Index of Contributors

A

• Abraham Mark MS Presentation
  Monday, June 1, 2015
  HG E1.2, 16:30-17:00
  
  MS Presentation
  
  The GROMACS Road to Performance-Portable Programming Abstractions, Mark Abraham (KTH, Sweden)
  
  Co-Authors: Roland Schulz (Oak Ridge National Laboratory, USA); Teemu Murtola (Royal Institute of Technology, Sweden); Berk Hess (Royal Institute of Technology, Sweden); Erik Lindahl (Royal Institute of Technology, Sweden)
  
  Scientists' computing resources are often both scarce and expensive, so high absolute performance is an ongoing key development objective for the GROMACS molecular simulation package. The increasingly wide range and heterogeneity of HPC hardware adds considerably to this challenge, and doubly so when such high portability is also required. It is critical to choose algorithms that match the hardware well. It is also necessary to implement them via abstractions appropriate to the computational task at hand, if we are to do so for reasonable development costs. This talk will discuss recent work in GROMACS that lifts the abstraction level to permit ongoing high performance and high portability.
• Afanasiev Michael MS Presentation
  Monday, June 1, 2015
  HG F3, 14:30-15:00
  
  MS Presentation
  
  A Data-Comprehensive Seismic Earth Model across the Scales, Michael Afanasiev (ETH Zurich, Switzerland)
  
  Co-Authors: Andreas Fichtner (ETH Zurich, Switzerland); Daniel Peter (ETH Zurich / Università della Svizzera italiana, Switzerland); Korbinian Sager (ETH Zurich, Switzerland); Saulé Zukauskaité (ETH Zurich, Switzerland); Laura Ermet (ETH Zurich, Switzerland)
  
  The Comprehensive Seismic Earth Model (CSEM) is a 3D multi-scale visco-elastic model of Earth's interior. Designed to combat scaling issues in seismic inversion, the CSEM incorporates models generated from a variety of forward problem solvers, on a variety of spatial scales. Approaching global-scale seismic inversion in this manner allows us to exploit the benefits of both high resolution numerical techniques, such as full waveform inversion via the spectral element method, and more classical inversion techniques, such as traveltime tomography, where each may contribute model updates within their specific regimes of validity.
  Poster
  
  Poster
  
  EAR-07 Towards a Data-Comprehensive Earth Model Across the Scales, Michael Afanasiev (ETH Zurich, Switzerland)
  
  Co-Authors: Andreas Fichtner (ETH Zurich, Switzerland); Daniel Peter (Università della Svizzera italiana, Switzerland/ ETH Zurich, Switzerland); Korbinian Sager (ETH Zurich, Switzerland); Saulé Zukauskaité (ETH Zurich, Switzerland); Laura Ermet (ETH Zurich, Switzerland)
  
  The 'Comprehensive Earth Model' (CEM) is a solver-independent seismic multi-scale model of the global distribution of density and visco-elastic parameters. Constructed from a locally-refined finite-element mesh, the model attempts to represent the Earth on all seismically accessible scales, with contributions from various inversion techniques. We report on the current state of the model, along with the results of a global scale full-waveform inversion, which makes use of the GPU-accelerated spectral-element code SPECFEM3D_GLOBE.
• Alexeev Dmitry Poster
  
  Poster
  
  MAT-22 Liquid Layering Effects on the Kapitza Resistance Between Few-Layer Graphene and Water, Dmitry Alexeev (ETH Zurich, Switzerland)
  
  Co-Authors: Jie Chen (ETH Zurich, Switzerland); Jens H. Walther (ETH Zurich, Switzerland/ Technical University of Denmark, Denmark); Konstantinos P. Giapis (California Institute of Technology, USA); Petros Koumoutsakos (ETH Zurich, Switzerland)
  
  Kapitza resistance (R) between few-layer graphene (FLG) and water is systematically studied using molecular dynamics simulations. In contrast to monoatomic liquid/solid interfaces, R between FLG and water is found to decrease with temperature. The R exhibits a distinct dependence on the thickness of the FLG and the water layer that is attributed to large differences in their phonon mean free path. The density layering of the water near the interface has a significant impact on R with higher water density corresponding to a lower R. Based on these findings we propose two mechanisms to regulate the water layering as means to engineer the thermal transport properties of solid-liquid interfaces.
• Antoniadis Antonios MS Presentation
  Tuesday, June 2, 2015
  HG E1.1, 11:00-11:30
  
  MS Presentation
  
  High-Order Schemes for Schemes for Atmospheric Dynamics on Unstructured Meshes, Antonios Antoniadis (Cranfield University, United Kingdom)
  
  Co-Authors: Panagiotis Tsoutsanis (Cranfield University, United Kingdom)
  
  This paper presents an extension of the Weighted Essentially Non-Oscillatory (WENO) type schemes for the non-hydrostatic compressible Euler equations in conjunction with two and three-dimensional unstructured meshes. The schemes are suitable for regional and global climate models dynamical cores. Their potential lies in their simplicity; accuracy; robustness; non-oscillatory properties; versatility in handling any type of grid topology; computational and parallel efficiency. The WENO schemes (up to 5th -order) are applied to two- and three-dimensional test cases: a 2D rising thermal bubble; the 2D density current and the 3D Robert smooth bubble.

B

• Baczynski Christian MS Presentation
  Tuesday, June 2, 2015
  HG E3, 12:00-12:30
  
  MS Presentation
  
  Chemistry-Coupled Radiative Transfer in Flash4: Towards Fully Self-consistent Massive Stellar Feedback on the ISM, Christian Baczynski (University Heidelberg, Germany)
  
  Co-Authors: Simon C. O. Glover (University of Heidelberg, Germany); Ralf S. Klessen (University of Heidelberg, Germany)
  
  We present a chemistry-coupled radiative transfer scheme for the magnetohydrodynamical simulation code FLASH4. We briefly describe its design philosophy and implementation, and show how non-equilibrium effects captured by the on-the-fly solution of a fast hydrogen chemical network changes the expansion behavior of ionization fronts obtained from simpler approximations. We characterize the differences by modeling a massive star embedded in a molecular clump and furthermore look at the combined expansion of ionization- and photon dominated region (PDR) fronts. We find that the pre-heating by UV-photons changes the stability of the shell typically found around HII regions.
• Bader Michael MS Presentation
  Tuesday, June 2, 2015
  HG E1.1, 11:30-12:00
  
  MS Presentation
  
  2D Adaptivity for 2.5D Flow Problems, Michael Bader (Technische Universität München, Germany)
  
  Co-Authors: Kaveh Rahnema (Technische Universität München, Germany); Oliver Meister (Technische Universität München, Germany)
  
  Many geophysical fluid dynamics problems feature a noticeably smaller extension in the vertical dimension compared to the horizontal dimensions. For such scenarios, we evaluate the potential of 2.5D grids that extend dynamically adaptive triangular meshes by a uniformly refined third dimension. While we tolerate a slight increase in the total number of degrees of freedom, we aim at improved performance due to vectorisation along the vertical dimension. Examples will be presented for simulation of tsunamis and porous media flow.
• Ban Nikolina Contributed Talk
  Wednesday, June 3, 2015
  HG E21, 10:50-11:10
  
  Contributed Talk
  
  Is Convection-resolving Resolution Necessary for Climate Simulations?, Nikolina Ban (ETH Zurich, Switzerland)
  
  Co-Authors: Jürg Schmidli (ETH Zurich, Switzerland); Christoph Schär (ETH Zurich, Switzerland)
  
  Convection-resolving models (CRMs) operate at very high resolution (grid spacing between 1 and 3km) where convective processes (i.e., thunderstorms and rain showers) can be explicitly represented. The use of CRMs is feasible in numerical weather prediction since recently, but is computationally challenging in climate applications. Here we present pilot CRM climate simulations at horizontal resolution of 2.2km across a greater Alpine region conducted on a Cray XE6 system using a setup with 2000 cores on a grid of 500x500x60 points. The results demonstrate the importance of high resolution in climate simulations for assessing short-term precipitation events.
• Bank Claudia MS Presentation
  Tuesday, June 2, 2015
  HG E21, 12:00-12:30
  
  MS Presentation
  
  A Bayesian MCMC Method to Estimate Selection Coefficients from Mutagenesis Experiments: Insights in to Adaptation, Claudia Bank (EPFL, Switzerland)
  
  Co-Authors: Sebastian Matuszewksi (EPFL, Switzerland); Jeffrey D. Jensen (EPFL, Switzerland)
  
  With the advent of next-generation sequencing technologies and advances in bioengineering, the field of experimental evolution is developing novel approaches that allow for unique opportunities to study adaptation. In particular, genetic mutations can be systematically engineered and tracked in bulk population competitions, allowing for the simultaneous assessment of hundreds of mutational effects. Here, we present a Bayesian Monte Carlo Markov Chain approach that allows for the estimation of fitness effects and experimental power from such large-scale data. Results are presented in the framework of classical evolutionary models, many of which can now be experimentally evaluated.
• Bastian Peter MS Presentation
  Tuesday, June 2, 2015
  HG E1.2, 10:30-11:00
  
  MS Presentation
  
  High-performance Computing for Flows in Porous Media , Peter Bastian (Universität Heidelberg, Germany)
  
  Co-Authors: Steffen Müthing (Universität Heidelberg, Germany)
  
  This talk will focus on the efficient solution of porous media flow problems with discontinuous Galerkin methods. For the fast solution of the linear systems arising in the elliptic flow equation a hybrid preconditioner based on subspace correction in the conforming finite element subspace is employed. The transport equation is solved with explicit methods as part of an operator splitting scheme for the coupled system. The performance of the implementation based on the sum factorization approach is assessed on various architectures and computational results for density-driven flow and miscible displacement are presented.
• Bauer Peter MS Presentation
  Tuesday, June 2, 2015
  HG E1.1, 10:30-11:00
  
  MS Presentation
  
  High-Performance Computing and Big Data Challenges for NWP and Climate, Peter Bauer (ECMWF, United Kingdom)
  
  Co-Authors:
  
  There is growing concern that the energy cost to produce weather forecasts, to quantify their uncertainty, and to handle the associated I/O load from enhanced observing systems as well as large forecast ensembles will be unaffordable with current HPC and data management technologies. NWP centres face enormous challenges due to the rising cost of energy associated with running complex high-resolution models on more and more processors in time for delivering forecasts with substantial societal benefit.
• Becsek Barna Poster
  
  Poster
  
  LS-04 FEM/FD Immersed Boundary FSI Simulations, Barna Becsek (Universiy of Bern, Switzerland)
  
  Co-Authors: Lea Conen (Università della Svizzera italiana, Switzerland); Rolf Krause (Università della Svizzera italiana, Switzerland); Dominik Obrist (University of Bern, Switzerland)
  
  Immersed boundary simulations have been under development for physiological flows, allowing for elegant handling of fluid-structure interaction modelling with large deformations due to retained domain-specific meshing. We couple a structural system in Lagrangian representation that is formulated in a weak form with a Navier-Stokes system discretized through a finite differences scheme. We build upon a proven highly scalable imcompressible flow solver that we extend to handle FSI. We aim at applying our method to investigating the hemodynamics of Aortic Valves. The code is going to be extended to conform to the new hybrid-node supercomputers.
• Bejarano Jeremiah MS Presentation
  Tuesday, June 2, 2015
  HG F1, 11:30-12:00
  
  MS Presentation
  
  Solving Optimal Taxation Models through Scalable, Reusable Simulation, Jeremiah Bejarano (University of Chicago, USA)
  
  Co-Authors: Christian Baker (Brigham Young University, USA); Richard W. Evans (Brigham Young University, USA); Kenneth L. Judd (Stanford University, USA); Kerk L. Phillips (Brigham Young University, USA)
  
  Mirrlees (1971) first formulated optimal taxation of a heterogeneous population as the solution to a bi-level optimization problem. This formulation is practical to solve only when people differ in a single dimension. The multidimensional problem lacks an organizing principle to simplify an otherwise enormous number of constraints, and solutions often do not satisfy constraint qualification. We present a method to circumvent these issues by decomposing the problem into two steps: one-time creation of a reusable sample of the impacts of various policies and a quick search that takes an interchangeable distribution function over the heterogeneity and finds the corresponding optimal policy.
• Bekas Costas Minisymposium
  Minisymposium
  
  MS17 Big Data Analytics in Science
  
  Organizer: Costas Bekas (IBM Research, Switzerland)
  
  Co-Organizers:
  
  Big Data driven scientific discovery presents a new paradigm next to theory, experimentation and simulation. It relies on the foundational concept of discovery pipelines that start with data acquisition and curation, introduce knowledge representation and exploration and finally lead to scientific discovery. Thus, data flow as well as advanced and deep analytic algorithms and new appropriate computer architectures will be central in our discussion. This new paradigm challenges the state of the art in Big Data and computing, as it demands unprecedented degrees of integration and interoperability as well as new levels of combined computational and data centric performance.
• Benedetto Antonio Poster
  
  Poster
  
  MAT-06 Computer Simulation and Neutron Scattering Investigations of Phospholipid Bilayers in Water Solutions of Room-Temperature Ionic Liquids, Antonio Benedetto (Paul Scherrer Institut, Switzerland)
  
  Co-Authors: Pietro Ballone (Università di Roma La Sapienza, Italy)
  
  The recent development of the so-called ionic liquid (IL) family has expanded the number of ionic systems that could be used to affect the behaviour of biosystems through the properties of their interfacial water. Our study concerns the microscopic mechanisms underlying these effects, and relies on the combination of neutron scattering and molecular dynamics (MD) simulations. We present the results for the interaction of imidazolium-based ILs with phospholipid bilayers [JPhysChemB 2014, 118, 12192]. Neutron scattering and MD simulations (i) confirm the tendency of cations to be absorbed into the lipid phase, enhancing the penetration of water that (ii) apparent changes its relaxation time.
• Berzins Martin MS Presentation
  Monday, June 1, 2015
  HG E1.2, 16:00-16:30
  
  MS Presentation
  
  Using Abstractions in Heterogeneous Architectures for Scalable Thermal Radiation Calculations with AMR and the Uintah Framework, Martin Berzins (University of Utah, USA)
  
  Co-Authors:
  
  The use of abstractions in achieving performance portability for challenging petascale and beyond calculations is considered in the context of the Uintah framework applied to an engineering design study with one trillion variables that concerns a large 6000 cubic M clean coal-boiler currently under design for power generation. The role of task-based abstractions, runtime systems and portable execution layers such as Kokkos will be explored on heterogenous architectures consisting of CPUs GPUs and Xeon Phis. The success of this approach will be demonstrated through scalability results involving the most challenging part of the calculation involving a thermal radiation calculation that coupled all variables.
• Bianco Mauro Poster
  
  Poster
  
  CLI-04 GridTools: A Tool for Stencil Methods on Grids, Mauro Bianco (CSCS / ETH Zurich, Switzerland)
  
  Co-Authors: Lucas Benedicc (CSCS / ETH Zurich, Switzerland); Paolo Crosetto (ETH Zurich, Switzerland); Oliver Fuhrer (MeteoSwiss, Switzerland); Carlos Osuna Escamilla (ETH Zurich, Switzerland); Thomas C. Schulthess (CSCS / ETH Zurich, Switzerland)
  
  GridTools develops a set of tools dedicated to the stencil motifs for regular and block-structured grid applications using finite difference or finite volume real world applications. The tools will be provided as C++ template libraries in order to allow: 1) understandability and maintainability of codes; 2) portable efficiency of execution; 3) easy interoperability with other libraries; 4) relying on well established development tools. GridTools supports multicores and GPUs implementing algorithms with performance aligned with STELLA. The GrdTools APIs guarantee productivity and customization of the user programs, thus providing an effective tool for developing grid applications.
• Bignucolo Olivier Poster
  
  Poster
  
  LS-03 Backbone Hydration Determines the Folding Signature of Amino Acid Residues, Olivier Bignucolo (University of Basel, Switzerland)
  
  Co-Authors: Hoi Tik Alvin Leung (University of Basel, Switzerland); Stephan Grzesiek (University of Basel, Switzerland); Simon Bernèche (Swiss Institute of Bioinformatics, Switzerland)
  
  The relation between the sequence of a protein and its structural propensities remains largely unknown. In order to elucidate the side-chain dependent process, we studied peptides of the sequence EGAAXAASS (X = Gly, Ile, Tyr, Trp) through precise comparison of molecular dynamics trajectories and NMR residual dipolar coupling measurements. We show that the formation of internal hydrogen bonds underlying a helical-turn is key to reproduce the experimental values. The computer simulations reveal that the driving force arises from the lack of hydration of the peptide chain on either side of the bulky side-chain, which can act as a nucleation point initiating the folding process.
• Bocharov Dmitry Poster
  
  Poster
  
  MAT-11 EXAFS Spectra Interpretation Using Molecular Dynamics and DFT Simulations, Dmitry Bocharov (Paul Scherrer Institut, Switzerland)
  
  Co-Authors: Matthias Krack (Paul Scherrer Institut, Switzerland); Alexei Kuzmin (University of Latvia, Latvia)
  
  Extended X-Ray Absorption Fine Structure EXAFS spectroscopy is a powerful tool to study the local atomic structure of materials. The interpretation of the EXAFS spectra is a non-trivial task which is addressed in our study using two approaches, combining ab initio EXAFS theory with molecular dynamics simulations or first-principles electronic structure calculations. The atomic scale simulations were performed by the CP2K code for two systems: (i) pure and Cr-doped UO2 which are employed as nuclear fuel and (ii) pure ScF3 which attracted attention due to its negative thermal expansion over a wide range of temperatures. The advantages and limitations of both approaches are discussed.
• Boehm Christian MS Presentation
  Monday, June 1, 2015
  HG F3, 13:30-14:00
  
  MS Presentation
  
  Data Assimilation for Source Encoding Strategies in Full-Waveform Seismic Inversion, Christian Boehm (ETH Zurich, Switzerland)
  
  Co-Authors: Andreas Fichtner (ETH Zurich, Switzerland); Alessandro Lechmann (ETH Zurich, Switzerland)
  
  Full-Waveform seismic tomography infers the material properties of the Earth based on the observation of seismograms. Since the costs of solving this inverse problem are roughly proportional to the number of seismic events, source-encoding strategies can reduce the computational effort substantially. We exploit statistical properties of a sample average approximation to gradually incorporate more information into the problem formulation and to update the weights of the encoded sources. This alternates with a Newton-type method based on mini-batch Hessian approximations to update the material model.
  Minisymposium
  Minisymposium
  
  MS03 Data Assimilation and Predictive Models in the Earth Sciences
  
  Organizer: Christian Boehm (ETH Zurich, Switzerland)
  
  Co-Organizers: Andreas Fichtner (ETH Zurich, Switzerland); Dave A. May (ETH Zurich, Switzerland)
  
  Nearly all fields in the Earth sciences combine numerical models and data measurements to infer unknown parameters or to predict the future state of a dynamical system. With the widespread availability of massively parallel supercomputers researchers can continually increase both the resolution and the physical complexity within their numerical models. By combining observations and model simulations, data assimilation offers a powerful tool to iteratively solve these systems, which may be highly nonlinear with extremely large numbers of unknowns. This minisymposium aims at discussing examples from various Earth science applications, including geodynamics, geomagnetism and seismic imaging.
• Borazjani Iman MS Presentation
  Monday, June 1, 2015
  HG E21, 14:00-14:30
  
  MS Presentation
  
  Fluid-Structure Interaction Simulations of Heart Valves on Parallel Computing Clusters, Iman Borazjani (University at Buffalo SUNY, USA)
  
  Co-Authors: Hafez Asgharzadeh (University at Buffalo, USA); Mohammadali Hedayat (University at Buffalo, USA)
  
  We will discuss our new developments for simulating mechanical heart valves in left-ventricle geometries. Specifically, we discuss the development of a Newton-Krylov solver with an analytical jacobian, which has improved the performance of our flow solver on curvilinear grids by several folds, with parallel efficiency of 80-90%. Furthermore, we discuss advancements in parallelized implementation of platelet activation models into numerical simulations to quantify the probability of blood clot formation in heart valves. We present the results of our recent fluid-structure interaction simulations, and discuss the flow physics. Finally, we discuss our future plans to advance FSI simulations.
• Braun Lukas Poster
  
  Poster
  
  LS-07 Investigation of the Effect of Ca2+ binding on Tissue Transglutaminase, Lukas Braun (ETH Zurich, Switzerland)
  
  Co-Authors: Florian Herzog(ETH Zurich, Switzerland); Viola Vogel (ETH Zurich, Switzerland)
  
  Tissue transglutaminase (tTG) is a Ca2+ activated cross-linking enzyme that is involved in apoptosis, wound healing and extracellular-matrix remodeling. To date no crystal structure of the active state is published. Molecular dynamics simulations based on the structure of the inactive form were performed in order to determine the native mode of calcium binding and its impact on the enzymes conformation. The study was based on five putative binding sites (S1 to S5) characterized by Kiraly et al. MD simulations revealed the opening of a potential substrate channel near the active site, as well as a mechanism that may explain the experimentally observed cooperativity between two sites.
• Breuer Alexander MS Presentation
  Tuesday, June 2, 2015
  HG E1.2, 12:00-12:30
  
  MS Presentation
  
  Minimizing Time- and Energy-to-Solution of SeisSol, Alexander Breuer (Technische Universität München, Germany)
  
  Co-Authors: Michael Bader (Technische Universität München, Germany); Alexander Heinecke (Intel Corporation, USA)
  
  This talk gives an overview of SeisSol's performance. SeisSol solves the elastic wave equations with a ADER-DG-FEM. Together with hardware-aware optimizations on three generations of CPUs and the Xeon Phi coprocessor, the talk covers a broad range of characteristics and strategies common to high order DG-FEMs. At single-node level we present a detailed study including the cross-architecture influence of the convergence order, clock frequency and vector instruction set on time- and energy-to-solution. The second part shows our completely redesigned communication scheme aiming at overlapping communication and computation on homogeneous supercomputers in a very natural way.

