (2006).
3. Chikatamarla, S., Ansumali, S., Karlin, I. Phys. Rev. Lett., Vol 97 (1): No. 010201 (2006).

S. Kerkemeier(1), C.E. Frouzakis(1), A.G. Tomboulides(2) , P.F. Fischer(3), K. Boulouchos(1)

1. Aerothermochemistry and Combustion Systems Laboratory,Swiss Federal Institute of Technology, Zurich, Switzerland
2. University of Western Macedonia, Kozani, Greece
3. Mathematics and Computer Science Division, Argonne National Laboratory, U.S.A

About 80% of our total energy needs for transportation or energy production results from combustion. Even modest gains in the efficiency of combustion translate into significant energy savings and reduced pollutant emissions. Further development of next generation low-emission devices like Homogeneous Charge Compression Ignition (HCCI) engine, Lean Premixed Prevaporized (LPP) turbines in terms of enhanced performance and efficiency and reduced pollutant emissions can be significantly aided by our ability to better understand and predict autoignition in the presence of fluctuations of velocity, temperature and composition. At the same time, turbulent autoigniting flows pose a challenging problem due to the direct coupling between turbulent mixing and the complex chemistry of the slow pre-ignition reactions. Recent advances in computational power together with high-order scalable algorithms have enabled simulations which resolve all relevant fluid and chemical time and length scales. The aim of the this project is to provide fundamental understanding of autoignition using direct numerical simulation on large parallel machines. In this poster we describe recent algorithmic developments, optimization strategies and performance considerations of our spectral element code for low Mach number reactive flows, preliminary results from 3-D autoignition simulations and the associated challenges to post-process and analyze the results of these large-scale computations.