C

• Cabezón Rubén Contributed Talk
  Wednesday, June 3, 2015
  HG E3, 11:10-11:30
  
  Contributed Talk
  
  Upgrading Smoothed Particle Hydrodynamics Calculations, Rubén Cabezón (University of Basel, Switzerland)
  
  Co-Authors: Domingo García-Senz (Universitat Politècnica de Catalunya, Spain)
  
  In the last years there has been a change of paradigm within the SPH calculations. What was considered as 'standard' in SPH has been renewed by a set of recent formalisms that have helped to overcome various long-lasting inherent problems of this numerical technique. These new additions include, among others, the use of new kernels resistant against pairing instability, and more accurate gradient calculation via an integral approach. In this talk I will focus in how these simple changes can substantially improve our SPH calculations at almost no cost.
• Caimo Alberto Poster
  
  Poster
  
  CSM-12 Snowball Sampling for Modeling Large Networks, Alberto Caimo (Università della Svizzera italiana, Switzerland)
  
  Co-Authors: Alessandro Lomi (Università della Svizzera italiana, Switzerland); Rolf Krause (Università della Svizzera italiana, Switzerland); Garry Robins (Melbourne School of Psychological Sciences, Australia); Alex Stivala (Melbourne School of Psychological Sciences, Australia); Johan Koskinen (University of Manchester, United Kingdom); David Rolls (Melbourne School of Psychological Sciences, Australia); Peng Wang, Swinburne University of Technology, Noshir Contractor (Northwestern University, USA)
  
  The exponential random graph model (ERGM) is a statistical model for analyzing social networks. However, estimating ERGM parameters is a computationally intensive procedure that imposes severe limits on the size of networks that can be fitted. Recently, it has been shown that conditional estimation can be used to estimate ERGM parameters by estimating parameters for smaller conditionally independent subsets of the network. Snowball sampling can be used to generate such subsets. A large number of relatively small samples can be estimated in parallel, taking advantage of parallel computing to allow estimation of much larger networks than previously possible.
• Callegari Simone Poster
  
  Poster
  
  LS-05 HPC-ABGEM: Simulating Population Dynamics and Genetics of the Worldwide Human Dispersal, Simone Callegari (University of Zurich, Switzerland)
  
  Co-Authors: John D. Weissmann (University of Zurich, Switzerland); Natalie Tkachenko (University of Zurich, Switzerland); G. Lake (University of Zurich, Switzerland); Ch. P. E. Zollikofer (University of Zurich, Switzerland)
  
  Paleoanthropological and genetic evidence point to an African origin and subsequent worldwide dispersal of Homo sapiens, who interbred during this range expansion with other human populations (e.g. Neanderthals). Understanding this phenomenon requires modelling the underlying population dynamics and genetics as well as the influence of geography and climate. Within the PASC project HPC-ABGEM, we developed a high-performance agent-based simulation framework able to tackle such issues. We will present simulations of this large-scale process, reconstruct the genetic history of our agents, as well as discuss some properties of discrete-time stochastic individual-based models.
• Calvo Flavio Poster
  
  Poster
  
  PHY-01 3-D Radiation Magnetohydrodynamic Simulations of the Near Surface Layers of the Sun, Flavio Calvo (University of Geneva, Switzerland)
  
  Co-Authors: Oskar Steiner (Kiepenheuer-Institut für Sonnenphysik, Germany); Bernd Freytag (Uppsala University, Sweden)
  
  We carry out numerical radiation magnetohydrodynamics simulations of the near surface layers of the Sun using the facilities at CSCS. The simulations reproduce the well known granular structure of the solar surface with excellent fidelity. Simulations with magnetic fields also reproduce magnetic bright points in the intergranular space as is observed on the Sun. Simulations without magnetic fields show tinier non-magnetic bright points, which have not been observationally detected so far but our simulations predict their existence and basic physical properties of them. The simulated model atmospheres also serve for the computation of synthetic polarimetric intensity maps.
• Canning Andrew Poster
  
  Poster
  
  MAT-01 A Hybrid OpenMP/MPI Solver for First-principles Plane Wave Materials Science Codes, Andrew Canning (Lawrence Berkeley National Laboratory, USA)
  
  Co-Authors:
  
  First-principles materials science codes based on density functional theory (DFT) and using plane waves (PW) have become the largest user (by method) of computer cycles at scientific computer centers around the world. We present a hybrid OpenMP/MPI Conjugate Gradient based iterative eigensolver that allows this approach to scale to tens of thousands of cores on modern many core parallel computers. Performance results will be presented for the Cray XE6 and XC30 architectures.
• Castelli Ivano E. Contributed Talk
  Wednesday, June 3, 2015
  HG E1.2, 10:30-10:50
  
  Contributed Talk
  
  High-Throughput Computational Screening of Rare-Earth Perovskite Oxides for Light Harvesting Applications, Ivano E. Castelli (EPFL, Switzerland)
  
  Co-Authors: Nicola Marzari (EPFL, Switzerland)
  
  Many codes and pseudopotentials (PP) are available and attempts to evaluate errors inside simulations have been made. We compare equations of state (EOS) using various libraries of PP inside the DFT code QUANTUM Espresso with all-electron data. A PP is considered good when the squared deviation of the two EOS is small. Using these values, together with the energy cutoffs required to converge phonons, heats of formation, and computational costs, we propose two libraries, called Standard Solid State Pseudopotentials (SSSP), with focus on efficiency and accuracy. We use the SSSP and a screening technique based on stability and bandgap to search for novel light harvesting rare-earth perovskites.
• Ceriotti Michele MS Presentation
  Monday, June 1, 2015
  HG F1, 13:30-14:00
  
  MS Presentation
  
  Mapping the Structure of Complex Materials Using Sketch-Map, Michele Ceriotti (EPFL, Switzerland)
  
  Co-Authors:
  
  A crucial step in understanding the behavior of complex materials is the introduction of coarse-grained order parameters that describe meta-stable states and transitions on the free energy surface, with a simpler, more intuitive representation than using the Cartesian coordinates of all the atoms. I will present sketch-map, a non-linear dimensionality reduction algorithm loosely based on multidimensional scaling, which obtains automatically such a simplified description, mapping with exquisite detail the stability of different conformations. Applications ranging from polypeptides to small clusters will be presented, demonstrating the potential and transferability of the method.
• Chakroun Imen Poster
  
  Poster
  
  CSM-13 Stencil-Based Exascale Simulations Using an N-Dimensional Array Toolkit, Imen Chakroun (IMEC, Belgium)
  
  Co-Authors: Imen Chakroun (IMEC, Belgium); Zubair Wadood Bhatti (Katholieke Universiteit Leuven, Belgium); Tom Vander Aa (IMEC, Belgium); Roel Wuyts (Katholieke Universiteit Leuven, Belgium/ IMEC, Belgium); Wolfgang Demeuter (Vrije Universiteit Brussel, Belgium)
  
  The goal of a scientist running simulations on top of exascale systems is to take advantage of the latest HPC technologies which are likely to guarantee better performances. Yet, its goals are to solve scientific problems rather than associated software engineering tasks. We introduce ExaShark, a large-scale n-dimensional array toolkit offered as a high-level library for scientists to help computing exascale simulations. It particularly integrates Patus for preprocessing applications that need stencil computations. While ExaShark handles the distributions of the grid using a plethora of back-end communication protocols, Patus computes optimized stencils on inner regions of the grid.
• Charles Joseph Poster
  
  Poster
  
  CLI-02 Computational and Energy Efficiency Optimizations of the Air Quality Prediction Model COSMO-ART, Joseph Charles (CSCS / ETH Zurich, Switzerland)
  
  Co-Authors: William Sawyer (CSCS / ETH, Switzerland); Oliver Fuhrer (MeteoSwiss, Switzerland); Heike Vogel (Karlsruhe Institute of Technology, Germany); Bernhard Vogel (Karlsruhe Institute of Technology, Germany); Teresa Beck (University of Heidelberg, Germany); John Linford (ParaTools, USA)
  
  COSMO-ART is a regional atmospheric GCM consisting of the COSMO forecast model coupled with the chemical transport model ART (Aerosols and Reactive Trace gases). We present several numerical approaches investigated within the FP7-funded Exa2Green project to optimize the energy footprint and performance of the model system. This effort utilizes evolving COSMO optimizations within the PASC initiative. Furthermore, algorithmic changes demonstrated in the PRACE 2IP WP8 and an accelerated version of the Kinetics PreProcessor are used to improve the integration of chemical kinetics and thus lower energy requirements and computational cost of the gas phase chemistry model.
• Charrier Dominic Etienne Poster
  
  Poster
  
  EAR-03 Discontinuous Galerkin Methods for Variable Viscosity Stokes, Dominic Etienne Charrier (ETH Zurich, Switzerland)
  
  Co-Authors: Sascha M. Schnepp (ETH Zurich, Switzerland); Dave A. May (ETH Zurich, Switzerland)
  
  We rigorously evaluate the applicability of two Interior Penalty Discontinuous Galerkin methods, namely the Nonsymmetric and Symmetric Interior Penalty methods, to variable viscosity Stokes problems. Discontinuous Galerkin (DG) methods are less restrictive regarding the construction of inf-sup stable discretizations of velocity and pressure. In addition, the respective inf-sup constants are insensitive to the element aspect ratio. To date, DG methods have not been extensively used for the considered geodynamic applications. Our numerical tests examine the stability, accuracy and robustness of the methods for cases with both smooth and discontinuous coefficients.
• Chen Jacqueline Invited Presentation
  Wednesday, June 3, 2015
  HG F30, 09:00-10:00
  
  Invited Presentation
  
  IP5 Towards Exascale Simulation of Turbulent Combustion, Jacqueline Chen (Sandia National Laboratories, USA)
  
  Chair: Gretar Tryggvason (University of Notre Dame, USA)
  
  Abstract
  Exascale computing will enable combustion simulations in parameter regimes relevant to next-generation combustors burning alternative fuels. The first principles direct numerical simulations (DNS) are needed to provide the underlying science base required to develop vastly more accurate predictive combustion models used ultimately to design fuel efficient, clean burning vehicles, planes, and power plants for electricity generation. However, making the transition to exascale poses a number of algorithmic, software and technological challenges. As Moore's Law and Dennard scaling come to an end exascale computing will be achieved only through massive concurrency. Addressing issues of data movement, power consumption, memory capacity, interconnection bandwidth, programmability, and scaling through combustion co-design are critical to ensure that future combustion simulations can take advantage of emerging computer architectures. Co-design refers to a computer system design process where combustion science requirements influence architecture design and constraints inform the formulation and design of algorithms and software. The current state of petascale DNS of turbulent combustion will be reviewed followed by a discussion of current combustion exascale combustion co-design topics: 1) architectural modeling and simulation of the behavior of combustion applications on future extreme architectures; and 2) programming model and runtime for heterogeneous, hierarchical machines with inherent variability. While bulk synchronous programming and data parallelism have been operative at the petascale, the movement to exascale requires a shift towards asynchronous programming, where to extract maximum parallelism, both data and task parallelism accessing disjoint sets of fields is required. An example from a recent refactorization of a combustion DNS code, S3D, using an asynchronous model, Legion, with dynamic runtime analysis at scale will be presented.
  
  Biography
  Jacqueline H. Chen is a Distinguished Member of Technical Staff at the Combustion Research Facility at Sandia National Laboratories. She has contributed broadly to research in petascale direct numerical simulations (DNS) of turbulent combustion focusing on fundamental turbulence-chemistry interactions. In collaboration with computer scientists and applied mathematicians she is the founding Director of the Center for Exascale Simulation of Combustion in Turbulence (ExaCT). She received several DOE INCITE Awards in the last few years and the Asian American Engineer of the Year Award in 2009.
• Cheng Bingqing Poster
  
  Poster
  
  MAT-10 Direct Path Integral Estimators for Isotope Fractionation Ratios, Bingqing Cheng (EPFL, Switzerland)
  
  Co-Authors: Michele Ceriotti (EPFL, Switzerland);
  
  We demonstrate that it is possible to evaluate the isotope fractionation ratios directly from path-integral molecular dynamics simulations. The corresponding estimators are derived from the ratio of the partition functions of the isotope substituted systems, which is obtained by a virtual substitution of the isotopes of the tagged atom. Since it avoids the mass integration altogether it is more convenient, computationally advantageous, and immune to errors in the integral for the isotope mass. Finally, we demonstrate the efficiency of these estimators by applying them to investigate the isotope fractionation ratios in the gas-phase Zundel cation, and in a few simple hydrocarbons.
• Chopard Bastien Poster
  
  Poster
  
  LS-13 Optimal Deployment of Multiscale Applications on a HPC Infrastructure, Bastien Chopard (Universiy of Geneva, Switzerland)
  
  Co-Authors: C. Bonadonna (University of Geneva, Switzerland); P. Albuquerque(University of Applied Sciences Western, Switzerland); J.-L. Falcone (University of Geneva, Switzerland); M. Ben Belgacem (University of Geneva, Switzerland); P. Künzli (University of Geneva, Switzerland); F. Brogi (University of Geneva, Switzerland); E. Rossi (University of Geneva, Switzerland); M. Bagheri (University of Geneva, Switzerland)
  
  The goal of our project is to deploy a general framework for multiscale, multi-science applications on a distributed HPC resource. The framework, defined and tested within the European project MAPPER, includes tools and provides a methodology to design, implement and run a multiscale model. We are currently deploying this framework on the Swiss HPC infrastructure, and augmenting it with a performance prediction tool which will allow an optimal placement on the distributed HPC resource. As a test application, we are developing a model for the transport of volcanic ashes that combines models for plume dynamics, advection-diffusion-sedimentation transport and aggregation of fine airborne particles. All these phenomena are produced by different physical processes and take place at different spatial and temporal scales. The corresponding numerical codes have different requirements, some needing HPC resources, but are all tightly coupled.
• Christen Patrik Poster
  
  Poster
  
  LS-06 Image-Based Computation of in Vivo Bone Loading History in Patients, Patrik Christen (ETH Zurich, Switzerland)
  
  Co-Authors: Ralph Müller (ETH Zurich, Switzerland)
  
  We have recently developed and implemented a novel image-based computational method to derive a bones loading history from its microstructure as assessed by high-resolution computed tomography. Bone strives for uniform tissue loading, thus the loading history is found by scaling a set of physiologically realistic forces until their summed resulting tissue loading is most homogeneous. Using micro-finite element analysis to calculate tissue loading, this approach provided estimates in agreement with in vivo force measurements in mouse, dog, and human bone. Since clinical image resolution is limited in patients, we here present our new method and its resolution dependency and reproducibility.
• Cisneros G. Andres MS Presentation
  Monday, June 1, 2015
  HG E3, 14:00-14:30
  
  MS Presentation
  
  Extension and Implementation of Ewald-Based Methods for Classical Simulations of a Density-Based Force Field, G. Andres Cisneros (Wayne State University, USA)
  
  Co-Authors: Robert E. Duke (Wayne State University, USA); Oleg Starovoytov (University of Houston, USA); Jean-Philip Piquemal (Pierre-and-Marie-Curie University, France)
  
  We have developed a novel polarizable force field, called the Gaussian Electrostatic Model (GEM). GEM is designed to reproduce each term of the QM intermolecular interaction by fitting electronic densities using Hermite Gaussians. GEM enables the evaluation of intermolecular interactions for molecular systems with errors below chemical accuracy for each component, and provides a novel procedure to obtain distributed multipoles (GEM--DM). The use of Hermite Gaussian result in the need to evaluate a large number of two center (Coulomb and overlap) integrals. I will discuss the extension of two approximate Ewald methods (PME and FFP) to evaluate these integrals and timings for H2O.
• Coltice Nicolas MS Presentation
  Monday, June 1, 2015
  HG F3, 16:00-16:30
  
  MS Presentation
  
  Towards Joint Reconstruction of Paleo Mantle-Lithosphere Dynamics Using Data Assimilation, Nicolas Coltice (Université Lyon 1, France)
  
  Co-Authors:
  
  Data assimilation becomes suitable when the observations are sufficient in number and accurate enough, and if the forward model contains sufficient complexity to necessary for predictions. When it comes to reconstructing convection in the Earth's mantle, data assimilation raises questions for both observations and models. I will review inverse methods used so far, for which seismic tomography is central, and present another angle for data assimilation: using tectonic data as the data to match. I will describe preliminaries and present a sequential data assimilation scheme, and a variational data assimilation scheme, both using tectonic information as the data to match.
• Commercon Benoit MS Presentation
  Tuesday, June 2, 2015
  HG E3, 10:30-11:00
  
  MS Presentation
  
  Multifrequency Radiation Hydrodynamics on Adaptive-mesh-refinement using the Flux-limited Diffusion Approximation, Benoit Commercon (CNRS, France)
  
  Co-Authors: M. Gonzalez (ENS de Lyon, France); N. Vaytet (ENS de Lyon, France); J. Masson (ENS de Lyon, France); V. Debout (Université Paris-Diderot, France); R. Teyssier (University of Zurich, Switzerland)
  
  I will present our last developments to model multifrequency radiation hydrodynamics within the adaptive-mesh-refinement code RAMSES (Teyssier 2002). We employ the flux-limited-diffusion approximation and use a combination of explicit (for hydrodynamics) and implicit (for radiation transport) schemes, as well as an efficient scheme for adaptive-time-stepping. I will present our numerical implementation and the tests we have performed. Last, I will present application to star formation problems (protostellar collapse).
• Croce Roberto Contributed Talk
  Wednesday, June 3, 2015
  HG E22, 10:50-11:10
  
  Contributed Talk
  
  Modelling and Simulation of Branching Morphogenesis, Roberto Croce (ETH Zurich, Switzerland)
  
  Co-Authors: Dzianis Menshykau (ETH Zurich, Switzerland); Tamas Kurics (ETH Zurich, Switzerland); Dagmar Iber (ETH Zurich, Switzerland)
  
  Organogenesis is a dynamic, self-organizing process. Many of the individual regulatory components, e.g. signalling molecules and their regulatory interactions, have been identified in experiments. However, an integrative mechanistic understanding of the regulatory processes is missing. Computational modelling provides a formalism to formulate and test hypotheses. We present and discuss an according Turing model for the simulation of branching morphogenesis and focus on its numerical challenges. Finally, we show a quantitative comparison between simulated- and real branching morphogenesis in vitro, acquired at the D-BSSE Single Cell Facility.

D

• De Lorenzi Michele Contributed Talk
  
  , -
  
  Contributed Talk
  
  titolo abstract, Michele De Lorenzi (CSCS / ETH Zurich, Switzerland)
  
  Co-Authors: blabla
  
  abastract text
• De Vita Alessandro MS Presentation
  Monday, June 1, 2015
  HG F1, 13:00-13:30
  
  MS Presentation
  
  On-The-Fly Machine Learning of QM Forces for Big Data-Augmented Molecular Dynamics, Alessandro De Vita (King's College London, United Kingdom)
  
  Co-Authors: Marco Caccin (King's College London, United Kingdom); Zhenwei Li (University of Basel, Switzerland); James Kermode (University of Warwick, United Kingdom)
  
  I will present a 'big-data'-based MD technique where QM-accurate information is either located in massive databases or generated on the fly if unavailable, and used to predict atomic forces via Bayesian inference [1]. The approach is efficient as it concentrates QM subroutine calls where/when 'chemically novel' configurations are encountered along the system's trajectory. Interestingly, QM-zone partitioning followed by execution by standard O(N3) QM engines is predicted to be a better option than using O(N) QM methods when dealing with large QM zones in machine learning-augmented QM/MM calculations running on HPC platforms. [1] Z. Li, J. R. Kermode and A. De Vita, PRL 114, 096405 (2015).
• Decaix Jean MS Presentation
  Monday, June 1, 2015
  HG E22, 13:30-14:00
  
  MS Presentation
  
  CFD Computations of a Cavitation Vortex Rope, Jean Decaix (HES-SO//Valais-Wallis, Switzerland)
  
  Co-Authors: Sebastien Alligné (Power Vision Engineering Sàrl, Switzerland); Andres Müller (EPFL, Switzerland); Cécile Münch (HES-SO//Valais-Wallis, Switzerland); François Avellan (EPFL, Switzerland)
  
  Due to the increase in renewable energies in the electricity market, hydraulic power plants have to run at off-design operating points. For Francis turbines, this leads to the formation of a vortex rope known to promote cavitation surge. If the flow discharge is higher than the nominal one, an axisymmetric cavitation vortex rope develops in the draft tube cone. Such a vortex rope leads to high pressure fluctuations that interact with the entire power plant and limit the range of operating points. To better understand the physics of the phenomenon and to provide input to 1D models, two-phase 3D CFD calculations are performed and compared with experimental data
• Deparis Simone MS Presentation
  Monday, June 1, 2015
  HG E21, 13:00-13:30
  
  MS Presentation
  
  Numerical Modeling of the Fluid Dynamics in the Heart, Simone Deparis (EPFL, Switzerland)
  
  Co-Authors: L. Dede (EPFL, Switzerland); E. Faggiano (Univeristà di Pavia, Italy); M. Fedele (Univeristà di Pavia, Italy); A. Gerbi (EPFL, Switzerland); A. Quarteroni (EPFL, Switzerland); A. Tagliabue (Politecnico di Milano, Italy)
  
  The challenge of the modeling and simulation of blood flow dynamics in the heart and in the aortic root includes at least the following components: large deformations of the computational domain, opening and closing of the valves, interaction with the cardiac muscle, and changing flow regime. In this talk we focus mainly on the flow components. We first look at the flow in an idealized left ventricle. The flow is discretized by finite elements and stabilized by the variational multiscale method. The valve is modeled by a zero-dimensional resistance model. We then present a three-dimensional model for the leaflets. We conclude by showing simulations of blood flow in the aortic root.
• Diaz Julien MS Presentation
  Tuesday, June 2, 2015
  HG E1.2, 13:30-14:00
  
  MS Presentation
  
  Discontinuous Galerkin Approximations for Seismic Wave Propagation in a HPC Framework, Julien Diaz (INRIA, France)
  
  Co-Authors: Hélène Barucq (INRIA, France); Lionel Boillot (INRIA, France); Marie Bonnasse-Gahot (INRIA, France); Henri Calandra (Total, France); Stéphane Lanteri (INRIA, France)
  
  The accurate representation of the heterogeneities of the subsurface requires using high order numerical methods based on unstructured meshes. Discontinuous Galerkin Methods (DGM) are becoming very popular in the geophysicist community, but they still have to be improved to tackle realistic 3D configurations. Issues are different regarding the equation regime. In time domain, stability depends on a penalization parameter and efficiency of the code requires task-based programming to take advantage of heterogeneous architecture. In harmonic domain, the size of the global DG linear system is huge and Hybridizable DG methods are a promising framework to reach reasonable dimensions.
• Dib Linda MS Presentation
  Tuesday, June 2, 2015
  HG E21, 11:30-12:00
  
  MS Presentation
  
  Bayesian Model in Medical Sciences, Linda Dib (University of Lausanne, Switzerland)
  
  Co-Authors:
  
  Coevolving sites play a critical role the structure and function of a protein. I have recently developed two methods for the detection of such signature. The first method is a combinatorial and compares the homologous sequences whereas the other method is probabilistic and looks for the evolution of amino acids along the specie tree. The latter was implemented in a ML and Bayesian framework. I applied both methods on a large set of proteins. Among these proteins, I took particular care to analyze the Amyloid B protein that is implied in Alzheimer disease. The method revealed that well-known fragments of the protein are co-evolving revealing an evolutionary signature in this protein.
  Minisymposium
  Minisymposium
  
  MS12 Bayesian Life Science Models
  
  Organizer: Linda Dib (University of Lausanne, Switzerland)
  
  Co-Organizers: Xavier Meyer (University of Lausanne, Switzerland; Swiss Institute of Bioinformatics, Switzerland; University of Geneva, Switzerland)
  
  The development of Bayesian methods in health sciences and biology has increased tremendously over the past decades since they learn about the data by exploring different hypothesis based on a priori knowledge. These models are computationally demanding and may require several million of iterations to reach convergence. Very few algorithms have explored the possibilities of using HPC approaches to speed up Bayesian methods. In this meeting we aim to provide an opportunity to bring together world experts in the fields of Bayesian statistics and HPC, to discuss the potential of new methodologies, and to investigate how to build Bayesian models that meet HPC platform requirements.
• Doetsch Joseph MS Presentation
  Monday, June 1, 2015
  HG E22, 16:30-17:00
  
  MS Presentation
  
  Modeling Induced Seismic Hazard During Geothermal Reservoir Creation, Joseph Doetsch (ETH Zurich, Switzerland)
  
  Co-Authors: Valentin Gischig (ETH Zurich, Switzerland); Dimitrios Karvounis (ETH Zurich, Switzerland); Stefan Wiemer (ETH Zurich, Switzerland)
  
  To enable the large-scale exploitation of deep geothermal energy for electricity generation in Switzerland, solutions must be found for two fundamental and coupled problems: (1) How to create an efficient heat exchanger in the hot underground that can produce energy for decades and (2) at the same time keeping the risk posed by induced earthquakes to acceptable levels' Numerical simulations of different complexity and physical detail are currently being developed for an Advanced Traffic Light System (ATLS) to predict the risk of in induces earthquakes in real-time. This contribution presents developments of the induced seismicity prediction codes towards high performance computing.
• Dolfi Michele Poster
  
  Poster
  
  MAT-12 Large Scale Density Matrix Renormalization Group Calculations, Michele Dolfi (ETH Zurich, Switzerland)
  
  Co-Authors: Adrian Kantian (KTH Royal Institute of Technology / Stockholm University, Sweden); Alexandr Kosenkov (ETH Zurich, Switzerland); Bela Bauer (Microsoft Research, USA); Thierry Giamarchi (University of Geneva, Switzerland); Matthias Troyer (ETH Zurich, Switzerland)
  
  The Density Matrix Renormalization Group (DMRG) is a very efficient method to study one-dimensional quantum systems. In higher dimensions, particularly in 2d, this technique requires to operate on matrices with a bond dimension 'm' scaling exponentially in the system width. We present the techniques used in our massively parallel DMRG application which allow the study of challenging systems up to m=30'000. Our test cases is the Hubbard models in weakly coupled chains and ladders. We apply the DMRG to this models to investigate theoretical models for superconductivity in cuprate materials and organic super conductors by studying the pairing correlations.
• Donadio Davide MS Presentation
  Monday, June 1, 2015
  HG F1, 15:30-16:00
  
  MS Presentation
  
  Machine Learning Approaches to Simulate Nanoscale Heat Transport, Davide Donadio (Ikerbasque, Spain)
  
  Co-Authors: Claudia Mangold (ETH Zurich, Switzerland); Emanuele Bosoni (University of Milano Bicocca, Italy); Davide Campi (University of Milano Bicocca, Italy); Joerg Behler (Ruhr University Bochum, Germany); Marco Bernasconi (University of Milano Bicocca, Italy)
  
  Simulating thermal transport in nanostructures at atomistic level often requires the use of large-scale models, thus making it necessary to recur to empirical potentials. However, systems with complex chemistry, such as materials for phase change memories (PCM) and advanced thermoelectrics (TE), are poorly described by standard potentials. Here we propose an alternative approach to compute phonon properties and the thermal conductivity of nanostructures, which makes use of neural network potentials fitted on a large database of density functional theory calculation. We demonstrate its application on GeTe nanostructures for PCM and on nanostructured Ge-Mn alloys for TE application.
• Donfack Simplice MS Presentation
  Tuesday, June 2, 2015
  HG E21, 14:30-15:00
  
  MS Presentation
  
  Improving the Applicability of Highly Efficient Compilers to a Wider Class of Problems, Simplice Donfack (Università della Svizzera italiana, Switzerland)
  
  Co-Authors: Olaf Schenk (Università della Svizzera italiana, Switzerland); Bram Reps (University of Antwerpen, Belgium); Wim Vanroose (University of Antwerpen, Belgium); Drosos Kourounis (Università della Svizzera italiana, Switzerland)
  
  We propose an approach that allows current solvers to adapt to future architectures and continue to scale at exascale, and this by removing the bottleneck introduced in the communication process. Our objective is to increase the arithmetic intensity, that is the number of floating-point operations performed per bytes fetched in the memory during the execution of the solvers, indeed reduce the number and the volume of the data exchanged among the processors and the memory. We use the help of efficient stencil compilers such as PLUTO and PATUS to increase the arithmetic intensity of these solvers.
• Draxl Claudia MS Presentation
  Monday, June 1, 2015
  HG F1, 16:30-17:00
  
  MS Presentation
  
  From Sharing to Big-Data Analytics: The NoMaD Project, Claudia Draxl (Humboldt-Universität zu Berlin, Germany)
  
  Co-Authors: Luca M. Ghiringhelli (Fritz-Haber-Institut der Max-Planck-Gesellschaft, Germany); Jan Vybiral (Charles University, Czech Republic); Sergey V. Levchenko (Fritz-Haber-Institut der Max-Planck-Gesellschaft, Germany); Matthias Scheffler (Fritz-Haber-Institut der Max-Planck-Gesellschaft, Germany)
  
  Essentially every new commercial product depends on new or improved materials. To identify such materials, computational approaches are increasingly employed, with millions of CPU hours spent every day and an enormous amount of data created. The goal of the Novel Materials Discovery (NoMaD) Project is not boosting more high-throughput calculations. The issue of NoMaD is: What to do with the data? In this talk, I will introduce the NoMaD Repository (https://nomad-repository.eu) that was established to host, organize, and share materials data, and I will demonstrate an example how big-data analytics based on statistical-learning approaches can lead to new insight into materials properties.
• Drescher Lukas Poster
  
  Poster
  
  MAT-03 A Spherical P-method for Full-potential Electronic Structure Problems, Lukas Drescher (Technische Universität Berlin, Germany)
  
  Co-Authors: Reinhold Schneider (Technische Universität Berlin, Germany); Kersten Schmidt (Technische Universität Berlin, Germany)
  
  A major challenge in full-potential electronic structure calculations stems from singularities due to the nuclear Coulomb potential. An hp-adaptive FEM was recently shown to yield exponential convergence at the cost of a large number of degrees of freedom (NDOF) that accumulate at the nuclei. In this work, we propose a spherical discretization near the nuclei that is adapted to the core singularity. By coupling this approach to a FEM on the complementary domain in the mortar element framework, we obtain a pure p-method that reduces the NDOF at the nuclei substantially. The associated efficiency gain versus hp-FEM is shown on a numerical example from molecular Kohn-Sham DFT.
• Dubey Anshu Minisymposium
  Minisymposium
  
  MS09 Abstractions Interplay in Domain Frameworks to Tackle Heterogeneity
  
  Organizer: Anshu Dubey (Lawrence Berkeley National Laboratory, USA)
  
  Co-Organizers: Brian Van Straalen (Lawrence Berkeley National Laboratory, USA)
  
  Current generation of domain frameworks have componentization and composability implemented in a stable and portable programming model as their primary abstractions. As platforms began to become heterogeneous some domain frameworks added abstractions to tackle the performance-portability trade-offs, for which the current best practices are no longer adequate. However, there is a growing separation between the shortening life-cycles for future programming models and the life-cycle of a community software framework. This minisymposium seeks to collect and document the abstraction-lifting efforts of multiphysics frameworks with a hope of starting the discussion for closing this gap.

E

• El Mellouhi Fadwa Poster
  
  Poster
  
  MAT-19 The Computational Design of Pb Free and Stable Hybrid Materials for Solar Cells, Fadwa El Mellouhi (Qatar Environment and Energy Research Institute, Qatar)
  
  Co-Authors: Fadwa El Mellouhi (Qatar Foundation, Qatar); Akinlolu Akande (Trinity College Dublin, Ireland); Sergey Rashkeev (Qatar Foundation, Qatar); Carlo Motta (Trinity College Dublin, Ireland); Stefano Sanvito (Trinity College Dublin, Ireland); Sabre Kais (Qatar Foundation, Qatar); Fahhad Alharbi (Qatar Foundation, Qatar)
  
  In the past four years, the solar cell field has experienced an unprecedented meteoritic emergence of a new family of solar cell technologies; namely, perovskites solar cells (PSC) using (CH3NH3)PbI3 as absorber. However, two main challenges prevent deploying PSC technology: the presence of the toxic element lead (Pb) and their structural instability. Our efforts to design lead free family of hybrid materials demonstrate that the design of hybrid materials containing organic cations might require careful considerations among them the size of the cell used during the screening process. Our strategy will be presented as well as some resulting promising compounds.

F

• Fahlke Jorrit MS Presentation
  Tuesday, June 2, 2015
  HG F3, 14:00-14:30
  
  MS Presentation
  
  EXA-DUNE - Flexible PDE Solvers, Numerical Methods and Applications, Jorrit Fahlke (University of Münster, Germany)
  
  Co-Authors: Peter Bastian (University of Heidelberg, Germany); Christian Engwer (University of Münster, Germany); Dominik Göddeke (University of Stuttgart, Germany); Steffen Müthing (University of Heidelberg, Germany); Dirk Ribbrock (Technische Universität Dortmund, Germany)
  
  DUNE is a software framework for the flexible construction of solvers for PDEs. It has been used successfully on up to O(10^6) cores using MPI. In the EXA-DUNE project we make DUNE suitable for EXA-scale computing. This includes using shared-memory parallelism, wide SIMD units, and accelerator devices while requiring minimal changes in the way users of the framework implement their solvers. In addition, some applications require new numerical methods to overcome restrictions like limited memory and bandwidth. Solving these challenges gives us a set of methods, which we present here. They allow constructing applications that take advantage of EXA-scale computers.
• Fisicaro Giuseppe Poster
  
  Poster
  
  MAT-02 A Solver for the Generalized Poisson Equation in Wet-environments Electronic-structure Calculations, Giuseppe Fisicaro (University of Basel, Switzerland)
  
  Co-Authors: L. Genovese (CEA, France); S. Goedecker (University of Basel, Switzerland)
  
  The computational study of chemical reactions in complex wet environments is critical for applications in many fields. In the present work we propose an algorithm, based on the Preconditioned Conjugate Gradient (PCG) method, to handle the Generalized Poisson equation. It allows to solve iteratively the minimization problem with some ten iterations. The algorithm takes advantage of a preconditioning procedure based on the BigDFT Poisson solver for the standard Poisson equation. It exhibits very high accuracy and parallel efficiency for different Boundary conditions. The solver has been integrated to the BigDFT main electronic-structure package and it will be released as a independent program.
• Folini Doris Poster
  
  Poster
  
  CLI-06 Tuning a Global Climate Model - a Costly Hunt in Parameter Space, Doris Folini (ETH Zurich, Switzerland)
  
  Co-Authors: Martin Wild (ETH Zurich, Switzerland)
  
  Global climate models have to rely on parameterizations for a number of sub-grid-scale processes. Associated free parameters have to be adjusted such as to obtain a 'physically meaningful climate', a process referred to as 'model tuning'. We report on the tuning of one specific global climate model: MPI-ESM-HAM, the Max Planck Earth System Model (MPI-ESM) coupled to the Hamburg Aerosol Module (HAM). And we carve out some questions related to the fact that model tuning is, in essence, an expensive search in a parameter space, where the parameter space and the score function are not necessarily well defined 'ab initio'.
• Forti Davide Poster
  
  Poster
  
  LS-01 A Parallel Block Preconditioner for Fluid-Structure Interaction Problems in Hemodynamics, Davide Forti (EPFL, Switzerland)
  
  Co-Authors: Simone Deparis (EPFL, Switzerland); Gwenol Grandperrin (EPFL, Switzerland); Alfio Quarteroni ( EPFL, Switzerland)
  
  Modeling blood flow in large arteries is based on fluid-structure coupled problems. Because of the strong added mass effect, the continuity between kinematic and dynamic variables needs to be enforced at every time iteration, ideally using fully-implicit monolithic algorithms. To deal with the computational complexity induced by monolithic schemes, a parallel solution framework is mandatory. We present a parallel block preconditioner that exploits each of the physics involved in the FSI problem. We show the scalability of our solver implemented in LifeV applied to patient-specific femoropopliteal bypasses.
• Fournier Alexandre MS Presentation
  Monday, June 1, 2015
  HG F3, 16:30-17:00
  
  MS Presentation
  
  Data Assimilation and Predictive Models in Geomagnetism, Alexandre Fournier (Institut de Physique du Globe de Paris, France)
  
  Co-Authors: Julien Aubert (Institut de Physique du Globe de Paris, France); Lars Nerger (Alfred Wegener Institute, Germany); Sabrina Sanchez (Institut de Physique du Globe de Paris, France)
  
  Assimilating geomagnetic data in numerical models of Earth's core dynamics is a challenging problem, since the information contained in the geomagnetic record is intrinsically restricted to the large scales of the poloidal geomagnetic field at the core-mantle boundary. After a general introduction on terrestrial magnetism and the observation of Earth's magnetic field, I will report more specifically on recent efforts carried out to investigate the feasibility of resorting to three-dimensional, buoyancy-driven, numerical dynamo models for geomagnetic data assimilation practice. For this symposium, I will discuss in particular the computational cost associated with this exercise.
• Fuhrer Oliver Contributed Talk
  Wednesday, June 3, 2015
  HG E21, 10:30-10:50
  
  Contributed Talk
  
  Co-designing a System for Regional Weather and Climate Prediction, Oliver Fuhrer (MeteoSwiss, Switzerland)
  
  Co-Authors: Xavier Lapillonne (MeteoSwiss, Switzerland); Carlos Osuna (ETH Zurich, Switzerland); Andrea Arteaga (ETH Zurich, Switzerland); Stefan Ruedisuehli (ETH Zurich, Switzerland); Thomas Schulthess (CSCS / ETH Zurich, Switzerland)
  
  Adapting weather and climate models to current and emerging hardware architectures is a formidable challenge. We present the design decisions, implementation, performance results and learnings from an effort to re-design an existing, widely used community code (COSMO) to hybrid architectures. Using the concrete example of MeteoSwiss' next-generation operational forecasting system, we illustrate the gains in scientific possibilities, time-to-solution and energy-to-solution.
• Fütterling Valentin Contributed Talk
  Wednesday, June 3, 2015
  HG E1.1, 10:30-10:50
  
  Contributed Talk
  
  Core Algorithms for High-Performance, Interactive Rendering of Large-Scale Scientific Data, Valentin Fütterling (Fraunhofer, Germany)
  
  Co-Authors: Carsten Lojewski (Fraunhofer Institute for Industrial Mathematics, Germany); Franz-Josef Pfreundt (Fraunhofer Institute for Industrial Mathematics, Germany); Achim Ebert (Technische Universität Kaiserslautern, Germany)
  
  The efficient analysis of large-scale scientific data demands interactive rendering techniques unconstrained by bus bottlenecks or memory limits. This favors CPU-based solutions over accelerators (GPUs). Ray-tracing provides the ideal method, supporting a variety of primitives and naturally incorporating ambient occlusion as a view-point neutral lighting strategy. Our contributions are novel algorithms for coherent and incoherent ray-traversal that significantly outperform previous solutions. We demonstrate the efficiency of our core algorithms implemented in shared and distributed memory rendering systems with the capability of rendering billions of primitives interactively.

G

• Galli Giulia Invited Presentation
  Tuesday, June 2, 2015
  HG F30, 09:00-10:00
  
  Invited Presentation
  
  IP3 Materials Discovery and Scientific Design by Computation: a Revolution Still in the Making, Giulia Galli (University of Chicago, USA)
  
  Chair: Nicola Marzari (EPFL, Switzerland)
  
  Abstract
  The need of advanced materials for sustainable energy resources and next generation information technology requires the development of integrated scientific strategies, encompassing theoretical innovations, and computational and laboratory experiments. Substantial progress has been made in the last two decades in understanding and predicting the fundamental properties of materials and molecular systems from first principles, i.e. from numerical solutions of the basic equations of quantum mechanics. However the field of ab initio predictions is in its infancy; some formidable theoretical and computational challenges lie ahead of us, including the collection and use of data generated by simulations. We will describe recent progress and successes obtained in predicting properties of matter by quantum simulations, and discuss algorithmic challenges in connection with the use of evolving high-performance computing architectures. We will also discuss open issues related to the validation of the approximate, first principles theories used in large-scale quantum simulations.
  
  Biography
  Giulia Galli is the Liew Family Professor of Electronic Structure and Simulations in the Institute for Molecular Engineering, at the University of Chicago, Senior Scientist at Argonne National Laboratory and Senior Fellow of the Computational Institute of the University of Chicago. She holds a PhD in physics from the International School of Advanced Studies (SISSA) in Trieste, Italy. She was the head of the Quantum Simulations Group at the Lawrence Livermore National Laboratory and Professor of Chemistry and Physics at the University of California, Davis, prior to joining the University of Chicago. She is a Fellow of the American Physical Society (APS) and of the American Academy of Arts and Sciences (AAAS). She received an award of excellence from the Department of Energy (2000) and the Science and Technology Award from the Lawrence Livermore National Laboratory (2004). She served as chair of the Division of Computational Physics of the American Physical Society (2006-2007) and of the Extreme Physics and Chemistry of Carbon Directorate of the Deep Carbon Observatory (2010-2013). Her research activity is focused on quantum simulations of systems and processes relevant to condensed matter physics, physical chemistry, molecular engineering and nano-science.
• Gantner Robert Poster
  
  Poster
  
  CSM-06 Higher-Order Quasi-Monte Carlo for Bayesian Inversion of Parametric PDEs, Robert Gantner (ETH Zurich, Switzerland)
  
  Co-Authors: Christoph Schwab (ETH Zurich, Switzerland)
  
  We consider multiparametric partial differential equations to which we apply the Bayesian approach to inverse problems. A diffusion equation with distributed, parametric coefficient is considered as model problem. Applications include uncertainty quantification in groundwater flow and shape uncertainty computations. Such problems involve high- or infinite-dimensional integrals, requiring both suitable methods and their massively parallel implementation. We focus here on novel higher-order QMC methods, explaining their efficient construction and their massively parallel application to the problems at hand. Numerical examples consider a parametric space with up to $s=1024$ dimensions.
• Gao Hao MS Presentation
  Monday, June 1, 2015
  HG E21, 13:30-14:00
  
  MS Presentation
  
  Coupling in Vivo Human Mitral Valve to the Left Ventricle, Hao Gao (University of Glasgow, United Kingdom)
  
  Co-Authors: Boyce E. Griffith (University of North Carolina, USA); Colin Berry (University of Glasgow, United Kingdom); Xiaoyu Luo (University of Glasgow, United Kingdom)
  
  I will present an integrated model of mitral valve (MV) coupled with the left ventricle (LV) by using an immersed boundary method along with a finite element description of valvular and myocardial mechanics. The model is derived from clinical images and takes into account of the important valvular features, left ventricular contraction, nonlinear soft tissue mechanics, fluid structure interaction, and the MV-LV interaction. The integrated MV and LV model can simulate the cardiac function both in diastole and systole. Further work is required to ensure that the highly complex valvular-ventricular interaction, and the fluid-structure interaction, can be reliably represented.
• Gasparotto Piero Poster
  
  Poster
  
  MAT-16 Recognizing Molecular Patterns by Machine Learning: An Agnostic Structural Definition of the Hydrogen Bond, Piero Gasparotto (EPFL, Switzerland)
  
  Co-Authors: Michele Ceriotti (EPFL, Switzerland)
  
  Atomistic simulations are getting increasingly more accurate, and make it possible to model predictively increasingly complex problems. In order to rationalize their behavior, a very promising approach is the use of machine-learning techniques (e.g. pattern recognition and dimensionality reduction) to provide a fully automated framework to understand structural patterns and to accelerate the exploration of the complex conformational landscape in a simulation. Here we demonstrate the use of a Probabilistic Analysis of Molecular Motifs (PAMM) method to provide an agnostic definition of the hydrogen bond, and discuss its application to the description of ab initio water models.
• Ghiringhelli Luca Minisymposium
  Minisymposium
  
  MS02 Big Data Analytics for Novel Materials Discovery
  
  Organizer: Luca Ghiringhelli (Fritz Haber Institute of the Max Planck Society, Germany)
  
  Co-Organizers: Matthias Scheffler (Fritz-Haber-Institut der Max-Planck-Gesellschaft, Germany)
  
  To achieve deeper and novel scientific insight, to decide what new materials should be studied next as most promising novel candidates, and to identify interesting anomalies, the big data generated in computational materials science must be shared and efficient and physically meaningful analytic tools must be developed. The field has evolved in two main directions: constructing large databases of internally consistent materials data and development of advanced, machine-learning based, analysis techniques. Speakers of this symposium will discuss the strong and weak points of the different strategies adopted to perform big-data analytics, towards the discovery of new materials.
• Ghosh Sunayana MS Presentation
  Tuesday, June 2, 2015
  HG E22, 11:00-11:30
  
  MS Presentation
  
  Adaptive Inexact Spectral Deferred Correction Methods for Long-Time Integration, Sunayana Ghosh (Zuse Institute Berlin, Germany)
  
  Co-Authors: Martin Weiser (Zuse Institute Berlin, Germany)
  
  Inexact spectral deferred correction (SDC) method for solving non-stiff initial value problems (IVPs) are addressed. In many IVPs the approximate evaluation of the right hand side (rhs) is obtained much faster. This allows us to exploit the trade-off between accuracy and computational cost in evaluation of rhs. Error propagation in the SDC method is studied to obtain expressions of global errors w.r.t. local errors. Work models relating local errors to computational costs are derived. Error and work models are combined to obtain an adaptive tolerance selection by minimizing total work. Inexact SDC is then tested on long-time integration of damage evolution in oscillatory fatigue problems.
• Giefers Heiner MS Presentation
  Tuesday, June 2, 2015
  HG E1.1, 14:00-14:30
  
  MS Presentation
  
  Enabling Energy-Efficient Exascale Computing: Acceleration of HPC Kernels with Reconfigurable Hardware, Heiner Giefers (IBM Research, Switzerland)
  
  Co-Authors: Heiner Giefers (IBM Research, Switzerland); Raphael Polig (IBM Research, Switzerland); Jan van Lunteren (IBM Research, Switzerland); Christoph Hagleitner (IBM Research, Switzerland)
  
  Recent studies have shown that arising big data analytics problems cannot be performed at economically affordable energy if we rely on the traditional scaling of Von-Neumann microprocessors based on Moore's law. A drastic reduction of energy consumption requires the transition from the current computing-centric model to a more data-centric model in which data is pushed to functional units tailored to the specific workload requirements. In this talk, we present an architecture for such a near-data processing approach and show initial results that unveil the energy-efficiency of custom processing pipelines implemented in reconfigurable hardware.
• Giofrè Daniele Poster
  
  Poster
  
  MAT-17 Self-Assembly of Water-Methanol Mixtures at the Surface of Graphite, Daniele Giofrè (EPFL, Switzerland)
  
  Co-Authors: Michele Ceriotti (EPFL, Switzerland)
  
  AFM experiments observed the formation of regular, striped patterns at the interface between a methanol/water solution and a graphite surface. The time scale for the formation of these patterns is beyond direct observation by atomistic simulations. Nevertheless, molecular dynamics shows segregation of methanol at the interface, and a high degree of topological and geometric order in the H-bond network. The regular pattern can then be explained as a regular array of defects originating due to the small mismatch between the periodicity of the substrate and the characteristic length-scale of the H-bond patterns.
• Giraud Luc MS Presentation
  Tuesday, June 2, 2015
  HG E21, 14:00-14:30
  
  MS Presentation
  
  On Resilience in Krylov Subspace Solvers, Luc Giraud (INRIA, France)
  
  Co-Authors: E. Agullo (INRIA, France); L. Giraud (INRIA, France); P. Salas (University of Sherbrooke, Canada); E.F. Yetkin (INRIA, France); M. Zounon (INRIA, France)
  
  In this talk we will discuss possible numerical remedies to survive data loss in some numerical linear algebra solvers namely Krylov subspace linear solvers and some widely used eigensolvers. Assuming that a separate mechanism ensures fault detection, we propose numerical algorithms to extract relevant information from available data after a fault. After data extraction, well chosen part of missing data is regenerated through interpolation strategies to constitute meaningful inputs to numerically restart. We will also present some preliminary investigations to address soft error detection again at the application level in the conjugate gradient framework.
• Glöss Andreas Poster
  
  Poster
  
  MAT-07 DBCSR: Accelerated Sparse Matrix Multiplication Library, Andreas Glöss (ETH Zurich, Switzerland)
  
  Co-Authors: Alfio Lazzaro (ETH Zurich, Switzerland); Hans Pabst (Intel Semiconductor AG, Switzerland); Ole Schütt (ETH Zurich, Switzerland); Peter Messmer (NVIDIA, Switzerland); Joost VandeVondele (ETH Zurich, Switzerland); Jürg Hutter (University of Zurich, Switzerland)
  
  DBCSR is a sparse matrix storage, manipulation and multiplication library. It is used in the CP2K quantum chemistry program and its multi-layered structure automatically takes care of and optimizes several computational aspects like parallelism (MPI, OMP, Accelerators), data (cache) locality and on-the-fly filtering. Here we describe the challenges and our solutions for designing the DBCSR library with a special focus on the backend layer and the backends (CPU, CUDA, OpenCL and MIC). The latter, as key components for accessing all available computational resources on a modern multicore hybrid system concurrently, will be explained in more detail.
• Golze Dorothea Poster
  
  Poster
  
  MAT-13 Local Density Fitting within a Gaussian and Plane Waves Approach, Dorothea Golze (University of Zurich, Switzerland)
  
  Co-Authors: Marcella Iannuzzi (University of Zurich, Switzerland); Jürg Hutter (University of Zurich, Switzerland)
  
  A local density fitting technique (LRI) is introduced for Kohn-Sham (KS) density functional theory calculations using a mixed Gaussian and plane waves (GPW) approach within the CP2K program package. In LRI, the atomic pair densities are approximated by an expansion in one-center fit functions reducing the prefactor for building the KS matrix. Furthermore, the scalability of the grid-based terms with respect to number of CPUs can be simplified and improved. Results are presented for liquid water demonstrating that the SCF step can be sped up by a factor of at least three for large and medium-sized systems. Reaction energies obtained by LRI prove that the approach is also accurate.
• Grayver Alexander Poster
  
  Poster
  
  EAR-06 Large-Scale Geo-Electromagnetic Modeling with Adaptive High-Order FEM, Alexander Grayver (ETH Zurich, Switzerland)
  
  Co-Authors: Tzanio Kolev (Lawrence Livermore National Laboratory, USA)
  
  Electromagnetic methods of geophysics aim at studying the subsurface electrical conductivity distribution, and typically require solution of a large number of problems derived from Maxwell's equations. This study investigates the use of adaptive high-order finite elements (FE) to discretize these problems in large-scale parallel settings. We present a new scalable algorithm for solving the resulting systems, based on optimal block-diagonal and auxiliary-space preconditioning. A particular advantage of our solver is that it can handle arbitrarily high-order FE on unstructured and non-conforming locally refined meshes. The meshes are refined by using efficient goal-oriented error estimator.
• Grosser Tobias Minisymposium
  Minisymposium
  
  MS18 Domain Specific Compilation for Scientific Computing
  
  Organizer: Tobias Grosser (ETH Zurich, Switzerland)
  
  Co-Organizers: Torsten Hoefler (ETH Zurich, Switzerland)
  
  Languages and compilation strategies that exploit domain specific knowledge are well known as effective tools for the automatic generation of optimized program code. However, the increasingly complex heterogeneous hardware at the core of todays HPC platforms as well as the large base of existing HPC applications makes the use and integration of domain specific compilation techniques challenging. In this symposium, we present and discuss recent approaches that address the challenge of domain specific optimizations in the context of heterogeneous systems, scientific computing problems as well as the integration of domain specific languages with existing code bases.
• Größlinger Armin MS Presentation
  Tuesday, June 2, 2015
  HG F1, 15:00-15:30
  
  MS Presentation
  
  ExaStencils: Domain-Specific Stencil Code Generation and Optimization, Armin Größlinger (Universität Passau, Germany)
  
  Co-Authors:
  
  Project ExaStencils aims to develop a code generator which generates high performance multigrid stencil codes for a certain class of problems. Going from the continuous mathematical high-level description of the problem down to highly optimized low-level code requires to make choices which make the description more concrete and, in the end, yield a fast code. ExaStencils has defined 4 layers of domain-specific languages at different levels of abstractions to express these choices. In this talk, we give an overview of the current state of the language layers, the code generator and the use of domain knowledge in the code transformations and optimizations.
• Gurdal Yeliz Poster
  
  Poster
  
  MAT-15 Pyrphyrin Adsorption on Reconstructed and Ideal Au(111) Surface, Yeliz Gurdal (University of Zurich, Switzerland)
  
  Co-Authors: Marcella Iannuzzi (University of Zurich, Switzerland); Sandra Luber (University of Zurich, Switzerland); Jürg Hutter (University of Zurich, Switzerland)
  
  We study adsorption of Pyrphyrin on ideal and reconstructed Au(111) surfaces and compare favorable adsorption geometries of the molecule by DFT simulations. The most stable adsorption state is the one with adsorption of CN links along Au axis. The interaction to the surface is dominated by van der Waals, but the orientation and deformation of the molecule are determined by the attraction between the CN groups and the closest Au atoms. Adsorption geometries of both monomer and dimer show differences with respect to the modeled Au surface. We will investigate the electronic structure of Pyrphyrin/Au by calculating molecular energy states, N1s XPS Spectra and STM images.
• Gysi Tobias MS Presentation
  Tuesday, June 2, 2015
  HG F1, 13:30-14:00
  
  MS Presentation
  
  MODESTO: Data-Centric Analytic Optimization of Complex Stencil Programs on Heterogeneous Architectures, Tobias Gysi (ETH Zurich, Switzerland)
  
  Co-Authors: Tobias Grosser (ETH Zurich, Switzerland); Torsten Hoefler (ETH Zurich, Switzerland)
  
  An efficient implementation of stencil computations requires code transformations such as loop tiling and loop fusion. In the course of the HP2C project we developed the STELLA stencil library that goes a long way in applying efficient code transformations for different target architectures. Nevertheless, STELLA still requires some manual tuning. In this talk, we introduce MODESTO, a model-driven stencil optimization framework, that for a stencil program suggests program transformations optimized for a given target architecture. Using a compile-time performance model and mathematical optimization we successfully tune example stencils of the COSMO atmospheric model.

H

• Haber Eldad MS Presentation
  Monday, June 1, 2015
  HG E21, 16:30-17:00
  
  MS Presentation
  
  On the Solution of Large Scale Inverse Problems that "cannot be solved", Eldad Haber (University of British Columbia, Canada)
  
  Co-Authors: Lars Ruthotto (Emory University, USA)
  
  Many problems in earth science involve with the solution of inverse problems on very large domains and with multiple sources and frequency. Solving such problems was considered to be impossible until recently. In this talk we explore algorithmic and software development that enable the solution of these problems. First, we explore the use of adaptive mesh refinement coupled with the use of direct solvers. Second, we discuss the use of stochastic programming techniques and finally we discuss the parallelization of the algorithm using Julia We use the controlled source EM as a model problem and demonstrate how it can be solved using the above techniques
• Hadjidoukas Panagiotis MS Presentation
  Tuesday, June 2, 2015
  HG F30, 15:00-15:30
  
  MS Presentation
  
  Hierarchical Bayesian Models on HPC Platforms, Panagiotis Hadjidoukas (ETH Zurich, Switzerland)
  
  Co-Authors: Panagiotis Angelikopoulos (ETH Zurich, Switzerland); Steven Wu (ETH Zurich, Switzerland); Petros Koumoutsakos (ETH Zurich, Switzerland)
  
  Hierarchical Bayesian modeling provides an inference framework to fuse heterogeneous data into engineering applications. Hierarchical models suffer from extensive computational demands arising from multiple model evaluations and data intensity. We tackle both challenges using our parallel framework for uncertainty quantification. We combine state of the art parallelized sampling schemes to achieve multiple levels of nested parallelism and manage large data volumes generated by the simulations. We demonstrate our approach to the calibration of Pharmacokinetic models using heterogeneous experimental data measurements on supercomputing platforms.
• Halpern Federico Contributed Talk
  Wednesday, June 3, 2015
  HG E3, 10:30-10:50
  
  Contributed Talk
  
  Electromagnetic Turbulence Simulations of the Tokamak Scrape-Off Layer, Federico Halpern (EPFL, Switzerland)
  
  Co-Authors: Trach-Minh Tranh (EPFL, Switzerland); Felix Musil (EPFL, Switzerland); Paolo Ricci (EPFL, Switzerland); Fabio Riva (EPFL, Switzerland); Christoph Wersal EPFL, Switzerland)
  
  We present the new version of the Global Braginskii Solver (GBS) a code addressing tokamak boundary turbulence. GBS employs a 3D cartesian communicator and a parallel multigrid Poisson/Ampere solver, achieving excellent parallel scalability, and it is being ported into manycore and hybrid architectures. Two new capabilities resulted from improved algorithms: the inclusion of electromagnetic fluctuations at realistic plasma size, and a non-Boussinesq Poisson operator for the electrostatic potential. Simulations of the turbulent plasma dynamics are benchmarked against state-of-the-art imaging diagnostics in tomakak experiments, showing remarkable agreement in many observables.
• Hariri Farah Contributed Talk
  Wednesday, June 3, 2015
  HG E3, 10:50-11:10
  
  Contributed Talk
  
  Towards Petascale Particle-In-Cell (PIC) Simulations and Beyond, Farah Hariri (EPFL, Switzerland)
  
  Co-Authors: T. M. Tran (EPFL, Switzerland); A. Jocksch (CSCS / ETH Zurich, Switzerland); S. Brunner (EPFL, Switzerland); C. Gheller (CSCS / ETH Zurich, Switzerland); L. Villard (EPFL, Switzerland)
  
  Our focus is to port Particle-In-Cell (PIC) codes applied for studying turbulence in magnetic fusion relevant plasmas to various computational platforms equipped with GPUs and MICs. To this end, a so-called 'PIC ENGINE' framework containing the main features of the PIC algorithm has been designed and implemented with different approaches ensuring the locality of the method. Our implementation on the GPU is done using the OpenACC parallel programming standard. The algorithmic performance is tested on the Cray XC30 - Piz Daint - platform. Results show that our optimizations lead to significant performance improvements and will be compared to a reference parallel OpenMP implementation.
• Hehn Andreas Poster
  
  Poster
  
  PHY-07 High-Temperature Series Expansions for Quantum Lattice Models, Andreas Hehn (ETH Zurich, Switzerland)
  
  Co-Authors: Matthias Troyer (ETH Zurich, Switzerland)
  
  We present a framework to calculate high-order series expansions for quantum lattice models which allows to exploit the many levels of parallelism offered by today's supercomputers. Using this framework, we performed high-order high-temperature series expansions for frustrated spin-1/2 Heisenberg and Bose-Hubbard models on various lattice geometries.
• Helluy Philippe MS Presentation
  Tuesday, June 2, 2015
  HG E1.2, 14:30-15:00
  
  MS Presentation
  
  Asynchronous OpenCL/MPI Numerical Simulations of Conservation Laws, Philippe Helluy (Université de Strasbourg, France)
  
  Co-Authors: Thomas Strub (Axessim, France)
  
  Hyperbolic conservation laws are important mathematical models for describing many phenomena in physics or engineering. The Discontinuous Galerkin (DG) methods is a popular method for solving conservation laws. We present how we have implemented the DG method in the OpenCL/MPI framework in order to achieve high efficiency on recent GPU or multicore processors. The implementation relies on a splitting of the DG mesh into sub-domains and sub-zones. Different kernels are compiled according to the zones properties. In addition, we rely on the OpenCL asynchronous task graph in order to overlap OpenCL computations, memory transfers and MPI communications.
• Hill Sean MS Presentation
  Tuesday, June 2, 2015
  HG F30, 12:00-12:30
  
  MS Presentation
  
  The Human Brain Project, Sean Hill (EPFL, Switzerland)
  
  Co-Authors:
  
  The aim of the Human Brain Project (HBP) is to integrate global neuroscience knowledge and data into supercomputer-based models and simulations to accelerate our understanding of the human brain. To do this, HBP will deliver six collaborative ICT platforms: Neuroinformatics, Brain Simulation, High Performance Computing, Medical Informatics, High Performance Computing, Neuromorphic Computing and Neurorobotics. The HBP will create new technologies for interactive supercomputing, visualization and big data analytics; federated analysis of globally distributed data; simulation of the brain; objective classification of disease; and neuromorphic computing systems based on brain-like principles.
• Hirn Matthew MS Presentation
  Monday, June 1, 2015
  HG F1, 14:30-15:00
  
  MS Presentation
  
  Quantum Energy Regression Using Scattering Transforms, Matthew Hirn (École normale supérieure, France)
  
  Co-Authors: Stéphane Mallat (École normale supérieure, France); Nicolas Poilvert (Pennsylvania State University, USA)
  
  Physical functionals are usually computed as solutions of variational problems or from solutions of partial differential equations, which may require huge computations for complex systems. Quantum chemistry calculations of molecular energies is such an example. We present a novel approach for the regression of quantum mechanical energies based on the scattering transform of an intermediate electron density representation. The scattering transform is a deep convolutional network, composed of iterated wavelet transforms, that possesses appropriate invariant and stability properties for quantum energy regression. Numerical experiments give state of the art accuracy on planar organic molecules.
• Hodel Florian Poster
  
  Poster
  
  MAT-18 Solvation Effects on Electronic Energy Differences of a Co-Cubane, Florian Hodel (University of Zurich, Switzerland)
  
  Co-Authors: Sandra Luber (University of Zurich, Switzerland); Juerg Hutter (University of Zurich, Switzerland)
  
  We investigated the influence of different methods on electronic energy differences of ligand exchange reactions of a Co(II)-cubane based water oxidation catalyst. We compared average electronic energy differences obtained from sampling with born-Oppenheimer molecular dynamics using Kohn-Sham density functional theory to electronic energy differences of geometry optimized structures. In doing so, we also compared different levels of solvation ranging from hundreds of water molecules, first solvation shell, implicit solvent to in vacuo calculations, and finally also investigated barriers and free energies using the nudged elastic band method and metadynamics.
• Hoffmann Volker Poster
  
  Poster
  
  PHY-05 Chaos in Terrestrial Planet Formation, Volker Hoffmann (University of Zurich, Switzerland)
  
  Co-Authors: Simon Grimm (University of Zurich, Switzerland); Ben Moore (University of Zurich, Switzerland); Joachim Stadel (University of Zurich, Switzerland)
  
  The formation and evolution of planetary systems involves a variety of physical processes acting over a range of timescales. Here, we focus on simulations of terrestrial planet formation. We give an overview of formation models, relevant physics, and how simulations on HPC platforms drive the research. We then describe how these simulations are highly chaotic, such that identical initial conditions can rapidly diverge due to variations in the ordering of compute operations. This has implications for the predictive power of individual runs and suggests that a statistical approach is required.

I

• Iannuzzi Marcella Contributed Talk
  Wednesday, June 3, 2015
  HG F1, 10:50-11:10
  
  Contributed Talk
  
  Formation of Defects and Self-Healing Processes in Single Layer Hexagonal Boron Nitride Supported on Rh(111), Marcella Iannuzzi (University of Zurich, Switzerland)
  
  Co-Authors:
  
  The Nanomesh (nm) is the Moire-like structure formed by a single hBN layer grown on Rh. It can be functionalized by sputtering ions, while the pristine form can be recovered by annealing. Argon implantation leads to the formation of vacancy and interstitial defects. The interstitials are atoms trapped beneath the sp2 layer. The vacancies are formed when B and N atom are kicked out by Ar ion impact. DFT-based optimisations and MD to provide better understanding of the observed phenomena. We go beyond the static picture and address the defect formation kinetics, their mobility, leading to aggregation and eventually the self-healing of the sp2 lattice. H.Y. Cun et al. ACS Nano 8 (2014)
• Ismail-Zadeh Alik MS Presentation
  Monday, June 1, 2015
  HG F3, 15:30-16:00
  
  MS Presentation
  
  Data Assimilation in Geodynamical Modelling: Methods and Applications, Alik Ismail-Zadeh (Karlsruhe Institute of Technology, Germany)
  
  Co-Authors:
  
  Present geophysical, geochemical, and geodetic observations together with geological information provide a clue to understanding dynamics of the Earth interior in the past. Assimilation of present observations allows to constrain mantle temperature/composition and flow in the past using dynamical models. Quantitative tools are required to assimilate the data and hence to solve inverse retrospective problems in geodynamics. The basic inversion methods (adjoint and quasi-reversibility) including the methodologies uncertainties and requirements for HPC, and two case studies related to mantle/lithosphere dynamics and to a volcanic lava flow will be presented and discussed.
• Iwainsky Christian MS Presentation
  Tuesday, June 2, 2015
  HG F3, 11:30-12:00
  
  MS Presentation
  
  How Many Threads will be Too Many? On the Scalability of OpenMP Implementations, Christian Iwainsky (Technische Universität Darmstadt, Germany)
  
  Co-Authors: Sergei Shudler (German Research School for Simulation Sciences, Germany); Alexandru Calotoiu (German Research School for Simulation Sciences, Germany); Alexandre Strube (Forschungszentrum Jülich, Germany); Michael Knobloch (Forschungszentrum Jülich, Germany); Christian Bischof (Technische Universität Darmstadt, Germany); Felix Wolf (Technische Universität Darmstadt, Germany)
  
  Exascale systems are expected to exhibit much higher degrees of parallelism in terms of number of cores per node. While OpenMP is used today for exploiting parallelism on the level of individual nodes, it is unclear how well OpenMP implementations will scale to much higher numbers of threads. In this work, we apply automated performance modeling to examine the scalability of OpenMP constructs across different compilers and platforms, such as Intel Xeon multi-board, Intel Xeon Phi, and Blue Gene with compilers from GNU, IBM, Intel, and PGI. The resulting models reveal a number of scalability issues in implementations of OpenMP constructs and show unexpected differences between compilers.

J

• Jaggi Martin MS Presentation
  Tuesday, June 2, 2015
  HG F30, 14:30-15:00
  
  MS Presentation
  
  Communication Efficient Distributed Training of Machine Learning Models, Martin Jaggi (ETH Zurich, Switzerland)
  
  Co-Authors: Virginia Smith (UC Berkeley, USA); Martin Takáč (Lehigh University, USA); Jonathan Terhorst (UC Berkeley, USA); Sanjay Krishnan (UC Berkeley, USA); Thomas Hofmann (ETH Zurich, Switzerland); Michael I. Jordan (UC Berkeley, USA)
  
  Communication remains the most significant bottleneck in the performance of distributed optimization algorithms for large-scale machine learning. We propose a communication-efficient framework, COCOA, that uses local computation in a primal-dual setting to dramatically reduce the amount of necessary communication. We provide a strong convergence rate analysis for this class of algorithms, as well as experiments on real-world distributed datasets with implementations in Spark. In our experiments, we find that as compared to state-of-the-art mini-batch versions of SGD and SDCA algorithms, COCOA converges to the same .001-accurate solution quality on average 25× as quickly.
• Jahanbakhsh Ebrahim MS Presentation
  Monday, June 1, 2015
  HG E22, 17:00-17:30
  
  MS Presentation
  
  Silt Erosion Simulation Using Finite Volume Particle Method, Ebrahim Jahanbakhsh (EPFL, Switzerland)
  
  Co-Authors:
  
  Silt erosion is a destructive phenomenon that may occur in hydropower plants. This work presents a new erosion prediction model in which, fluid flow and solid deformation equations are discretized by Finite Volume Particle Method (FVPM), and silt contact forces are calculated according to Hertz contact theory. FVPM is a meshless method which includes many of the desirable features of the mesh-based finite volume method. To obtain a more accurate and robust model, a new FVPM formulation is presented in which particle interaction integrals are evaluated exactly and efficiently. To validate the new silt erosion model, 2D and 3D erosion cases are simulated and compared with experimental data.
• Jansen Gunnar MS Presentation
  Monday, June 1, 2015
  HG E22, 15:30-16:00
  
  MS Presentation
  
  Modeling 3D THM Processes in Geothermics with Continuum Mechanics, Gunnar Jansen (University of Neuchatel, Switzerland)
  
  Co-Authors: Reza Sohrabi (University of Neuchatel, Switzerland); Boris Galvan (University of Neuchatel, Switzerland); Stephen A. Miller (University of Neuchatel, Switzerland)
  
  The geothermal community is challenged by the Energy Strategy 2050 goal to uncover the potential of renewable energy resources in Switzerland. To this end we implement a novel finite element based simulation tool for rock and fluid physics. It efficiently couples the dominant processes in the subsurface, such as fracture nucleation and growth using continuum mechanics, anisotropic multiphase fluid flow and heat transfer to provide a deeper understanding. We aim at state of the art performance by using massively parallel tools and modern accelerator techniques. We present first results from 3D fully coupled THM-simulations and evaluate the performance from a series of benchmark simulations.
• Jensen Jan H. MS Presentation
  Monday, June 1, 2015
  HG E3, 16:00-16:30
  
  MS Presentation
  
  Quantum Biochemistry, Jan H. Jensen (University of Copenhagen, Denmark)
  
  Co-Authors:
  
  I will give an overview of my groups latest work in applying quantum chemistry to biochemical problems such as enzyme catalysis, enzyme design, protein-ligand binding and protein structure determination by NMR. Preview: http://youtu.be/V45gEsJXGmI
• Jiang Yan-Fei MS Presentation
  Tuesday, June 2, 2015
  HG E3, 11:00-11:30
  
  MS Presentation
  
  Numerical Algorithms to Solve the Time-Depedent Radiative Transfer Equation Based on VET and Finite Volume Method, Yan-Fei Jiang (Harvard-Smithsonian Center for Astrophysics, USA)
  
  Co-Authors: James Stone (Princeton University, USA); Shane Davis (University of Virginia, USA)
  
  Two algorithms to solve the time dependent radiation (magneto)hydrodynamic equations will be described. The first method solves the two radiation momentum equations, which are closed with the variable Eddington tensor based on time-independent radiative transfer equation using short characteristics. The second method solves the time-dependent radiative transfer equation for specific intensities directly based on finite volume method. Both methods are accurate for both optically thick and optically thin regimes and can be used to model both gas pressure and radiation pressure dominated flows. Application of both methods to study black hole accretion disks will also be described.
• Jocksch Andreas Poster
  
  Poster
  
  PHY-02 A Bucketsort Algorithm on GPUs for the Particle-In-Cell (PIC) Method, Andreas Jocksch (CSCS / ETH Zurich, Switzerland)
  
  Co-Authors: Farah Hariri (EPFL, Switzerland); Trach-Minh Tran (EPFL, Switzerland); Stefan Brunner (EPFL, Switzerland); Claudio Gheller (CSCS / ETH Zurich, Switzerland); Laurent Villard (EPFL, Switzerland)
  
  The Particle-In-Cell (PIC) method is effectively used in many scientific simulation codes. We present variants of an in place bucketsort algorithm for the PIC method targeting GPUs. In order to optimise performance data locality is required. This relies on efficient sorting algorithms. We discuss the different datastructures that are used where particles are distributed to buckets or stored contiguously. Finally, we show that the performance increases with the amount of storage provided and with the orderliness of the particles. The overall PIC algorithm performs at its best if the sorting is applied.
• Judd Kenneth MS Presentation
  Monday, June 1, 2015
  HG F30, 15:30-16:00
  
  MS Presentation
  
  An Agenda for Bringing Computational Science to Economics, Kenneth Judd (Stanford University, USA)
  
  Co-Authors:
  
  Computational science is becoming the third pillar of science. Unfortunately, economics still largely relies on only two pillars - theory and observations. This can change but only after identifying the mathematical and computational methods that match the needs of economic modeling. I will point to the unique features of economics modeling, and illustrate how a combination of computational tools can be used to address a policy issue in the area of climate change.
• Juillard Michel MS Presentation
  Monday, June 1, 2015
  HG F30, 17:00-17:30
  
  MS Presentation
  
  DYNARE, a Toolbox for Solving and Estimating DSGE Models, Michel Juillard (Banque de France, France)
  
  Co-Authors:
  
  Dynare is a software tool to solve, simulate and estimate dynamic stochastic general equilibrium models. This requires heavy use of numerical methods: such as solving large sparse complementarity problems, function approximation with the perturbation approach or MCMC simulation of the posterior density. Dynare rests on a strict separation of the specification of a particular model and the use of standard software tools implied the development of a specificmodeling language. The numerical routines are provided as a Matlab/Octave toolbox. Dynare is open source so as to make used algorithms transparent. Challenges ahead include occasionally binding constraints and large shocks.

K

• Kaman Tulin Contributed Talk
  Wednesday, June 3, 2015
  HG F3, 10:50-11:10
  
  Contributed Talk
  
  Large Eddy Simulation of Turbulent Combustion with Finite Rate Chemistry Model, Tulin Kaman (ETH Zurich, Switzerland)
  
  Co-Authors: Xiaoxue Gong (Stony Brook University, USA); Ying Xu (Stony Brook University, USA); James Glimm (Stony Brook University, USA)
  
  We present Large-Eddy simulations (LES) of the combustion chamber of a scram jet experiment with finite rate chemistry, using only the fundamental chemical reaction equations. The main focus of this work is to provide necessary guidance to the mesh resolution needed to perform a full-domain numerical simulation of a three-dimensional scram jet, which is M=7 experimental aircraft. The combustion process modeled by the finite rate chemistry is based on the notion that it can be resolved on a larger scale than the levels needed to resolve turbulence.
• Keyes David Invited Presentation
  Wednesday, June 3, 2015
  HG F30, 12:00-12:50
  
  Invited Presentation
  
  IP6 Algorithmic Adaptations to Extreme Scale, David Keyes (King Abdullah University of Science and Technology, Saudi Arabia)
  
  Chair: Olaf Schenk (Università della Svizzera italiana, Switzerland)
  
  Abstract
  Algorithmic adaptations are required to use anticipated exascale hardware near its potential, since the code base has been engineered to squeeze out flops. Instead, algorithms must now squeeze out synchronizations, memory, and transfers, while extra flops on locally cacheable data represent small costs in time and energy. Today's scalable solvers, in particular, exploit frequent global synchronizations. After decades of programming model stability with bulk synchronous processing (BSP), new programming models and new algorithmic capabilities (to take advantage, e.g., of forays in data assimilation, inverse problems, and uncertainty quantification) must be co-designed with the hardware. We briefly recap the architectural constraints, highlight some work at KAUST, and outline future directions.
  
  Biography
  David Keyes is Director of the Extreme Computing Research Center at KAUST. He earned a BSE in aerospace and mechanical sciences from Princeton in 1978 and PhD in applied mathematics from Harvard in 1984. Keyes works at the interface between parallel computing and the numerical analysis of PDEs, with a focus on scalable implicit solvers. Newton-Krylov-Schwarz (NKS), Additive Schwarz Preconditioned Inexact Newton (ASPIN), and Algebraic Fast Multipole (AFM) methods are methods he helped name and popularize. Before joining KAUST as a founding dean in 2009, he led scalable solver software projects in the ASCI and SciDAC programs of the US Department of Energy, ran university collaboration programs at NASA's ICASE and the LLNL's ISCR, and taught at Columbia, Old Dominion, and Yale Universities. He is a Fellow of SIAM and AMS.
• Klatt Torbjörn MS Presentation
  Tuesday, June 2, 2015
  HG E22, 15:00-15:30
  
  MS Presentation
  
  The Parallel-in-Time Integration Library PFASST++ Applied to Molecular Dynamics, Torbjörn Klatt (Forschungszentrum Jülich, Germany)
  
  Co-Authors: Robert Speck (Forschungszentrum Jülich, Germany); Mathias Winkel (Università della Svizzera italiana, Switzerland); Daniel Ruprecht (Università della Svizzera italiana, Switzerland); Matthew Emmett (Lawrence Berkeley National Laboratory, USA)
  
  Iterative parallel-in-time integration methods like the 'parallel full approximation scheme in space and time' (PFASST) provide parallelism along the temporal axis by integrating multiple time-steps simultaneously. Along the line of a high-order Boris-SDC integrator for charged particles in external magnetic fields we present the evolution of three implementations of PFASST. Starting from a prototyping Python framework, we explain different implementation strategies and conceptual challenges. We then present results using the HPC library libpfasst and describe our new, lightweight PFASST++ code, highlighting the lessons learned from the interplay with existing HPC applications.
• Koestler Harald MS Presentation
  Tuesday, June 2, 2015
  HG F3, 10:30-11:00
  
  MS Presentation
  
  EXASTENCILS - Advanced Stencil-Code Engineering, Harald Koestler (University of Erlangen-Nuremberg, Germany)
  
  Co-Authors: Christian Lengauer (Universität Passau, Germany); Armin Größlinger (Universität Passau, Germany); Sven Apel (Universität Passau, Germany); Jürgen Teich (University of Erlangen-Nuremberg, Germany); Frank Hannig (University of Erlangen-Nuremberg, Germany); Matthias Bolten (University of Wuppertal, Germany); Ulrich Rüde (University of Erlangen-Nuremberg, Germany)
  
  The central goal of ExaStencils is to develop a radically new software technology for applications with exascale performance. To reach this goal, the project focusses on a comparatively narrow but very important application domain of stencil codes. The aim is to enable a simple and convenient formulation of problem solutions in this domain. The software technology developed in ExaStencils shall facilitate the highly automatic generation of a large variety of efficient implementations via the judicious use of domain-specific knowledge in each of a sequence of optimization steps such that, at the end, exascale performance results.
  Minisymposium
  Minisymposium
  
  MS16 Software for Exascale Computing
  
  Organizer: Harald Koestler (University of Erlangen-Nuremberg, Germany)
  
  Co-Organizers: Matthias Bolten (University of Wuppertal, Germany)
  
  This minisymposium in centered around the German Priority Programme 'Software for Exascale Computing' (SPPEXA). SPPEXA addresses fundamental research on the various aspects of HPC software, which is particularly urgent against the background that we are currently entering the era of ubiquitous massive parallelism. It drives research towards extreme-scale computing in six areas or research directions: computational algorithms, system software, application software, data management and exploration, programming, and software tools.
• Kolev Tzanio MS Presentation
  Tuesday, June 2, 2015
  HG E1.2, 11:00-11:30
  
  MS Presentation
  
  Scalable High-Order Finite Element Discretizations and Solvers with MFEM, hypre and BLAST, Tzanio Kolev (Lawrence Livermore National Laboratory, USA)
  
  Co-Authors: J. Cerveny (Lawrence Livermore National Laboratory, USA); V. Dobrev (Lawrence Livermore National Laboratory, USA); A. Grayver (Lawrence Livermore National Laboratory, USA); R. Rieben (Lawrence Livermore National Laboratory, USA); I. Karlin (Lawrence Livermore National Laboratory, USA); M. Kumbera (Lawrence Livermore National Laboratory, USA)
  
  High-order (HO) finite element (FE) discretizations are a natural fit for future HPC hardware. In this talk we present our work on scalable HO FE software that combines the modular FE library MFEM (mfem.googlecode.com), the hypre library of linear solvers (llnl.gov/casc/hypre), and the HO shock hydrodynamics code BLAST (llnl.gov/casc/blast). We discuss the FE abstractions provided by MFEM, the use of hypre data structures/kernels for efficient parallel assembly, and demonstrate the benefits with respect to scaling/GPU acceleration in BLAST. We also consider HO adaptive refinement and present recent work on scalable algebraic multigrid for electromagnetic problems in this context.
  MS Presentation
  Tuesday, June 2, 2015
  HG E22, 14:30-15:00
  
  MS Presentation
  
  Parallel Time Integration with Multigrid, Tzanio Kolev (Lawrence Livermore National Laboratory, USA)
  
  Co-Authors: R. Falgout (Lawrence Livermore National Laboratory, USA); S. Friedhoff (Tufts University, USA); V. Dobrev (Lawrence Livermore National Laboratory, USA); S. MacLachlan (Tufts University, USA); N. A. Petersson (Lawrence Livermore National Laboratory, USA); U. Yang (Lawrence Livermore National Laboratory, USA)
  
  In this talk we examine an optimal-scaling parallel time integration method, multigrid-reduction-in-time (MGRIT). MGRIT applies multigrid to the time dimension by iteratively solving all time steps simultaneously. This results in a non-intrusive approach that wraps existing time evolution codes. MGRIT allows for various time integrators, adaptive refinement/coarsening in time and space and handles non-linear problems through FAS multigrid. Aspects of the theory, software implementation in the XBraid library (llnl.gov/casc/xbraid), and practical result for a variety of problems will be presented, e.g., explicit/implicit time integration, non-linear diffusion and compressible Navier-Stokes.
• Kosakowski Georg Poster
  
  Poster
  
  ENG-01 OpenGeoSys-GEM: A Coupled Thermo-Hydro-Chemical(-Mechanical) Code THC(M) for Geoscientific and Engineering Applications, Georg Kosakowski (Paul Scherrer Institut, Switzerland)
  
  Co-Authors: Dmitrii A. Kulik (Paul Scherrer Institut, Switzerland); Allan Leal (Paul Scherrer Institut, Switzerland); Haibing Shao (Helmholtz Centre for Environmental Research - UFZ, Germany); Wenqing Wang (Helmholtz Centre for Environmental Research - UFZ, Germany); Olaf Kolditz (Helmholtz Centre for Environmental Research - UFZ, Germany)
  
  We present a coupled reactive transport code that uses the open-source, multi-platform, multi-physics Finite Element package OpenGeoSys for solving flow and transport equations and the Gibbs Energy Minimization (GEM) based chemical solvers GEMS3K and Reaktoro for obtaining chemical equilibrium speciation. The coupled code can be used to simulate systems from lab to some extend field scale with very complex chemistry (e.g. several non-ideal solution phases in the system). The code runs on various platforms, including massively parallel systems where it uses an overlapping domain decomposition approach in combination with execution of multiple threads for solving chemical equilibria.
• Kotlikoff Laurence MS Presentation
  Monday, June 1, 2015
  HG F30, 16:00-16:30
  
  MS Presentation
  
  Generational Policy and Aging in Closed and Open Dynamic General Equilibrium Models, Laurence Kotlikoff (Boston University, USA)
  
  Co-Authors:
  
  This talk will review my recent co-authored work on generational policy and aging in closed and open dynamic general equilibrium models.
• Kourounis Drosos MS Presentation
  Monday, June 1, 2015
  HG F3, 14:00-14:30
  
  MS Presentation
  
  Full Waveform Inversion for the Identifiable Subspace, Drosos Kourounis (Università della Svizzera italiana, Switzerland)
  
  Co-Authors: Olaf Schenk (Università della Svizzera italiana, Switzerland)
  
  Full-waveform inversion (FWI) optimizes subsurface model estimates to derive high-fidelity geological models. A mathematically sound method is described for selecting the part of the geological parameters that is best identifiable from the seismic acquisition geometry. This is combined with either interior-point or sequential quadratic programming methods for performing FWI for the subset of parameters that have been characterized as identifiable. Numerical results are presented on several examples of increased complexity.
• Krause Rolf MS Presentation
  Tuesday, June 2, 2015
  HG E22, 13:30-14:00
  
  MS Presentation
  
  Numerical Simulation of Skin Transport using Parareal, Rolf Krause (Università della Svizzera italiana, Switzerland)
  
  Co-Authors: Andreas Kreienbuehl (Università della Svizzera italiana, Switzerland); Daniel Ruprecht (Università della Svizzera italiana, Switzerland); Robert Speck (Università della Svizzera italiana, Switzerland); Arne Naegel (University of Frankfurt, Germany); Gabriel Wittum (University of Frankfurt, Germany)
  
  The talk will present recent results on the applicability of the time-parallel Parareal algorithm to the in-silico modeling of permeation of chemicals through human skin. For our analysis we use a brick-and-mortar setup as a model problem. A C++ library implementing Parareal is combined with the ug4 simulation framework, which provides the spatial discretization and parallelization. By numerical examples, we investigate Parareal's performance on an anisotropic domain with large jumps in the diffusion coefficients. While speedup from the time parallelization is shown to be possible, load balancing is identified as an important aspect.
  Minisymposium
  Minisymposium
  
  MS07 Modeling and Simulation in Geo-Energy and Hydro-Power
  
  Organizer: Rolf Krause (Università della Svizzera italiana, Switzerland)
  
  Co-Organizers: Thomas Driesner (ETH Zurich, Switzerland)
  
  Mathematical modeling and numerical simulation are of fundamental importance for the development of new and innovative technologies in the energy sector. In this minisymposium, we focus on the exchange of methods, concepts and ideas in modeling and simulation of complex application problems from the two areas geo-energy and hydro-power. This includes the efficient numerical solution of coupled nonlinear systems, or problems involving interface effects and highly complex geometries.
• Kreienbuehl Andreas MS Presentation
  Tuesday, June 2, 2015
  HG E22, 14:00-14:30
  
  MS Presentation
  
  Parallel-in-Time Simulation of Black Hole Formation, Andreas Kreienbuehl (Università della Svizzera italiana, Switzerland)
  
  Co-Authors: Pietro Benedusi (Università della Svizzera italiana, Switzerland); Daniel Ruprecht (Università della Svizzera italiana, Switzerland); Rolf Krause (Università della Svizzera italiana, Switzerland)
  
  We present and discuss an application of the parallel-in-time integration method Parareal to the Einstein gravity equations for a collapsing massless scalar field in spherical symmetry. We show that Parareal captures the correct black hole formation event and generates the proper mass scaling law. Moreover, if the computational load is properly balanced in time, Parareal features speedup when compared to the serial two-step Lax-Wendroff Richtmyer scheme. As we are using a two-level approach in both space and time, we furthermore analyze the influence of the spatial interpolation. Finally, the efficiency of different Parareal implementations is compared with respect to energy consumption.
• Kremer Kurt MS Presentation
  Monday, June 1, 2015
  HG F1, 14:00-14:30
  
  MS Presentation
  
  Soft Matter and Data Repositories, a Personal View, Kurt Kremer (Max Planck Institute for Polymer Research, Germany)
  
  Co-Authors:
  
  Compared to many low molecular weight materials soft matter is not that perfectly characterized. E.g. polydispersity directly affects material properties. Furthermore chemical variation, for both synthetic and biological soft matter, is huge. Thus there are two options to provide data repositories, i.e. by experiment or simulation and quantum chemistry or by combination of both. I will present some thoughts of how, mainly based on theoretical work, they might be extended to form a basis of generally applicable data repositories.
• Kronbichler Martin MS Presentation
  Tuesday, June 2, 2015
  HG E1.2, 14:00-14:30
  
  MS Presentation
  
  Shared Memory Parallelization Strategies for Matrix-free Finite Element Operator Evaluation, Martin Kronbichler (Technische Universität München, Germany)
  
  Co-Authors: Katharina Kormann (Technische Universität München, Germany)
  
  Matrix-free finite element operator evaluation for quadratic and higher order hexahedral bases has been shown to be faster per degree of freedom than sparse matrix-vector products for linear elements. The key to success is substituting a memory intensive kernel by fast tensorial integration in a way that suits modern hardware. Our shared-memory implementation dynamically schedules integration tasks on non-conflicting cells and faces, offering coloring, partitioning, and more abstract flow graph concepts. Results show that hardware trends with a growing number of cores per memory controller and widening vector units require particular emphasis on memory access patterns into solution vectors.
• Kuczaj Arkadiusz MS Presentation
  Monday, June 1, 2015
  HG E1.2, 14:30-15:00
  
  MS Presentation
  
  Towards Computing Aerosol Flow and Exposure for In-Vitro Research Purposes, Arkadiusz Kuczaj (Philip Morris Int. & University Twente, Switzerland/Netherlands)
  
  Co-Authors: Markus Nordlund (Philip Morris International, Switzerland)
  
  Systems Toxicology combines high content experimental data obtained at the molecular, cellular, organ, organism, and with computational and mathematical sciences aims to identify Pathways of Toxicity. Linking conditions occurring in a respiratory tract with those in the exposure systems requires detailed understanding of flow and deposition of aerosols. The presentation will showcase the application of Computational Fluid Dynamics in simulating flow in a human lung airway model and outlines opportunities for leveraging simulated data to develop well-designed in-vitro respiratory toxicology experiments that have the potential to allow in-vitro to in-vivo exposure extrapolations.
• Kuiper Rolf MS Presentation
  Tuesday, June 2, 2015
  HG E3, 11:30-12:00
  
  MS Presentation
  
  Hybrid Radiation Transport Methods for Star and Planet Formation, Rolf Kuiper (University of Tübingen, Germany)
  
  Co-Authors:
  
  Many astrophysical studies require an highly accurate treatment of radiation transport to properly determine heating and cooling timescales, observables, or dependent physical properties. At the same time the radiation transport algorithm should be as fast as possible to allow for an efficient usage of computing resources. Hybrid approaches try to combine the accuracy of a high-order radiation transport method with the speedup of sensible approximations. Exemplary, I will present the theory and numerics of our own hybrid radiation transport scheme developed for hydrodynamical studies of star formation and accretion disks. Benchmark tests and science applications will be addressed.

L

• Lagardère Louis Poster
  
  Poster
  
  MAT-05 Achieving Linear Scaling in Computational Cost for a Fully Polarizable MM/Continuum Embedding, Louis Lagardère (UPMC, France)
  
  Co-Authors: Stefano Caprasecca (Università di Pisa, Italy); Sandro Jurinovich (Università di Pisa, Italy); Benjamin Stamm (Pierre-and-Marie-Curie University, France); Filippo Lipparini (Pierre-and-Marie-Curie University, France)
  
  We present a new, efficient implementation of a fully polarizable QM/MM/continuum model based on an induced-dipoles polarizable force field and on the Conductor-like Screening Model as a polarizable continuum in combination with a self-consistent field QM method. A linear scaling with respect to both the computational cost and the memory requirements is achieved thanks to the use of the recently developed ddCOSMO model for the continuum and the Fast Multipole Method for the force field, together with an efficient iterative procedure. Therefore, it becomes possible to include in the classical layer as much as several tens of thousands of atoms with a limited computational effort.
• Lanteri Stéphane MS Presentation
  Tuesday, June 2, 2015
  HG E1.2, 11:30-12:00
  
  MS Presentation
  
  Scalable High Order Finite Element Solvers for Computational Nanophotonics, Stéphane Lanteri (INRIA, France)
  
  Co-Authors: Claire Scheid (Université de Nice - Sophia Antipolis, France); Jonathan Viquerat (INRIA, France)
  
  The numerical study of electromagnetic wave propagation in interaction with nanometer scale structures generally relies on the solution of the system of time-domain Maxwell equations, taking into account an appropriate physical dispersion model for characterizing the material properties of the involved nanostructures at optical frequencies. In this talk, we will present our recent efforts aiming at the design of high performance Discontinuous Galerkin Time-Domain (DGTD) methods for 3d nanophotonic applications. We will consider more particularly DGTD methods for solving the system of Drude-Maxwell equations.
  Minisymposium
  Minisymposium
  
  MS15 Recent Advances on Scalable High-Order Finite Element Type Schemes for PDEs
  
  Organizer: Stéphane Lanteri (INRIA, France)
  
  Co-Organizers: Peter Bastian (Universität Heidelberg, Germany)
  
  The availability of exaflop computers will enable the numerical treatment of more challenging physical problems. However, the simulation of such complex physical phenomena will require very accurate and scalable numerical schemes. In this context, the objective of this minisymposium is to address these needs for different types of PDE models underlying challenging physical problems from modern computational sciences. Related topics are among others, high order or spectral accuracy, space and time adaptivity, finite element methods, finite volume methods, discontinuous Galerkin methods, multiscale methods, stencil codes, etc.
• Lazzaro Alfio Contributed Talk
  Wednesday, June 3, 2015
  HG E1.1, 10:50-11:10
  
  Contributed Talk
  
  A Communication Avoiding Scheme for Sparse Matrix Multiplication, Alfio Lazzaro (ETH Zurich, Switzerland)
  
  Co-Authors: Ole Schuett (ETH Zurich, Switzerland); Florian Schiffmann (ETH Zurich, Switzerland); Joost VandeVondele (ETH Zurich, Switzerland)
  
  DBCSR is a sparse matrix library designed to efficiently perform sparse matrix-matrix multiplication, among other operations. It is MPI and OpenMP parallel, and can exploit accelerators. It is developed as part of CP2K, where it provides core functionality for linear scaling electronic structure theory. The multiplication algorithm is based on Cannon's algorithm, whose scalability is limited by the MPI communication time. We present a novel implementation based on a 2.5D algorithm, which takes in account the sparsity of the problem in order to reduce the MPI communication.
• Leißa Roland MS Presentation
  Tuesday, June 2, 2015
  HG F1, 14:30-15:00
  
  MS Presentation
  
  AnyDSL: A Compiler-Framework for Domain-Specific Libraries, Roland Leißa (Saarland University, Germany)
  
  Co-Authors: Richard Membarth (Deutsches Forschungszentrum für Künstliche Intelligenz, Germany)
  
  In this talk, we present AnyDSL, a compiler-framework that allows to define arbitrary domain-specific abstractions in the form of a library. AnyDSL allows to express hierarchies of abstractions and efficient transformation to lower-level abstractions through refinement. At the lowest level, the code can be optimized via exposed compiler functionality such as partial evaluation or target code generation. Further, we will present a DSL for stencil codes that achieves the same performance as hand-optimized implementations on CPUs as well as on GPUs.
• Lelièvre Tony MS Presentation
  Monday, June 1, 2015
  HG E3, 14:30-15:00
  
  MS Presentation
  
  Accelerated Dynamics in Molecular Simulations, Tony Lelièvre (Ecole des Ponts, France)
  
  Co-Authors:
  
  I will present multiscale-in-time strategies to accelerate molecular dynamics simulations. Starting from a metastable stochastic process, the idea is to use the underlying jump process between metastable states in order to accelerate exit events in a statistically consistent way. For example, it is possible to parallelize in time the integration of the dynamics. These ideas have been proposed A.F. Voter in the nineties, and their mathematical foundations have been recently understood thanks to the notion of quasi stationary distribution.
• Leonardi Filippo Poster
  
  Poster
  
  CSM-01 Approximation of Admissible Measure Valued Solutions for Incompressible Euler Equations Valued Solutions for Incompressible Euler Equations, Filippo Leonardi (ETH Zurich, Switzerland)
  
  Co-Authors: Siddhartha Mishra (ETH Zurich, Switzerland)
  
  We propose a new, first order numerical scheme for the approximation of incompressible Euler equations. This scheme has a number of interesting properties, which mimic the behaviour of the governing equations. We also propose a procedure, which allows a consistent approximation of admissible measure valued solutions. Numerical experiments demonstrate statistical convergence of quantities of interest.
• Leutwyler David Contributed Talk
  Wednesday, June 3, 2015
  HG E21, 11:10-11:30
  
  Contributed Talk
  
  Towards Continental-Scale Convection-Resolving Climate Simulations on GPUs, David Leutwyler (ETH Zurich, Switzerland)
  
  Co-Authors: Oliver Fuhrer (MeteoSwiss, Switzerland); Daniel Lüthi (ETH Zurich, Switzerland); Christoph Schär (ETH Zurich, Switzerland)
  
  Climate simulations using horizontal resolution of O(1km) allow to explicitly resolve deep convection. Precipitation processes are then represented much closer to first principles and allow for an improved representation of the water cycle. We present a set of convection-resolving simulations covering the European-scale using a COSMO-model prototype enabled for GPUs, integrated on a computational mesh of 1536x1536x60 grid points. Results illustrate the interactions between synoptic-scale and meso-scale atmospheric circulations, and the initiation of new convective cells by propagating cold pools. Furthermore we discuss computational benefits from using GPUs for climate simulations.
• Lieber Matthias MS Presentation
  Monday, June 1, 2015
  HG E1.1, 14:00-14:30
  
  MS Presentation
  
  Scalable Dynamic Load Balancing of Detailed Cloud Physics with FD4, Matthias Lieber (Technische Universität Dresden, Germany)
  
  Co-Authors:
  
  Load balancing of large-scale scientific simulations is a challenging task. My talk shows how the open source framework FD4 (Four-Dimensional Distributed Dynamic Data structures) enables scalable dynamic load balancing of detailed cloud physics in the atmospheric model COSMO-SPECS+FD4. The concept is based on separating data structures and decomposition such that the cloud physics model can be partitioned independently of the atmospheric model. To increase scalability, a hierarchical space-filling curve partitioning algorithm is used. Benchmarks on a Blue Gene/Q system show the scalability up to 256k ranks. FD4 can also be applied to other multiphase or multiphysics problems.
• Lipkova Jana Poster
  
  Poster
  
  LS-12 Tumor Induced Brain Deformations, Jana Lipkova (Technische Universität München, Germany)
  
  Co-Authors: Bjoern Menze (Technische Universität München, Germany); Petros Koumoutsakos (ETH Zurich, Switzerland); John Lowengrub (University of California, USA)
  
  Increased intracranial pressure is the most critical symptom of brain cancer, leading to brain deformation, tissue and nerve compression. We present novel approach for modelling tumour progression together with corresponding brain deformations in patient individual anatomies obtained from medical images. We use the diffuse domain approach to implicitly capture complex brain anatomy. The governing equations are then appropriately modified and extended to a larger, regular domain. This approach helps to reduce errors caused by anatomy missegmentation and allows the use of efficient numerical methods that would not be applicable to the original problem formulation.
• Lippert Thomas MS Presentation
  Tuesday, June 2, 2015
  HG F30, 13:30-14:00
  
  MS Presentation
  
  Scientific Big Data Analytics at the John von Neumann-Institute for Computing (NIC), Thomas Lippert (Forschungszentrum Jülich, Germany)
  
  Co-Authors:
  
  The importance of data analytics, management, sharing and preservation of very big, often heterogeneous or distributed data sets - besides the basic technical requirements transfer and storage - is of increasing significance for science, research and industry. The John von Neumann Institute for Computing, a joint institute by DESY, GSI and Forschungszentrum Jülich in Germany, is going to establish a call for project submission in the field of scientific big data analytics (SBDA). The goal is to extend and optimize the existing HPC and data services. A call for expressions of interest has been launched in order to identify and analyze the needs of the scientific communities.
• Lo Cynthia Poster
  
  Poster
  
  MAT-20 Towards an Ab Initio Electronic Transport Model for Photovoltaic Materials Design, Cynthia Lo (Washington University in St. Louis, United States)
  
  Co-Authors: Alireza Faghaninia (Washington University in St. Louis, USA); Maria Stoica (Washington University in St. Louis, USA)
  
  Improvements in computational models of electronic transport are necessary for accurate predictions of photovoltaic behavior. Existing efforts to solve the Boltzmann transport equation employ approximations that result in incorrect trends in the calculated properties. Instead, we propose an ab initio model that explicitly determines elastic and inelastic scattering rates and charge carrier mobility. We have validated the calculated mobility of GaAs and InN, as a function of temperature and carrier concentration, against experimental data. We now show how this model is being used to screen new transparent conducting materials for low-cost photovoltaics.
• Locans Uldis Poster
  
  Poster
  
  CSM-04 Dynamic Kernel Scheduler (DKS) - a Thin Software Layer Between Host Application and Hardware Accelerators, Uldis Locans (University of Latvia & Paul Scherrer Institute, Switzerland)
  
  Co-Authors: Andreas Adelmann (Paul Scherrer Institut, Switzerland); Andreas Suter (Paul Scherrer Institut, Switzerland)
  
  Hardware accelerators, such as GPUs and Intel MICs, provide a huge performance potential for HPC applications. However, due to different hardware architectures and development frameworks, taking full advantage of every device and creating manageable code is becoming a challenging task. Dynamic Kernel Scheduler (DKS) provides host application with an interface to schedule communication and task execution on the device and handles all the device and framework specific details necessary to execute these tasks on the accelerator. The concepts and first version of DKS will be presented as well as first results of integrating DKS in different applications (FFT Poisson, MC, Chi-square).
• Luber Sandra Poster
  
  Poster
  
  MAT-14 New Spectroscopic Approaches for Periodic Systems, Sandra Luber (University of Zurich, Switzerland)
  
  Co-Authors: Jürg Hutter (University of Zurich, Switzerland)
  
  Knowledge about local properties is extremely helpful for the analysis of structures and interactions. Calculations provide additional insight allowing, for instance, to quantify the contributions of solute and solvent molecules [1] or adsorbates on solids. We present novel, computationally efficient methods for the calculation of properties for periodic systems such as liquids and solids. These are applied to calculate, among others, vibrational spectra via ab initio molecular dynamics [2,3]. References: [1] S. Luber, J. Phys. Chem. A, 117 2760 (2013). [2] S. Luber, M. Iannuzzi, J. Hutter, J. Chem. Phys. 141 094503 (2014). [3] S. Luber, J. Chem. Phys. 141 234110 (2014).
• Ludwig Thomas MS Presentation
  Tuesday, June 2, 2015
  HG F30, 11:00-11:30
  
  MS Presentation
  
  Data Management in Climate Science - Cost-Benefit Considerations at DKRZ, Thomas Ludwig (DKRZ, Germany)
  
  Co-Authors:
  
  Climate science is highly data intensive and data is the raw material for the scientist for gaining new insights. The German Climate Computing Centre DKRZ stores dozens of Petabyte of modelling output on disks and tapes. Costs for this part of the infrastructure are increasing and we look into methods of how to reduce them. The talk will introduce the most important aspects of data management within climate science and cover various aspects during the lifecycle of data. We will explain how a well-balanced HPC system needs to be configured and which investment and operational costs are generated by it. A look at Exascale systems will conclude the presentation.
• Luisier Mathieu Poster
  
  Poster
  
  MAT-04 Accelerating Nano-device Simulations with Extreme-scale Algorithms and Software Co-integration, Mathieu Luisier (ETH Zurich, Switzerland)
  
  Co-Authors: Nicola Marzari (EPFL, Switzerland); Olaf Schenk (Università della Svizzera italiana, Switzerland)
  
  The ANSWERS project (Accelerating nano-device simulations with extreme-scale algorithms and software co-integration) is currently supported by the Platform for Advanced Scientific Computing (PASC). Its main objectives consist in developing numerical algorithms and software on hybrid manycore architectures specifically dedicated to ab-initio electronic structure and quantum transport calculations of realistic nanostructures. This poster intends to summarize the progresses of the project and highlight the future research directions.

M

• Maday Yvon MS Presentation
  Monday, June 1, 2015
  HG E3, 13:00-13:30
  
  MS Presentation
  
  A Fast Domain Decomposition Algorithm for Continuum Solvation Models, Yvon Maday (Université Pierre et Marie Curie, France)
  
  Co-Authors: Filippo Lipparini (Université Pierre et Marie Curie, France); Louis Lagardère (Université Pierre et Marie Curie, France); Benjamin Stamm (Université Pierre et Marie Curie, France); Eric Cancès (École des Ponts ParisTech, France / INRIA, France); Jean-Philip Piquemal (Université Pierre et Marie Curie, France); Benedetta Mennucci (Università di Pisa, Italy)
  
  Continuum solvation models are nowadays among the most popular tools in computational chemistry to include the effects of the chemical environment in the description of a molecular property or process. The strength of these models is their simplicity of use and cost-effectiveness. Continuum solvation models have been used together with both classical molecular mechanics (MM), quantum mechanics (QM) and hybrid (QM/MM) levels of theory to describe the solute. In this talk, I will present the basics of an efficient, parallel, linear scaling implementation of the conductor-like screening model (COSMO), based on a domain decomposition algorithm adapted to the integral formulation of the problem.
• Mantovani Filippo Contributed Talk
  Wednesday, June 3, 2015
  HG E1.1, 11:10-11:30
  
  Contributed Talk
  
  Scientific Computing Based on Mobile Embedded Technology, Filippo Mantovani (Barcelona Supercomputing Center, Spain)
  
  Co-Authors:
  
  In the late 1990s, mostly economic reasons led to the adoption of commodity processors in high-performance computing. This transformation has been so effective that in 2015 the TOP500 list is still dominated by x86-based supercomputers. In 2015, the largest commodity market is the one of smartphones and tablets, most of which are built with ARM-based SoCs. This leads to the suggestion that once mobile SoCs deliver sufficient performance, mobile SoCs can help reduce the cost of HPC. In view of the experiences within the Mont-Blanc project, this talk will describe possibilities, results and challenges raised when developing HPC platforms from mobile embedded technology.
• Marcolongo Aris Poster
  
  Poster
  
  PHY-03 Ab Initio Atomic Heat Transport Via Green-Kubo Formalism, Aris Marcolongo (EPFL, Switzerland)
  
  Co-Authors: Baroni Stefano (International School for Advanced Studies, Italy); Umari Paolo (Padua University, Italy)
  
  Green Kubo formulas have never been used in combination to a fully ab initio framework in order to compute the thermal conductivity of liquids. We derive here an expliticit expression for the adiabatic energy current, the key quantity in the Green-Kubo formalism, within density-functional theory and ready to be implemented in ab initio plane-wave based codes. The resulting methodology is demonstrated, as a proof of principle, by comparing ab initio and classical molecular simulations of a model liquid-Argon system, for which accurate inter-atomic potentials are derived by the force-matching method. A first application to liquid water is also presented.
• Markopoulos Alexandros MS Presentation
  Tuesday, June 2, 2015
  HG E21, 15:00-15:30
  
  MS Presentation
  
  Parallel Implementation of Hybrid FETI Solver with Communication Hiding Techniques, Alexandros Markopoulos (Technical University of Ostrava, Czech Republic)
  
  Co-Authors: Lubomir Riha (University of Ostrava, Czech Republic); Tomas Brzobohaty (University of Ostrava, Czech Republic); Tomas Kozubek (University of Ostrava, Czech Republic)
  
  We would like to present a massively parallel implementation of the Hybrid Total Finite Element Tearing and Interconnecting (FETI) method, which is designed to solve extremely large problems using multilevel decomposition combined with communication hiding techniques. In Hybrid FETI a relatively small number of neighboring subdomains is aggregated into clusters and each cluster is processed by a single compute node. This method significantly reduces main bottleneck of the two level FETI caused by solving the coarse problem. As of now the solver is able to solve problems larger than 4.3 billions of unknowns using 720 compute nodes with 64 GB of RAM.
• Marras Simone MS Presentation
  Tuesday, June 2, 2015
  HG E1.1, 12:00-12:30
  
  MS Presentation
  
  A Conservative and Grid Adaptive Stabilization Scheme for Spectral Elements Based on a Dynamic SGS Model for LES. Application in Numerical Weather Prediction, Simone Marras (Naval Postgraduate School, USA)
  
  Co-Authors: Francis X. Giraldo (Naval Postgraduate School, USA)
  
  The solution of the Euler equations by high-order spectral elements is prone to instabilities that must be damped in some way. We approach the problem of stabilization via an adaptive subgrid-scale scheme meant to treat the instabilities by modeling the sub-grid scale features of the flow. The equations are regularized via a dynamically adaptive stress proportional to the residual of the unperturbed equations. Its effect is close to none where the solution is smooth and it increases elsewhere. This is a first step toward LES for hurricanes and extreme weather forecast with the Nonhydrostatic Unified Model of the Atmosphere, NUMA, the dynamical core of the next-generation Navy model NEPTUNE.
  Minisymposium
  Minisymposium
  
  MS01 Advances in Numerics and Physical Modeling for Geophysical Fluid Dynamics
  
  Organizer: Simone Marras (Naval Postgraduate School, USA)
  
  Co-Organizers: Mariano Vázquez (Barcelona Supercomputing Center, Spain); Giovanni Tumolo (ICTP, Italy); Alex Breuer (Technische Universität München, Germany); Francis X. Giraldo (Naval Postgraduate School, USA)
  
  In view of the ever increasing parallelism of supercomputing platforms and the approaching era of exascale computing, high-resolution modeling is becoming a priority for Geophysical Fluid Dynamics (GFD) numerical modelers. Today's supercomputers require highly scalable numerical methods; high-resolution permits more resolved physical processes. This mini-symposium aims to bringing together scientists from different disciplines who are contributing to the next-generation GFD model development via the definition of advanced numerical methods and more accurate physical schemes. Interest is given to scalable numerics and high-resolution physics modeling such as, but not limited to, LES.
• Maruyama Naoya MS Presentation
  Monday, June 1, 2015
  HG F30, 13:30-14:00
  
  MS Presentation
  
  Domain-Specific Approaches in Scientific Computing, Naoya Maruyama (RIKEN, Japan)
  
  Co-Authors:
  
  Parallel programming with low-level interfaces has been the most viable choice in scientific computing for a long time. In such models, different parallelisms require different parallel programming interfaces, e.g., message passing for parallelism across nodes, threading for intra-node parallelism, and vector processing for SIMD and GPUs. Often applications are confronted with these multiple interfaces to fully exploit the current and future large-scale machines. We present our work toward higher-level programming models, allowing for a single program to run on different parallel platforms without much human intervention, and at the same time to achieve close to hand-tuned performance.
• Marzari Nicola MS Presentation
  Tuesday, June 2, 2015
  HG F30, 10:30-11:00
  
  MS Presentation
  
  The ADES Model for Computational Science, Nicola Marzari (EPFL, Switzerland)
  
  Co-Authors: Giovanni Pizzi (EPFL, Switzerland); Andrea Cepellotti (EPFL, Switzerland); Andrius Merkys (EPFL, Switzerland); Nicolas Mounet (EPFL, Switzerland); Riccardo Sabatini ( EPFL, Switzerland); Martin Uhrin (EPFL, Switzerland); Boris Kozinsky (Robert Bosch RTC, Cambridge MA, USA)
  
  Computational science has seen a meteoric rise in the scope, breadth, and depth of its efforts. Notwithstanding this prevalence and impact, it is often still performed using the renaissance model of individual artisans gathered in a workshop, under the guidance of an established practitioner. Great benefits could follow from adopting concepts and tools coming from computer science to manage, preserve, and share these computational efforts. I will illustrate here our vision for the four pillars that should sustain such effort (the ADES model: Automation, Data, Environment, and Sharing) and discuss their implementation in the open-source AiiDA platform (http://www.aiida.net).
• Mayer Lucio Minisymposium
  Minisymposium
  
  MS14 Radiative Transfer in Astrophysics; Methods and Applications to Multi-Scale Simulations
  
  Organizer: Lucio Mayer (University of Zurich, Switzerland)
  
  Co-Organizers: Romain Teyssier (University of Zurich, Switzerland)
  
  We plan to organize a Mini-Symposium focused on presenting and discussing methods and applications of radiative transfer in multi-scale simulations of astrophysical systems. The subject is timely as it reflects the main theme of the ongoing 'DIAPHANE' PASC project. Radiative transfer is the current bottleneck of multi-scale simulations across a wide range of scales and problems, from planet formation to galaxy formation and cosmological reionization. State-of-the-art methods employed in the major particle-based and grid-based codes worldwide will be presented and their range of application critically discussed. We will foster discussion of innovative algorithmic strategies.
• McRae Andrew MS Presentation
  Monday, June 1, 2015
  HG E1.1, 13:30-14:00
  
  MS Presentation
  
  Towards a 3D Dynamical Core Based on Mixed Finite Element Methods, Andrew McRae (Imperial College London, United Kingdom)
  
  Co-Authors: Colin Cotter (Imperial College London, United Kingdom); David Ham (Imperial College London, United Kingdom)
  
  There are known scaling issues with the otherwise-successful combination of C-grid finite difference methods on a global latitude-longitude grid. Direct extensions of C-grid methods to quasi-uniform grids have known inadequacies. Mixed finite element methods, recently unified under the label of finite element exterior calculus, provide a way of recovering desirable numerical properties but on arbitrary meshes. After some initial work in 2D, we implemented a method for the automated generation of sophisticated 'tensor-product' finite elements. The use of finite element methods also invites the possibility of high-order approaches, though we are not actively exploring this.
• Mehl Miriam MS Presentation
  Tuesday, June 2, 2015
  HG F3, 15:00-15:30
  
  MS Presentation
  
  ExaFSA -- Parallel Coupling of Structures, Turbulent Flow and Acoustics, Miriam Mehl (University of Stuttgart, Germany)
  
  Co-Authors: Florian Lindner (University of Stuttgart, Germany); Benjamin Uekermann (Technische Universität München, Germany)
  
  Fluid-Structure-Acoustics Interactions are one example for multiphysics simulations featuring a high model accuracy but also high computational costs which requires the use of massively parallel computers. Whenver we want to ensure flexibility in adding or exchanging physical fields and solver or set up a new model combination in a short timerange, such simulations use a partitioned approch combining independent existing single-physics solvers. We present functionalities, efficient numerics and parallelization strategies that we implemented in our coupling library preCICE comrising the complete functionality required to combine codes to a multiphysics simulation environment.
• Melie-Garcia Lester MS Presentation
  Tuesday, June 2, 2015
  HG E21, 10:30-11:00
  
  MS Presentation
  
  Bayesian Model for Decoding Organizational Principles of the Brain Anatomy, Lester Melie-Garcia (Centre hospitalier universitaire vaudois, Switzerland)
  
  Co-Authors: Ferath Kherif (Centre Hospitalier Universitaire Vaudois, Switzerland)
  
  A Bayesian Model is proposed as framework to decode the organizational principles of the brain anatomy. This problem aims at investigating how tissue properties ('trait') in the brain manifold predict 'states' represented by clinical variables. Our Bayesian Model deals with the problem of mapping thousands of multivariate variables to a small dimensional space having different sets of data informing the same model. We used the idea of 'Streaming Variational Bayes' approach based on different model evidences to tackle this difficulty. This framework is ideal when Neuroimaging data coming from different Hospitals are not allowed to be used due to confidentiality reasons.
• Merkys Andrius Poster
  
  Poster
  
  MAT-08 Developing Experimental & Theoretical Crystallography Open Databases, Andrius Merkys (EPFL, Switzerland)
  
  Co-Authors: Giovanni Pizzi (EPFL, Switzerland); Andrea Cepellotti (EPFL, Switzerland); Nicolas Mounet (EPFL, Switzerland); Saulius Grazulis (Vilnius University, Lithuania); Nicola Marzari (EPFL, Switzerland)
  
  The Crystallography Open Database (COD), launched as a grass-root initiative by an international group of scientists, has become the largest open-access resource to date for experimentally determined small-molecule crystal structures and is ready to be used as a source for large-scale automated analyses in various fields of computational chemistry, such as drug design and material research. A variety of data access and selection options, cross-links with other resources are made possible thanks to the open-access nature of the COD. Recently, a similar effort - the Theoretical COD - was launched alongside the COD, aimed to collect the results of atomistic simulations in an unified format.
• Métivier Ludovic MS Presentation
  Monday, June 1, 2015
  HG F3, 13:00-13:30
  
  MS Presentation
  
  Preconditioning Strategies for Multi-parameter Full Waveform Inversion, Ludovic Métivier (CNRS, France)
  
  Co-Authors: Romain Brossier (University of Grenoble Alpes, France); Stéphane Operto Géoazur (University of Sophia-Antipolis, France); Jean Virieux (University of Grenoble Alpes, France)
  
  Full Waveform Inversion (FWI) is a high-resolution seismic imaging method, based on the minimization of the misfit between observed and synthetic data. FWI is now routinely used in seismic exploration, as a velocity-building method. However, the formalism of FWI allows for the simultaneous reconstruction not only of P- and S-wave velocities, but also for parameters such as density, attenuation and anisotropy. This current limitation is due to trade-offs between parameters, making difficult to decipher between them only from seismic data. We propose to review and analyze the issues related to multi-parameter FWI, and propose a preconditioning strategy trying to mitigate these difficulties.
• Meuwly Markus MS Presentation
  Monday, June 1, 2015
  HG E3, 15:30-16:00
  
  MS Presentation
  
  Multi-Scale Exploration of Protein Interiors, Markus Meuwly (University of Basel, Switzerland)
  
  Co-Authors:
  
  Proteins exhibit dynamics on multiple length scales. This presentation will highlight computational approaches to characterize such motions ranging from fully atomistic simulations to network simulations. The focus will be on small ligand diffusion in globular proteins which is of physiological significance.
• Michelangeli Valentina MS Presentation
  Monday, June 1, 2015
  HG F30, 16:30-17:00
  
  MS Presentation
  
  Does it pay to get a Reverse Mortgage?, Valentina Michelangeli (Bank of Italy, Italy)
  
  Co-Authors:
  
  This paper uses data on single households from the Health and Retirement Study (HRS) to study the economic gains or losses associated with reverse mortgages. These data are examined within a dynamic structural life-cycle model featuring consumption, housing, and mobility decisions with uncertainty about both life span and mobility. I develop and apply new methods for solution and estimation based on a combination of four state-of-the-art mathematical programming tools. I find that reverse mortgages provide liquidity and a form of longevity insurance; however, moving becomes a risky proposition.
• Mikushin Dmitry Poster
  
  Poster
  
  PHY-04 Adopting CERN SixTrack Fortran Legacy Modeling Code to Perform Ensemble Simulations on GPU, Dmitry Mikushin (Università della Svizzera italiana, Switzerland)
  
  Co-Authors: Alexey Ivakhnenko (Applied Parallel Computing LLC); Igor Zacharov (EPFL / Eurotech, Switzerland); Eric McIntosh (CERN, Switzerland)
  
  In our talk we will present a reusable approach for poting 200K line Fortran codebase onto GPU: SixTrack model - the LHC Accelerator Beam Dynamics, a part of BOINC Distributed Volunteer Computing project at CERN. The model has been adapted to compile both with host compiler and CUDA Fortran. The problem of insufficient resource of parallelism is solved by performing ensemble simulations. GPU version executes most of the code on device, one block per simulation. Kernel intermixes single-threaded and parallel code portions implemented by postprocessing CUDA code at LLVM IR level. CPU and GPU performances are compared on standard SixDesk test set.
• Miller Stephen MS Presentation
  Monday, June 1, 2015
  HG E22, 14:30-15:00
  
  MS Presentation
  
  A New Volcanic Hydrothermal System in Java; the Lusi Mud Eruption and Aftermath, Stephen Miller (University of Neuchatel, Switzerland)
  
  Co-Authors: Maïté Faubert (University of Neuchatel, Switzerland); Reza Sohrabi (University of Neuchatel, Switzerland)
  
  In 2006 a magnitude 6.3 earthquake occurred in Yogyakarta on the island of Java. This relatively shallow earthquake resulted in 20,000 deaths and widespread destruction of the city. About 48 hours later, in the town of Sidoarjo about 250 km from the earthquake epicentre, mud began to spill out onto the surface and Lusi was born. Eruption rates peaked at 180,000 cubic meters of mud per day, and continues today as a vigorous geyser system, with geochemistry studies showing that is a geologic rarity of a newborn tectonic-scale hydrothermal system linked to the nearby volcano complex. We are modelling this system to gain insight of its workings and the possible exploitation of geothermal energy.
• Mindel Julian MS Presentation
  Monday, June 1, 2015
  HG E22, 14:00-14:30
  
  MS Presentation
  
  Challenges of Simulating Geothermal Reservoir Processes, Julian Mindel (ETH Zurich, Switzerland)
  
  Co-Authors: Thomas Driesner (ETH Zurich, Switzerland)
  
  Future geothermal power production in Switzerland requires circulating water through deep (~ 5km), fractured rock masses at temperatures exceeding 180°C. The extreme conditions massively limit direct, in situ observation and steering of reservoir processes, making numerical simulation a key research and development technology. The biggest challenges are a combination of complex but poorly know geometries (thin, irregular fracture networks in a large rock mass) and strongly coupled, non-linear processes (thermo-hydro-mechanical-chemical interactions). This contribution outlines the goals and challenges of numerical code development in the SCCER-SoE and related initiatives.
• Minion Michael MS Presentation
  Tuesday, June 2, 2015
  HG E22, 11:30-12:00
  
  MS Presentation
  
  High Order Temporal Integration Methods for Schroedinger Equations, Michael Minion (Lawrence Berkeley National Laboratory, USA)
  
  Co-Authors: Jeffrey Banks (Rensselaer Polytechnic Institute, USA)
  
  I will report on the construction of a temporal integration scheme designed to be used with a higher-order finite-difference spatial approximation to the nonlinear Schroedinger equation with perfectly matched layer boundary conditions for the paraxial wave equation. The temporal method is based on a semi-implicit spectral deferred corrections approach coupled with an alternating direction implicit (ADI) approximation to the (linear) implicit terms. The accuracy and efficiency of the methods, as well as a discussion of space-time parallelization of the method will be presented.
  Minisymposium
  Minisymposium
  
  MS13 Increasing Computational Efficiency for Temporal Integration
  
  Organizer: Michael Minion (Lawrence Berkeley National Laboratory, USA)
  
  Co-Organizers: Daniel Ruprecht (Università della Svizzera italiana, Switzerland); Marcus Grote (University of Basel, Switzerland); Rolf Krause (Università della Svizzera italiana, Switzerland)
  
  The temporal integration of ordinary or partial differential equations is a fundamental computational problem in many application areas such as molecular dynamics, weather and climate, structural mechanics, combustion, cosmology, and many others. This minisymposium will present innovative strategies for realizing better computational efficiency in cases where the temporal direction creates a computational bottleneck. Topics include exploiting disparate time scales in the governing equations, applying techniques from multigrid in the time direction, and using time parallel algorithms like parareal and PFASST.
• Moody Sandie Poster
  
  Poster
  
  CSM-03 Discrete Duality Finite Volume (DDFV) Method Applied to COSMO Horizontal Diffusion, Sandie Moody (University of Geneva, Switzerland)
  
  Co-Authors: Martin Jakob Gander (University of Geneva, Switzerland); Oliver Fuhrer (MeteoSwiss, Switzerland)
  
  At present, the horizontal components of the subgrid scale flux divergence of the averaged equation for mass of water constituents are not being calculated in the COSMO-Model. This is due to the discretization, whos stability is limited. We propose a new approach, namely a coupling of the DDFV and finite volume methods, to be implemented in the COSMO-Model. Our results show that the DDFV method is well adapted to the COSMO-Model as it is stable on any grid type, in particular a terrain-following grid including steep slopes. We also analyse computational costs and convergence rates.
• Moragues Ginard Margarida MS Presentation
  Monday, June 1, 2015
  HG E1.1, 17:00-17:30
  
  MS Presentation
  
  Preconditioned Variational Multiscale Stabilization (PVMS) for Low Mach Flows, Margarida Moragues Ginard (Barcelona Supercomputing Center, Spain)
  
  Co-Authors: Guillaume Houzeaux (Barcelona Supercomputing Center, Spain); Mariano Vázquez (Barcelona Supercomputing Center, Spain)
  
  In this work we present an implicit scheme for solving the Navier-Stokes equations at Low Mach regimes. The method can cope efficiently with both stiffness and numerical instabilities thanks to the following procedure. A local preconditioner is applied to the Navier-Stokes equations. Next, the preconditioned local system is discretized according to the Variational Multiscale Stabilization (VMS) method. Finally, a monolithic implicit Jacobi (i.e. fixed point) scheme is applied, which, in turn, uses an algebraic diagonal preconditioner on a GMRES iterative solver. The scheme is assessed on large-scale 3D problems, particularly on its accuracy, convergence and parallel efficiency.

N

• Nakajima Kengo MS Presentation
  Monday, June 1, 2015
  HG F30, 13:00-13:30
  
  MS Presentation
  
  ppOpen-HPC: Open Source Infrastructure for Development and Execution of Large-Scale Scientific Applications on Post-Peta Scale Supercomputers with Automatic Tuning (AT), Kengo Nakajima (The University of Tokyo, Japan)
  
  Co-Authors: Takahiro Katagiri (University of Tokyo, Japan)
  
  In this presentation, recent achievements and progress of the "ppOpen-HPC" project are presented. ppOpen-HPC is an open source infrastructure for development and execution of optimized and reliable simulation code on post-peta-scale (pp) parallel computers based on many-core architectures with automatic tuning (AT), and it consists of various types of libraries, which cover general procedures for scientific computation. An example of automatic tuning by ppOpen-AT on 3D FDM code of seismic simulations (Seism3D) will be presented. Moreover, recent achievements in the development of ppOpen-MATH/MG, which is a geometric multigrid solver in ppOpen-HPC, will be also presented.
  Minisymposium
  Minisymposium
  
  MS08 Scientific Libraries/Frameworks in Japan for Future HPC Systems
  
  Organizer: Kengo Nakajima (The University of Tokyo, Japan)
  
  Co-Organizers: Ryuji Shioya (Toyo University, Japan)
  
  Reliable software is the key technology for future HPC systems. In 2010, 'Development of System Software Technologies for Post-Peta Scale HPC (Post-Peta CREST)' was initiated by the Japan Science & Technology Agency. It aims at developing software exploiting maximum efficiency and reliability on future HPC systems in late 2010s. It includes 14 projects, which cover a wide range of research areas in software for HPC, such as system software, programming languages, compilers, numerical libraries, and application frameworks. This minisymposium offers an overview on recent activities of four projects in Post-Peta CREST related to scientific computing, and discusses future international/interdisciplinary collaborations.
• Nazarov Murtazo MS Presentation
  Monday, June 1, 2015
  HG E1.1, 16:00-16:30
  
  MS Presentation
  
  Nonlinear Stabilization Techniques for Finite Element Approximations of Fluid Problems, Murtazo Nazarov (Uppsala University, Sweden)
  
  Co-Authors:
  
  This talk will discuss recent developments in stabilized finite element approximations using nonlinear viscosity methods. The regularization terms are constructed by the residual of the system or entropy equations. The method is successfully applied to compressible and incompressible Navier-Stokes equations and variable density flows in two and three space dimensions. Then we discuss about maximum principle preserving continuous finite element schemes for scalar conservation laws. The new method does not require any a priori knowledge of quantities like local wave-speed, proportionality constant, or local mesh-size and it is independent of the cell type.
• Novo David MS Presentation
  Tuesday, June 2, 2015
  HG E1.1, 15:00-15:30
  
  MS Presentation
  
  Reconfigurable Computing for All: Efficient Use of FPGAs for Non-Hardware Experts, David Novo (EPFL, Switzerland)
  
  Co-Authors: Nithin George (EPFL, Switzerland); Mohsen Owaida (EPFL, Switzerland); Paolo Ienne (EPFL, Switzerland)
  
  Field-Programmable Gate Arrays (FPGAs) can achieve significant acceleration over traditional processor-based solutions. Unfortunately, their inaccessible programming model, which is inherited from circuit design and far from software-like productivity, is probably the main reason for many users to stay away from FPGAs. In this talk, I will present a cost-effective approach to develop domain-specific high-level synthesis tools. In particular, we leverage domain-specific languages (i.e., designed to specify applications within a particular domain, e.g., machine learning) to express the application at a high abstraction level while enabling the automatic generation of efficient hardware.

O

• Obrist Dominik MS Presentation
  Monday, June 1, 2015
  HG E21, 14:30-15:00
  
  MS Presentation
  
  HPC Framework for Aortic Valve Simulation with Hybrid Discretization for Fluid and Soft Tissue, Dominik Obrist (University of Bern, Switzerland)
  
  Co-Authors: Barna Becsek (University of Bern, Switzerland)
  
  The numerical simulation of aortic valves is a multi-physics problem involving large deformations of soft tissue and transient vortical flow fields. Whereas soft tissue is most appropriately discretized on unstructured meshes in Lagrangian formulation, the three-dimensional flow field is discretized on a structured Cartesian grid to obtain an efficient implementation on modern HPC platforms. The tissue dynamics on the unstructured mesh and the flow on the structured grid are coupled with the immersed boundary method. The parallelization of such a hybrid discretization approach raises interesting questions with respect to data locality and load balancing under a domain decomposition paradigm.
  Minisymposium
  Minisymposium
  
  MS04 Fluid-Structure Interaction in the Cardiovascular System
  
  Organizer: Dominik Obrist (University of Bern, Switzerland)
  
  Co-Organizers: Simone Deparis (EPFL, Switzerland)
  
  The mechanical interaction between blood and soft tissue is a central mechanism in the cardiovascular system (e.g. heart valve dynamics, myocardial motion). A detailed understanding of these processes is the basis for the development of diagnostic methods, therapeutic devices and prostheses. To this end, large fluid-structure problems have to be solved accurately and efficiently on high-performance computing platforms with sophisticated numerical methods and with parallelization paradigms, which comply with large structural deformations and a range of inherent time scales. This minisymposium presents the work of leading groups in this field.
• Omlin Samuel Poster
  
  Poster
  
  CSM-07 HPC.m - the MATLAB HPC Compiler and its Use for Solving 3D Poromechanics on Supercomputers, Samuel Omlin (University of Lausanne, Switzerland)
  
  Co-Authors: Ludovic Räss (University of Lausanne, Switzerland); Yuri Podladchikov (University of Lausanne, Switzerland);
  
  We present a pre-release version of HPC.m - the MATLAB HPC Compiler. HPC.m transforms simple MATLAB scripts in a few seconds into massively parallel near peak performance applications for CPU-, GPU- and MIC-supercomputers. The MATLAB scripts must employ basic syntax and follow a few simple rules. We have successively deployed this software in the development of supercomputing applications for the Piz Daint at CSCS in Lugano, Switzerland. We have shown real-world applications work, as for example a non-linear poro-visco-elasto-plastic 3D solver. Our code achieves near peak performance and scales linearly up to more than 2000 GPUs allowing for a spatial resolution of over 2000^3 grid points.
• Osuna Escamilla Carlos MS Presentation
  Monday, June 1, 2015
  HG E1.2, 17:00-17:30
  
  MS Presentation
  
  STELLA: Optimization of the COSMO Dynamical Core for Heterogeneous Platforms, Carlos Osuna Escamilla (ETH Zurich, Switzerland)
  
  Co-Authors: Tobias Gysi (ETH Zurich, Switzerland); Oliver Fuhrer (MeteoSwiss, Switzerland); Thomas C. Schulthess (CSCS / ETH Zurich, Switzerland / Oak Ridge National Laboratory, USA)
  
  Regional weather and climate models typically solve Partial Differential Equations (PDEs) equations using compact stencils on structured grids. STELLA (Stencil Loop Language) is a DSL written in C++ for such stencil computations. It abstracts the underlying programming models and optimization techniques (loop tiling, loop fusion) for multiple hardware architectures. The STELLA language allows to describe stencils with a unique source code in a concise way, close to the mathematical description of the PDEs. We present the STELLA syntax and the performance results achieved for the dynamical core of an important community code (COSMO) for GPU (5.8x speedup) and CPU (1.8x speedup).

P

• Pakrouski Kiryl Poster
  
  Poster
  
  PHY-11 Topological Quantum Computation and Fractional Quantum Hall: Phase Diagram of the 5/2 State, Kiryl Pakrouski (ETH Zurich, Switzerland)
  
  Co-Authors: Matthias Troyer (ETH Zurich, Switzerland); Michael R. Peterson (California State University Long Beach, USA); Thierry Jolicoeur (CNRS, France / University Paris-Sud, France); Vito W. Scarola (Virginia Tech, USA); Chetan Nayak (University of California, USA / Microsoft Research, USA);
  
  Interesting non-Abelian states, e.g., the Moore-Read Pfaffian and the anti-Pfaffian, offer candidate descriptions of the nu= 5/2 fractional quantum Hall state. To study the phase diagram of the nu= 5/2 state we numerically diagonalize a comprehensive effective Hamiltonian describing the fractional quantum Hall effect of electrons under realistic conditions in GaAs semiconductors. We conclude that the ground state is in the universality class of the Moore-Read Pfaffian state, rather than the anti-Pfaffian for small Landau level mixing strength. Our findings have important implications for the identification of non-Abelian fractional quantum Hall states.
• Parmigiani Andrea Poster
  
  Poster
  
  EAR-05 From Capillary to Bubbly Flow: the Fate of Low Reynolds Number, Buoyancy Driven Fluids Transport at Strong Porosity Transition, Andrea Parmigiani (ETH Zurich, Switzerland)
  
  Co-Authors: Salah Faroughi (GeorgiaTech, USA); Christian Huber (GergiaTech, USA); Olivier Bachmann (University of Zurich, Switzerland)
  
  Natural porous-media, as crystal-rich magma chambers, can display change in porosities (e.g. layered/stratified environment generated by settling and compaction of solid particles). In this poster, we investigate the fate of a buoyant, low-viscosity and non-wetting fluid injected in a porous medium saturated with a more viscous fluid that opens in a solid-free environment. We argue for this existence of a counter-intuitive process whereby the upward migration of the buoyant fluid is more efficient in low-permeability region (i.e. the porous medium). We investigate this scenario using pore-scale numerical calculations based on the lattice Boltzmann technique and laboratory experiments.
• Patmanidis Ilias Poster
  
  Poster
  
  LS-11 Structural Insights into Phosphoinositide 3-kinase Alpha (PI3Ks) Regulation using Molecular Dynamics Simulations, Ilias Patmanidis (University of Geneva, Switzerland)
  
  Co-Authors: Chiriano Gianpaolo (University of Geneva, Switzerland); Vadas Oscar (University of Geneva, Switzerland); Scapozza Leonardo (University of Geneva, Switzerland);
  
  Phosphoinositide 3-kinases (PI3Ks) are cytosolic lipid kinases phosphorylating the 3 inositol ring of inositol-phospholipids (PtdIns) which act as lipid second messengers regulating key cellular functions. We focused on the isoform of class I PI3Ks and different mutants (G106V, N345K, E545K, H1047R) associated with cancer and increased enzymatic activity. Our work aimed at understanding the dynamical behavior of PI3Ks by Molecular Dynamics (MD). The MD results shed light on the conformational changes and fluctuations upon activation of WT PI3Ks and mutants, complementing the experimental observations by X-ray crystallography and Hydrogen/Deuterium exchange coupled to mass spectrometry.
• Pedersen Andreas Contributed Talk
  Wednesday, June 3, 2015
  HG E1.2, 10:50-11:10
  
  Contributed Talk
  
  Lithiation / Delithiation of Tin-Oxide, Andreas Pedersen (ETH Zurich, Switzerland)
  
  Co-Authors: Petr A. Khomyakov (ETH Zurich, Switzerland); Mathieu Luisier (ETH Zurich, Switzerland)
  
  Based on first-principles calculations, we propose a microscopic model that explains the reversible lithiation/delithiation of tin-oxide anodes in Li-ion batteries. At the point when the irreversible regime ends, the anode is ordered and consist of layers of Li-oxide separated by Sn bilayers. During the following reversible lithiation, the Li-oxide undergoes two phase transformations, which give rise to a Li-enrichment of the oxide and the formation of a SnLi composite. The predicted anode volume expansion and voltage profile agree well with experiments, contrary to existing models.
• Peitsch Manuel Minisymposium
  Minisymposium
  
  MS05 HPC in Systems Biology
  
  Organizer: Manuel Peitsch (Philip Morris Int. & SIB, Switzerland)
  
  Co-Organizers:
  
  Systems Biology is a highly multidisciplinary approach to decoding life that combines computational with experimental methods to elucidate biological mechanisms and their dynamic behavior at all levels of biological organization. Systems Biology enables new approaches in pharmacology, toxicology, diagnostics, as well as drug development. The presentations will exemplify how computational Systems Biology can enable the understanding of biological systems both at the genomic and at the structural level, how mechanistic understanding is likely to change product risk assessment and how HPC will turn Big Data into knowledge about biological systems.
• Peter Daniel Poster
  
  Poster
  
  EAR-04 Forward and Adjoint Spectral-Element Simulations of Seismic Wave Propagation using Hardware Accelerators, Daniel Peter (Università della Svizzera italiana & ETH Zurich, Switzerland)
  
  Co-Authors: Brice Videau (University of Grenoble, France); Kevin Pouget (University of Grenoble, France); Dimitri Komatitsch (University of Aix-Marseille, France)
  
  Recent advances in regional and global-scale seismic inversions move towards full-waveform inversions which require accurate simulations of seismic wave propagation in complex 3D media, providing access to the full 3D seismic wavefields. We incorporate a code generation tool BOAST into an existing spectral-element code package to use meta-programming of computational kernels and generate optimized source code for both CUDA and OpenCL hardware accelerators. We show here applications of forward and adjoint seismic wave propagation on CUDA/OpenCL GPUs, validating results and comparing performances for different simulations and hardware usages.
• Peter Emanuel Karl Poster
  
  Poster
  
  PHY-08 Polarizable Coarse-Grained Water and Protein Models for Dissipative Particle Dynamics, Emanuel Karl Peter (Università della Svizzera italiana, Switzerland)
  
  Co-Authors: Kirill Lykov (Università della Svizzera italiana, Switzerland); Igor V. Pivkin (Università della Svizzera italiana, Switzerland)
  
  Polarizability effects play a major role in many biological processes from effects at lipid bilayers to protein folding. We developed a new method for modeling of polarizability effects in Dissipative Particle Dynamics (DPD). Based on a Drude oscillator model, we first implemented a new coarse-grained water model and applied it to simulations of water-membrane systems. Following that, we developed a model for simulations of proteins, which is based on the polarization of the protein backbone and a simplified representation of the sidechains. We believe that the new models can significantly expand the application range of DPD into the field of computational biophysics.
• Petra Cosmin MS Presentation
  Monday, June 1, 2015
  HG E21, 16:00-16:30
  
  MS Presentation
  
  StochJuMP - Parallel Algebraic Modeling for Stochastic Optimization in Julia, Cosmin Petra (Argonne National Lab, USA)
  
  Co-Authors: Joey Huchette (Massachusetts Institute of Technology, USA); Miles Lubin (Massachusetts Institute of Technology, USA)
  
  We present scalable algebraic modeling software, StochJuMP, for stochastic optimization as applied to power grid economic dispatch. It enables the user to express the problem in a high-level algebraic format with minimal boiler-plate and allows efficient parallel model instantiation across nodes and efficient data localization. Computational results on up to 2048 cores of the Blues cluster of Argonne National Laboraratory are presented showing that the model construction is both efficient and scalable. StochJuMP is configured with the parallel interior-point solver PIPS-IPM but is sufficiently generic to allow straightforward adaptation to other solvers.
• Pflüger Dirk MS Presentation
  Tuesday, June 2, 2015
  HG F3, 14:30-15:00
  
  MS Presentation
  
  Algorithm-Based Fault Tolerance with the Plasma Physics Code GENE, Dirk Pflüger (University of Stuttgart, Germany)
  
  Co-Authors: Alfredo Parra Hinojosa (Technische Universität München, Germany); Mario Heene (University of Stuttgart, Germany)
  
  Massively parallel simulation codes will have to be able to handle faults in the coming age of exascale computing. Faults are predicted to occur in the time-span of minutes or even below. We employ the combination technique for the simulation of hot fusion plasmas. It numerically decomposes a single and huge problem into many, but small partial problems that can be computed in parallel. This hierarchical splitting provides a new handle to treat faults without checkpoint-restart at the cost of sligthly slower convergence or slightly higher error. Furthermore, our algorithm-based fault tolerance does not depend on expensive recomputations of missing solutions.
• Pintarelli Simon Poster
  
  Poster
  
  CSM-09 Parallel Solver for the Space Inhomogeneous and Time Dependent Boltzmann Equation, Simon Pintarelli (ETH Zurich, Switzerland)
  
  Co-Authors: Philipp Grohs (ETH Zurich, Switzerland); Ralf Hiptmair (ETH Zurich, Switzerland)
  
  We present a high-performance implementation for the solution of the space inhomogeneous and time dependent Boltzmann equation. The phase space is discretized using finite elements for the physical domain and a polar spectral discretization based on Laguerre polynomials in velocity. Computations are done in 2+2+1 dimensions with an implicit/explicit split time stepping scheme on unstructured meshes. The polar spectral scheme requires no truncation in velocity and conserves mass, momentum and energy. Dirichlet type boundary conditions are included into a least squares formulation. Results for supersonic gas flows in complicated geometries will be presented.
• Piotrowski Zbigniew Poster
  
  Poster
  
  CLI-01 Alternate Direction Implicit Preconditioning for Geophysical Applications on CPU and GPU, Zbigniew Piotrowski (Institute of Meteorology and Water Management, Poland)
  
  Co-Authors: Milosz Ciznicki (Poznan Supercomputing and Networking Center, Poland)
  
  Multiscale solvers for geophysical flows often utilise implicit solvers, which demand efficient preconditioning. Alternate Direction Implicit preconditioners can improve flexibility and efficiency of the iterative solvers for modern atmospheric research, despite relying on tridiagonal solvers. In this presentation, utility of ADI preconditioning for global, mesoscale and local flows will be discussed and illustrated with computational examples. Implementation and performance of the preconditioned GCR solver of EULAG model on CPU and GPU will be presented.
• Piquemal Jean-Philip MS Presentation
  Monday, June 1, 2015
  HG E3, 13:30-14:00
  
  MS Presentation
  
  High Performance Polarizable Molecular Dynamics, Jean-Philip Piquemal (Université Pierre et Marie Curie, France)
  
  Co-Authors:
  
  In this talk, I will present new advances in the development of scalable polarizable force fields for extensive molecular dynamics simulations. The discussed techniques are grounded on solid applied mathematic techniques enabling gain ranging from to 2 to 3 orders in magnitude in time thanks to parallelism. Potential applications will be discussed.
• Plamada Andrei Contributed Talk
  Wednesday, June 3, 2015
  HG F1, 10:30-10:50
  
  Contributed Talk
  
  Double Counting Problem in DFT+DMFT Method: A Study in the Exact Limit, Andrei Plamada (ETH Zurich, Switzerland)
  
  Co-Authors: Peter Staar (IBM Research, Switzerland); Anton Kozhevnikov (CSCS / ETH Zurich, Switzerland); Thomas Schulthess (CSCS / ETH Zurich, Switzerland)
  
  The combination of density functional theory (DFT) with dynamical mean field (DMFT) has become a popular method for first principles studies of strongly correlated materials. Yet, one of the most severe drawbacks of the method is the double counting of correlations that have been included in both, DFT and DMFT, and that so far could be removed only by introducing an empirical parameter. Assuming that DFT and DMFT produce the exact density, we devise a method to derive the double counting correction from first principles. We validate it on metals, charge transfer and Mott insulators, and find that in all cases the predicted correction to be in agreement with the accepted values.
• Plessl Christian MS Presentation
  Tuesday, June 2, 2015
  HG E1.1, 14:30-15:00
  
  MS Presentation
  
  Custom Data-Flow Engines for Computational Nanophotonics, Christian Plessl (University of Paderborn, Germany)
  
  Co-Authors:
  
  Finite difference methods are widely used, highly parallel algorithms for solving differential equations. While simple from a computational point of view, these algorithms are hard to implement efficiently because their performance is essentially memory-bound. In this talk we discuss how the finite difference time domain (FDTD) method for solving Maxwell's equations can be translated to custom data-flow engines implemented on FPGAs. We evaluate our work with a case study from the computational nanophotonics domain and show that the resulting solution provides both, performance and energy-efficiency benefits over CPU and GPU implementations.
  Minisymposium
  Minisymposium
  
  MS19 Accelerating Scientific Computing with FPGAs
  
  Organizer: Christian Plessl (University of Paderborn, Germany)
  
  Co-Organizers: Heiner Giefers (IBM Research, Switzerland)
  
  Since the performance of CPUs started to stagnate, computing with field-programmable gate arrays (FPGAs) has attracted significant interest. FPGAs not only provide massive parallelism and but also allow for fully customizing the processing architecture to the application through a software programming process. As a result, FPGAs can outperform CPUs in speed and energy-efficiency by orders of magnitude for many applications. In this symposium we will bring together leading experts in FPGA computing with computational scientists with a selection of talks that elaborate on the foundations, programming approaches and case studies of the applications of FPGAs for scientific computing.
• Potthast Roland MS Presentation
  Monday, June 1, 2015
  HG F3, 17:00-17:30
  
  MS Presentation
  
  On Ensemble and Particle Filters for Large-Scale Data Assimilation, Roland Potthast (Deutscher Wetterdienst, Germany)
  
  Co-Authors: Andreas Rhodin (Deutscher Wetterdienst, Germany); Christoph Schraff (Deutscher Wetterdienst, Germany); Hendrik Reich (Deutscher Wetterdienst, Germany)
  
  Ensemble data assimilation techniques are of rapidly growing importance. Ensemble techniques allow to describe and forecast uncertainty of the analysis, but they also improve the assimilation result itself, by allowing estimates of the covariance or, more general, the prior and posterior probability distribution of atmospheric states. In our talk, we will first give a survey about recent activities of the German Meteorological Service DWD. Then, we present recent work on the further development of the ensemble data assimilation towards a particle filter for large-scale atmospheric systems, which keeps the advantages of the LETKF, but overcomes some of its limitations.
• Proehl Elisabeth Poster
  
  Poster
  
  EMD-01 Computing Stationary Markov Equilibria with Heterogeneity, Elisabeth Proehl (University of Geneva, Switzerland)
  
  Co-Authors:
  
  Dynamic stochastic general equilibrium models with heterogeneous agents and incomplete markets have to be solved numerically. Existing algorithms approximate the law of motion of aggregate variables parametrically using a limited number of moments of the cross-sectional distribution. They do not take full advantage of the stationary state distribution. In this paper, a computable expression for the exact law of motion is introduced which leads to the equilibrium's stationary state distribution. The algorithm, a hybrid of projection and perturbation methods, is shown to converge. Furthermore, approximation and convergence issues of some existing algorithms are rationalized.

R

• Rahmati Alireza MS Presentation
  Tuesday, June 2, 2015
  HG E3, 14:00-14:30
  
  MS Presentation
  
  TRAPHIC: an Efficient Radiative Transfer Method for SPH Simulations with Many Sources, Alireza Rahmati (University of Zurich, Switzerland)
  
  Co-Authors:
  
  I will talk about the radiative transfer (RT) code TRAPHIC (TRAnsport of PHotons In Cones). TRAPHIC is a multi-frequency RT code for use in Smoothed Particle Hydrodynamics (SPH) simulations. It solves the time-dependent RT equation by tracing photon packets emitted by source particles at the speed of light and in a photon-conserving manner through the simulation box. While the unique design of TRAPHIC makes it ideal for cosmological simulations with large number of sources, it is a general purpose ionizing RT code, and as I will show with several examples, has already been successfully employed in a range of problems.
• Reed Darren Poster
  
  Poster
  
  PHY-06 DIAPHANE: Building a Library for Radiation and Neutrino Transport in Hydrodynamic Simulations, Darren Reed (University of Zurich, Switzerland)
  
  Co-Authors: Lucio Mayer (University of Zurich, Switzerland); Cladio Gheller (ETH Zurich / CSCS, Switzerland); Ruben Cabezon (University of Basel, Switzerland); Ali Rahmati (University of Zurich, Switzerland)
  
  A central task of the 'DIAPHANE' PASC project is to develop a library of radiation and neutrino transport routines for astrophysical and cosmological simulations. Applications span a wide range: supernovae, the formation of planets, black holes, stars, and galaxies, and many other problems where progress has been limited by the importance of efficient modeling of radiation or neutrino transport. The library includes algorithms for both optically thick and optically thin gas allowing for 'adaptive' radiative transfer modelling. I demonstrate its current functionality and discuss some of the challenges and strategies in making the library portable enough for multiple codes.
• Reumann Matthias MS Presentation
  Monday, June 1, 2015
  HG E1.2, 13:30-14:00
  
  MS Presentation
  
  Big Data in Life Science: Harnessing Supercomputing to Make Big Data Accessible and Actionable, Matthias Reumann (IBM Research, Switzerland)
  
  Co-Authors:
  
  Big data in life science is experiencing the perfect storm: The volume is increasing exponentially with accelerating speed, the variety of data ranges from multi-omics information to lifestyle measures with the help of mobile devices backed by cloud infrastructures. State of the art analytical methods are generally limited by computational approaches. The researcher is forced to limit their investigation to selected data rather than exhaustively mining the data. We have demonstrated that supercomputing methodologies enables discovery beyond traditional methods. Add cognitive computing to supercomputing and the perfect storm can be conquered.
• Rietmann Max MS Presentation
  Tuesday, June 2, 2015
  HG E22, 12:00-12:30
  
  MS Presentation
  
  Load-Balanced Local Time Stepping for Large-Scale Wave Propagation, Max Rietmann (ETH Zurich, Switzerland)
  
  Co-Authors: Marcus Grote (University of Basel, Switzerland); Daniel Peter (Università della Svizzera italiana, Switzerland); Bora Ucar (INRIA, France); Olaf Schenk (Università della Svizzera italiana, Switzerland)
  
  Applications using finite element meshes often require regions of refinement to honor small-scale features. These localized smaller elements create a bottleneck for explicit time-stepping schemes due to the Courant-Friedrichs-Lewy stability condition. Recently developed local time stepping (LTS) algorithms reduce the impact of these small elements by locally adapting the time-step size to the size of the element. The multi-level nature of LTS introduces a strong load imbalance across processors. We examine the use of multi-constraint graph and hypergraph partitioning tools to achieve effective, load-balanced parallelization for CPUs and GPUs implemented in the seismology package SPECFEM3D.
• Rigazzi Alessandro MS Presentation
  Monday, June 1, 2015
  HG E22, 16:00-16:30
  
  MS Presentation
  
  Prediction of Elastostatic Friction for Rock-Like Surfaces with FEM, Alessandro Rigazzi (Università della Svizzera italiana, Switzerland)
  
  Co-Authors: Rolf Krause (Università della Svizzera italiana, Switzerland)
  
  We present the results of our recent work, in which we apply the Finite Element Method to study the frictional forces resulting from the elastostatic interactions of surface asperities. To do this, we generate several self-affine rough surfaces that are statistically similar to rock surfaces. Such surfaces are used as rigid obstacles onto which an elastic smooth cube is pushed and sheared to measure the maximal resistance opposed by the asperities. The contact problem gives rise to a huge nonsmooth system of equations. We solve this system with a solver based on our nonsmooth multigrid algorithm of optimal complexity for which we investigate scaling properties.
• Righi Marcello Contributed Talk
  Wednesday, June 3, 2015
  HG F3, 10:30-10:50
  
  Contributed Talk
  
  High-fidelity Aeroelasticity, Marcello Righi (Zurich University of Applied Sciences, Switzerland)
  
  Co-Authors: Jan Koch (Zurich University of Applied Sciences, Switzerland)
  
  Aeroelasticity considers the joint effects of aerodynamic, elastic and inertial forces. This discipline initially relied on analytical methods to model the underlying physics. Later on, low-fidelity methods such as doublet-lattice became the standard. High-fidelity, CFD-based approaches are of course possible but demanding, as, unlike most engineering disciplines, they require time-accurate simulations, often used to feed convenient Reduced-Order-Models. Availability of HPC might soon allow practitioners to systematically capture flow non-linearities in full. We would like to present the state of art and the current trends.
• Robert Sylvain Poster
  
  Poster
  
  CLI-03 Ensemble Kalman Particle Filter for Convective Scale Data Assimilation, Sylvain Robert (ETH Zurich, Switzerland)
  
  Co-Authors: Hans Rudolph Künsch (ETH Zurich, Switzerland)
  
  The Local Ensemble Transform Kalman Filter (LETKF) is the state of the art algorithm for data assimilation in weather forecasting. However, because it implicitly assumes the background distribution to be Gaussian, there is potential for improvement on the convective scale, where non-Gaussian features are present. We propose new algorithms which combine the LETKF with the Particle Filter in a way that maintains scalability and sample diversity. We demonstrate their performance in a simplified model of cloud convection and show how they can be adapted to work with the COSMO model. The latter is an ongoing cooperation with MeteoSchweiz and Deutscher Wetterdienst.
• Rosdahl Joakim MS Presentation
  Tuesday, June 2, 2015
  HG E3, 14:30-15:00
  
  MS Presentation
  
  Galaxy Evolution with Radiation-Hydrodynamics in RAMSES-RT, Joakim Rosdahl (Leiden University, Netherlands)
  
  Co-Authors:
  
  I present an implementation of radiation-hydrodynamics in the RAMSES code, which models the emission and propagation of radiation and its interaction with gas via ionisation, dissociation, heating, dust-scattering, and momentum transfer. We use moment-based radiative transfer, which has the advantage over ray-based implementations of handling an unlimited number of radiation sources without an effect on the computational load. I discuss the basic aspects of the implementation, highlighting advantages and disadvantages, and show examples where it has been used in studies of radiation feedback, the emission properties of the circum-galactic medium, and the escape of UV radiation from galaxies.
• Rossinelli Diego Poster
  
  Poster
  
  LS-08 Needle in a Bloody Haystack, Diego Rossinelli (ETH Zurich, Switzerland)
  
  Co-Authors: Dmitry Alexeev (ETH Zurich, Switzerland); Kirill Lykov (Università della Svizzera italiana, Switzerland); Igor Pivkin (Università della Svizzera italiana, Switzerland); Petros Koumoutsakos (ETH Zurich, Switzerland)
  
  We present petascale simulations of microfluidics involving 100,000 of Red Blood Cells (RBCs) and a Circulating Tumor Cell (CTC) in a complex microscale  geometry. The simulations assess the effectiveness of the channel geometry in isolating a single CTC in a 'haystack' of RBCs. The simulation software harnesses the compute power of K20X-accelerated supercomputers reaching up to 60% of their nominal peak in terms of Instructions-Per-Cycle over 10,000s of compute nodes. With the present software we aim to unravel the key requirements for designing effective microfluidics devices and arm manufacturers and cancer researchers with a ground breaking in-silico prototyping tool.
• Ruprecht Daniel Poster
  
  Poster
  
  CSM-05 Energy Efficiency of Parareal, Daniel Ruprecht (Università della Svizzera italiana, Switzerland)
  
  Co-Authors: Andrea Arteaga (ETH Zurich, Switzerland); Rolf Krause (Università della Svizzera italiana, Switzerland)
  
  Parallel-in-time methods have been shown to provide a promising approach to additional concurrency for solving time-dependent problems. However, little is known about their performance in metrics beside speedup. In the future, energy-to-solution for example will become one of the key aspects to judge a method's performance on an HPC system. The poster presents results investigating the Parareal method with respect to its energy efficiency using different parallelization paradigms. A simple theoretical model for the expected energy efficiency is compared against measurements to distinguish between overhead from Parareal's intrinsic limit on parallel efficiency and other sources.
• Ruthotto Lars MS Presentation
  Monday, June 1, 2015
  HG E21, 17:00-17:30
  
  MS Presentation
  
  Distributed and Parallel Algorithms for PDE Constrained Optimization in Julia, Lars Ruthotto (Emory University, USA)
  
  Co-Authors: Eldad Haber (University of British Columbia, Canada)
  
  This talk presents a Julia framework for the solution of large-scale PDE constrained optimization problems. It is based on a discretize-then-optimize approach, uses a (projected) Gauss-Newton method, and provides interfaces state-of-the-art linear solvers (both explicit and iterative). The framework uses Julia's potential for parallel and distributed computation. Being written in a dynamic language, it is easily extendable and yet fast as it will be outlined for a large electromagnetic test problem.
  Minisymposium
  Minisymposium
  
  MS11 Large-Scale Scientific Computing with Julia
  
  Organizer: Lars Ruthotto (Emory University, USA)
  
  Co-Organizers: Eldad Haber (University of British Columbia, Canada)
  
  Julia is a new and emerging dynamic high-level programming language designed for high performance computing. Julia enables rapid development of expressive and easily extensible code and therefore becomes increasingly popular in science and engineering. It overcomes some limitations often associated with established dynamic languages, most importantly, large computational overhead and imperfect scaling on parallel computers. This mini symposium contains examples that outline Julia's potential for the solution of large-scale real-world problems.
• Räss Ludovic Poster
  
  Poster
  
  EAR-08 Use of High Performance and Massively Parallel GPU Computing to Resolve Nonlinear Waves in Poromechanics, Ludovic Räss (University of Lausanne, Switzerland)
  
  Co-Authors: Samuel Omlin (University of Lausanne, Switzerland); Yuri Podladchikov (University of Lausanne, Switzerland)
  
  The Nvidia GPU chip architecture sets new challenges for High Performance Computing (HPC). In order to exploit the large number of Cuda cores optimally, new numerical applications need to be designed. We propose a HPC poromechanical solver that runs on massively parallel multi GPU platforms, such the Cray XC30 Piz Daint, at CSCS. We performed successful preliminary runs on the Cray XC30 Piz Daint at CSCS, and observed linear scaling of our application from 1 to 2048 GPU nodes. The use of supercomputers allowed us to capture the expected nonlinear waves, which results in localized fluid flow through high porosity pathways.

S

• Sakurai Tetsuya MS Presentation
  Monday, June 1, 2015
  HG F30, 14:00-14:30
  
  MS Presentation
  
  A Scalable Parallel Eigensolver for Large-Scale Simulations in Post-Peta Scale Environments, Tetsuya Sakurai (University of Tsukuba, Japan)
  
  Co-Authors: Yasunori Futamura (University of Tsukuba, Japan); Akira Imakura (University of Tsukuba, Japan)
  
  Large-scale eigenvalue problems arise in wide variety of applications such as nano-scale materials simulation, vibration analysis of automobiles, big data analysis, etc, and high performance solvers are required to exploit distributed parallel computing environments. In this talk, we present a contour-integral based parallel eigensolver for interior eigenvalue problems. This method has a good parallel scalability according to a hierarchical structure of the method. We also present some efficient implementations of the method for large-scale simulations. We show performance evaluations on state-of-art supercomputers with some applications.
• Salamin Nicolas MS Presentation
  Tuesday, June 2, 2015
  HG E21, 11:00-11:30
  
  MS Presentation
  
  Efficient Approaches to Model Evolution in Computational Biology, Nicolas Salamin (University of Lausanne, Switzerland)
  
  Co-Authors:
  
  Biological modeling is important to study the evolution of genes and organisms. The increased availability of genomic data is pushing for the development of HPC approaches to model evolutionary processes. The availability of Bayesian approaches has been essential to extend the realism of the evolutionary models. The development of these methods based on MCMC techniques is computationally intensive and requires HPC approaches to deal with the computational complexity. I will present recent work to optimize and parallelize MCMC techniques. I will discuss our efforts to extend modeling of adaptation from genomic data to complex phenotypic traits using hierachical Bayesian computations.
• Samuel Henri Poster
  
  Poster
  
  EAR-01 Advection of Material Interfaces via a Two-Way Particle Level Set Approach, Henri Samuel (CNRS, France)
  
  Co-Authors:
  
  Level set interface tracking is a powerful approach consisting in advecting a smooth function (often taken as the signed distance to the interface) and in reinitializing it permanently to maintain this convenient property. I have recently shown that replacing the standard one-way advection by a two-way wave equation, reduces considerably both the associated errors and complexity. To reduce the erroneous displacements during the reinitialization step, I have now combined this two-way Eulerian advection with Lagrangian particles, as in [Enright et al., JCP, 183, 2002]. Tests reveal a further order-of-magnitude reduction in mass error, making this new hybrid method very competitive.
• Sanan Patrick Poster
  
  Poster
  
  EAR-02 Aggressive Local Smoothing on Accelerators for Stokes Flow, Patrick Sanan (Università della Svizzera italiana, Switzerland)
  
  Co-Authors: Dave A. May (ETH Zurich, Switzerland); Olaf Schenk (Università della Svizzera italiana, Switzerland); Karl Rupp (Technische Universität Wien, Austria)
  
  Hybrid supercomputers offer attractive performance per Watt, yet require care to avoid communication-related bottlenecks. We investigate 'heavy smoothing' with a multigrid preconditioner, using aggressive coarsening with accelerator-enabled smoothing to reduce communication. This maintains scalability and admits a tradeoff between local work and non-local communication. We examine local polynomial smoothing as well as a recently-developed fine-grained ILU decomposition. As part of the GeoPC PASC project, we present portable software contributions within ViennaCL, PETSc, and pTatin3d and results applied to ill-conditioned linear systems from lithospheric dynamics.
• Sargent Thomas Invited Presentation
  Tuesday, June 2, 2015
  HG F30, 18:30-19:20
  
  Invited Presentation
  
  IP4 Computational Challenges in Macroeconomics, Thomas Sargent (New York University, USA)
  
  Chair: Felix Kübler (University of Zurich, Switzerland)
  
  Abstract
  Modern macroeconomic theory has provided many qualitative insights into the functioning of financial and labor markets in our complex modern economies. But determining the quantitative importance of frequently countervailing forces requires much more than the pencil and paper methods used by economic theorists. My talk will describe some new economic ideas brought by economic theory and how efforts to develop computational tools promise to help macroeconomists perform quantitative analyses that can inform policy choices.
  
  Biography
  Thomas J. Sargent is the W. R. Berkley Professor at New York University, a joint appointment in the Economics Department and the Stern School of Business. He has been a Senior Fellow at the Hoover Institution since 1987. He was awarded the 2011 Nobel Prize in Economics for his empirical research on cause and effect in the macroeconomy. Professor Sargent earned his BA from the University of California, Berkeley in 1964, and his PhD in economics from Harvard University in 1968. He has held Professor appointments at the University of Minnesota, University of Chicago, and Stanford University. Sargent was elected a fellow of the National Academy of Sciences in 1983.
• Sawyer William MS Presentation
  Monday, June 1, 2015
  HG E1.1, 14:30-15:00
  
  MS Presentation
  
  Using GPUs Productively for the ICON Climate Model, William Sawyer (CSCS / ETH Zurich, Switzerland)
  
  Co-Authors: Markus Wetzstein (CSCS / ETH Zurich, Switzerland); Leonidas Linardakis (Max Planck Institute for Meteorology, Germany)
  
  The Swiss National Supercomputing Centre (CSCS) has undertaken a full port of the Icosahedral Non-hydrostatic (ICON) climate model currently under development at the Max Planck Institute for Meteorology (MPI-M) and the German Weather Service (DWD). Central to this port is the co-design with MPI-M and DWD developers to ensure that the code changes are minimal and non-intrusive, do not impose any significant changes in CPU performance, and can be incorporated directly into the development trunk. In this talk we present some of the techniques which facilitate this co-design project and initial results indicating that the performance is directly related to the peak memory bandwidth.
  Poster
  
  Poster
  
  EMD-02 Towards the HPC-Inference of Causality Networks from Multiscale Economical Data, William Sawyer (CSCS / ETH Zurich, Switzerland)
  
  Co-Authors: Illia Horenko (Università della Svizzera italiana, Switzerland); Patrick Gagliardini (Università della Svizzera italiana, Switzerland)
  
  One of the challenges in economics is to learn about the causality relations based on available data. The presence of unresolved scale quantities may result in non-stationarity and non-homogeneity of the data-driven models. They can manifest themselves in a presence of trends, regime-transition behavior and lead to biased results when applying standard stationary causality inference approaches used in Economics (e.g., Granger causality, an approach that was distinguished with a Nobel prize in Economics in 2003). The main aim of this project is implementing an HPC-framework for multiscale causality inference and testing it on a set of the credit rating data of increasing complexity.
• Scheidegger Simon MS Presentation
  Tuesday, June 2, 2015
  HG F1, 10:30-11:00
  
  MS Presentation
  
  Scalable High-dimensional Dynamic Stochastic Economic Modeling, Simon Scheidegger (University of Zurich, Switzerland)
  
  Co-Authors: Johannes Brumm (University of Zurich, Switzerland); Dmitry Mikushin (Università della Svizzera italiana, Switzerland); Olaf Schenk (Università della Svizzera italiana, Switzerland)
  
  I will present a highly parallelizable and flexible computational method to solve high-dimensional stochastic dynamic economic models. Solving such models often requires the use of iterative methods, like dynamic programming. The solution method I will present includes the use of a fully adaptive sparse grid algorithm and the use of a mixed MPI-Intel TBB-CUDA/Thrust implementation to improve the interprocess communication strategy on massively parallel architectures. Numerical experiments on 'Piz Daint' (Cray XC30) at the Swiss National Supercomputing Centre show that high-dimensional international real business cycle models up to at least 50 countries can be solved efficiently.
  Minisymposium
  Minisymposium
  
  MS10 Emerging Applications of Computation in Economics
  
  Organizer: Simon Scheidegger (University of Zurich, Switzerland)
  
  Co-Organizers: Kenneth L. Judd (Stanford University, USA)
  
  Clearly, the world economy is an extremely complex system. Even when modeling only the most relevant features of a small part of this system, one easily ends up with a large and intricate formal structure. A further complication stems from the fact that human beings choose their actions based on expectations about an uncertain future. This feedback from the future makes economic modeling particularly difficult. This symposium provides an overview of recent developments in how state-of-the art computational economical modeling tackles these issues in the fields of macroeconomics, optimal taxation, game theory, climate change, and growth models.
• Schlemmer Linda Poster
  
  Poster
  
  CLI-05 Modifications of the Atmospheric Moisture Field as a Result of Cold-Pool Dynamics, Linda Schlemmer (ETH Zurich, Switzerland)
  
  Co-Authors: Cathy Hohenegger (Max Planck Institute for Meteorology, Germany)
  
  This study investigates the interplay between atmospheric moisture and deep convective clouds via cold-pool dynamics in a series of cloud-resolving modelling studies. More specifically the contribution of moisture advection, evaporation of rain and surface fluxes to the moisture budget over particular regions of the domain is investigated. To further identify the origin of the advected moisture additional scalars marking moisture originating from the surface and from the evaporation of rain are introduced into the model. It is seen that the major part of the moisture in the regions under question stems from moisture already present before the downdraft hit the sub-cloud layer.
• Schmedders Karl MS Presentation
  Tuesday, June 2, 2015
  HG F1, 12:00-12:30
  
  MS Presentation
  
  Polynomial Problems in Economics, Karl Schmedders (University of Zurich, Switzerland)
  
  Co-Authors:
  
  Economic models frequently lead to polynomial systems of equations or polynomial optimization problems. Although the literature on computational algebraic geometry has made great progress in solving polynomial systems of equations and polynomial optimization problems, many difficulties remain. In particular, currently we can only solve rather small problem instances on computers. In this talk, we will review some economic applications of polynomial methods and discuss some current technical limitations.
• Schneider Reinhold MS Presentation
  Monday, June 1, 2015
  HG E3, 17:00-17:30
  
  MS Presentation
  
  Tensor Product Approximation and Tensor Networks for Many Particle Schrödinger Equation, Reinhold Schneider (Technische Universität Berlin, Germany)
  
  Co-Authors: Max Pfeffer (Technische Universität Berlin, Germany)
  
  Hierarchical tensor formats and tensor trains introduced recently in numerical analysis have the same parametrization as tree tensor networks and matrix product states. These representations offer stable and robust approximation of high-order tensors and multi-variate functions by a low order cost. For many high dimensional problems, including many body quantum mechanics, uncertainty quantification etc., this approach has a certain potential to circumvent from the curse of dimensionality. For numerical computations, the groundstate computation can be cast into optimization problems constraint by restricting to set of d tensors of bounded multilinear ranks.
• Schnepp Sascha Poster
  
  Poster
  
  CSM-10 Pipelined Flexible Krylov Subspace Methods for Large-Scale Computing, Sascha Schnepp (ETH Zurich, Switzerland)
  
  Co-Authors: Patrick Sanan (Università della Svizzera italiana, Switzerland); Dave May (ETH Zurich, Switzerland)
  
  We present variants of Conjugate Gradient (CG), Conjugate Residual (CR), and Generalized Minimal Residual (GMRES) methods which are both pipelined and flexible, allowing overlap of global reductions with sparse matrix multiplies and inexact or variable preconditioner applications. The methods are aimed at hiding network latencies for high performance computing using large compute clusters. We demonstrate their effectiveness on synthetic problems and practical examples for nonlinear mantle convection with heterogeneous viscosity structure.
• Schoeberl Joachim MS Presentation
  Tuesday, June 2, 2015
  HG E1.2, 15:00-15:30
  
  MS Presentation
  
  High Performance Implementation of High Order Finite Element Operations, Joachim Schoeberl (Vienna University of Technology, Austria)
  
  Co-Authors:
  
  We discuss an object oriented design of finite element core functionality. It allows to separate the mathematical definition of the finite element basis functions, the efficient implementation of operations, and the calculation of stiffness matrices and residual vectors. We show how features of the C++11 programming language help to reduce code complexity and thus allow for additional performance optimization such as vectorization.The presented techniques are implemented in the open source finite element package NGSolve. We present examples including matrix-free solvers and high order tent-pitching methods for conservation laws.
• Sclauzero Gabriele Poster
  
  Poster
  
  MAT-09 DFT+DMFT Study of Strain and Interface Effects in D1 and D2 T2g-Perovskites, Gabriele Sclauzero (ETH Zurich, Switzerland)
  
  Co-Authors: Krzysztof Dymkowski (ETH Zurich, Switzerland); Claude Ederer (ETH Zurich, Switzerland)
  
  We address the metal-insulator transitions in epitaxial films of early-transition metal oxides through DFT+DMFT. We show that the Mott insulating phase in LaVO3 is robust against epitaxial strain, while LaTiO3 becomes metallic under compressive strains, in agreement with thin-film experiments. The combined effect of strain-induced changes in the crystal field splitting of t2g orbitals and the different t2g occupations of the two materials can account for this behaviour. The origin of the metallicity at the LaVO3/SrTiO3 interface is still unclear and calls for an explicit modelling of reconstruction effects at the thin-film/substrate interface.
• Sewer Alain MS Presentation
  Monday, June 1, 2015
  HG E1.2, 14:00-14:30
  
  MS Presentation
  
  Building Computable Biological Network Models and their Application to Product Risk Assessment, Alain Sewer (Philip Morris International R&D, Switzerland)
  
  Co-Authors: Manuel C. Peitsch (Philip Morris International, Switzerland); Julia Hoeng (Philip Morris International, Switzerland)
  
  Systems Toxicology is aimed at decoding the toxicological blueprint of active substances that interact with living systems. It integrates classic toxicology with network models and quantitative measurements of molecular and functional changes across multiple levels of biological organization. It combines high content experimental data obtained with computational and mathematical sciences to identify Pathways of Toxicity and the Key Events in Adverse Outcome Pathways. These findings are applied to perform a mechanism-by-mechanism evaluation of the impact of products on these pathways. The presentation will focus on the computational aspects of Systems Toxicology and its applications.
• Sidler Christof MS Presentation
  Tuesday, June 2, 2015
  HG E1.1, 13:30-14:00
  
  MS Presentation
  
  Examples of Accelerating Algorithms with FPGAs, Christof Sidler (Supercomputing Systems AG, Switzerland)
  
  Co-Authors:
  
  FPGAs can dramatically speed-up certain algorithms. This presentation will show some practical examples of algorithms accelerated by FPGAs.
• Solcà Raffaele Poster
  
  Poster
  
  CSM-08 Time and Energy to Solution Study of the Generalized Eigenvalue Solver, Raffaele Solcà (ETH Zurich, Switzerland)
  
  Co-Authors: Thomas Schulthess (ETH Zurich / CSCS, Switzerland)
  
  We study time and energy to solution of electronic structure simulation in terms of a model that relates application performance to machine parameters. We focus on the generalised eigenvalue problems for dense matrices, and study distributed memory architectures with multi-core CPU and hybrid CPU-GPU nodes. The model explains, under certain conditions, the empirically observed affine relationship between node-hours and the energy to solution consumed by the computation. It allows us to extract an effective dynamic energy and static power (which we relate to effective leakage) for the application running on different architectures.
• Spampinato Daniele MS Presentation
  Tuesday, June 2, 2015
  HG F1, 14:00-14:30
  
  MS Presentation
  
  A Basic Linear Algebra Compiler, Daniele Spampinato (ETH Zurich, Switzerland)
  
  Co-Authors: Markus Püschel (ETH Zurich, Switzerland)
  
  Many applications in media processing, control, graphics, and other domains require efficient small-scale linear algebra computations. However, most existing high performance libraries for linear algebra, such as ATLAS or Intel MKL, are geared towards large problems (matrix sizes in the hundreds and larger). We present LGen: a compiler for small-scale, basic linear algebra computations. The input to LGen is a fixed-size linear algebra expression; the output is a corresponding C function optionally SIMD-vectorized. We show benchmarks against libraries (e.g., Intel MKL) and alternative generators (e.g., C++ template-based Eigen). The typical speedup is about a factor of two to three.
• Springel Volker Invited Presentation
  Monday, June 1, 2015
  HG F30, 10:15-11:10
  
  Invited Presentation
  
  IP1 Simulating Cosmic Structure Formation, Volker Springel (University of Heidelberg, Germany)
  
  Chair: Ben Moore (University of Zurich, Switzerland)
  
  Abstract
  Numerical simulations on supercomputers play an ever more important role in astrophysics. They have become the tool of choice to predict the non-linear outcome of the initial conditions left behind by the Big Bang, providing crucial tests of cosmological theories. However, the problem of galaxy and star formation confronts us with a staggering multi-physics complexity and an enormous dynamic range that severely challenges existing numerical methods. In my talk, I review current strategies to address these problems, focusing on recent developments in the field such as hierarchical time integration schemes, improved particle- and mesh-based hydrodynamical solvers, and novel parallelization schemes.
  
  Short Biography
  Volker Springel became professor for Theoretical Astrophysics at Heidelberg University in 2010, where he now leads a research group at the Heidelberg Institute for Theoretical Studies (HITS). Springel previously was a group leader in computational cosmology at the Max-Planck-Institute for Astrophysics in Garching, after working as a postdoctoral researcher at the Harvard Center for Astrophysics in the United States. He received his PhD in 2000 from the Ludwig-Maximilian University in Munich, after studying physics at the University of Tübingen and the University of California, Berkeley.
• Staar Peter W. J. MS Presentation
  Tuesday, June 2, 2015
  HG F30, 14:00-14:30
  
  MS Presentation
  
  Big Data Based Materials Discovery, Peter W. J. Staar (IBM Research, Switzerland)
  
  Co-Authors:
  
  Traditionally, the discovery of new materials is extremely labor-intensive. This approach is not scalable and as such this field of research is ideally suited for a big data based cognitive computing approach. In this talk, we will present how this approach is applied in practice. We will discuss in detail the various aspects ranging from the used data-models to the algorithms applied. Special attention will be given to the design and construction the knowledge graph, which encodes the underlying data and knowledge model. We will discuss in detail how the nodes and edges in the graph are designed and how the algorithms can be used to refine and optimize the edges of the graph.
• Stamm Benjamin Minisymposium
  Minisymposium
  
  MS06 Mathematical Aspects of Multi-Scale Methods in Computational Chemistry
  
  Organizer: Benjamin Stamm (Sorbonne Universités, UPMC Univ Paris 06, France)
  
  Co-Organizers:
  
  In the last decades, multi-scale strategies have been the object of a wide and fruitful development. Such methodologies, including QM/MM methods and QM/Continuum methods, allow to model different chemical parts with different accuracies in the description of a molecular system. For strongly inhomogeneous systems, as it is the case for biological environments, such techniques are extremely important resulting in a large variety of real-world applications at an affordable computational price. The scope of this mini-symposium is to bring chemists and mathematicians working on the methodological development of such methods together and foster this inter-disciplinary research field.
• Steiner Oskar Poster
  
  Poster
  
  PHY-09 Polarized Radiative Transfer in Discontinuous Media, Oskar Steiner (Istituto Ricerche Solari Locarno, Switzerland)
  
  Co-Authors: Francesco Züger (Istituto Ricerche Solari Locarno, Switzerland); Luca Belluzzi (Istituto Ricerche Solari Locarno, Switzerland)
  
  Astrophysical numerical radiation hydrodynamic simulations often show shock fronts, contact discontinues, or steep gradients across which radiative transfer needs to be computed. In view of applications to the solar atmosphere, we here propose a new method for performing radiative transfer of polarized light through a discontinuous atmosphere using methods of shock capturing hydrodynamics. We use slope-limiter methods for the reconstruction of the source function.
• Stengel Holger MS Presentation
  Tuesday, June 2, 2015
  HG F3, 13:30-14:00
  
  MS Presentation
  
  EXASTEEL - Node Level Performance Analysis and Optimization in FETI-DP Methods, Holger Stengel (University of Erlangen-Nuremberg, Germany)
  
  Co-Authors: Axel Klawonn (University of Cologne, Germany); Martin Lanser (University of Cologne, Germany); Oliver Rheinbach (Technische Universität Bergakademie Freiberg, Germany); Gerhard Wellein (University of Erlangen-Nuremberg, Germany)
  
  We present the performance engineering process for a FETI-DP application (2D nonlinear hyperelasticity) used in the SPPEXA project EXASTEEL. Starting from a pure MPI code we incrementally develop a hybrid MPI/OpenMP parallel approach that allows efficient utilization of current multi-core compute nodes. The PETSc+MPI-based finite element assembly is parallelized with OpenMP and as a multithreaded direct solver we have chosen Pardiso. This hybrid parallel approach can extend the scalability range of FETI-DP methods by up to a factor of four. Besides a multi-core scaling analysis each step is accompanied with single-core runtime analysis revealing relevant performance bottlenecks.
• Stevens Bjorn Invited Presentation
  Monday, June 1, 2015
  HG F30, 11:10-12:00
  
  Invited Presentation
  
  IP2 The Great Leap, Bjorn Stevens (Max-Planck-Institute for Meteorology, Germany)
  
  Chair: Christoph Schär (ETH Zurich, Switzerland)
  
  Abstract
  Increasing computational power and advances in algorithms have made it possible to resolve an ever increasing fraction of the scales of atmospheric motion. While it remains inconceivable to resolve all the relevant scales of motion we are currently in the midst of a great leap across a range of scales that have posed some of the greatest challenges to climate science over the past sixty years. This leap is bringing wholly new insights into the structure of the climate system on both ends of the spectrum of atmospheric motions.
  
  Biography
  Bjorn Stevens's research blends modeling, theory and field work to understand the role of clouds and moist convection on the climate system. Stevens has made pioneering contributions to our understanding of how mixing and microphysical processes influence both cloud structure and its susceptibility to changes in the environment. Stevens has contributed more than 130 scholarly articles to the peer reviewed literature and received many honors, including the Clarence Leroy Meisinger Award of the American Meteorological Society for "pioneering advances in understanding and modeling of cloud-topped boundary layer.? He currently co-leads the World Climate Research Programme?s Grand Challenge on Clouds, Circulation and Climate Sensitivity.
• Sukys Jonas Contributed Talk
  Wednesday, June 3, 2015
  HG F3, 11:10-11:30
  
  Contributed Talk
  
  Petascale Simulations of Cloud Cavitation Collapse, Jonas Sukys (ETH Zurich, Switzerland)
  
  Co-Authors: Panagiotis Hadjidoukas (ETH Zurich, Switzerland); Diego Rossinelli (ETH Zurich, Switzerland); Fabian Wermelinger (ETH Zurich, Switzerland); Babak Hejazialhosseini (ETH Zurich, Switzerland); Petros Koumoutsakos (ETH Zurich, Switzerland)
  
  Simulation of cloud cavitation collapse pertains to the erosion of liquid-fuel injectors, hydropower turbines and ship propellers, or can even be harnessed in treating kidney stones by shock wave lithotripsy. Cavities undergo a collapse at higher pressure regions inside the flow and produce extreme pressure spots, possibly causing erosion. Numerical simulations of cloud cavitation collapse require two phase flow solvers capable of capturing interactions between multiple deforming bubbles. We present the two-phase flow simulations enabling quantitative prediction of cavitation of clouds containing up to 50'000 vapor bubbles on unprecedented resolution of up to 1 trillion grid points.

T

• Tagami Daisuke MS Presentation
  Monday, June 1, 2015
  HG F30, 14:30-15:00
  
  MS Presentation
  
  Development of a Numerical Library Based on Hierarchical Domain Decomposition, Daisuke Tagami (Kyushu University, Japan)
  
  Co-Authors: Ryuji Shioya (Toyo University, Japan); Masao Ogino (Nagoya University, Japan); Hiroshi Kawai (The University of Toyo, Japan); Yasuhi Nakabayashi (The University of Toyo, Japan)
  
  We have been developing an open source system software, ADVENTURE, which is a general-purpose parallel finite element analysis system and can simulate a large scale analysis model with supercomputer like K-computer. In the system, HDDM (hierarchical domain decomposition method), which is a very effective technique to large-scale analysis, was developed. The aim of this project is to develop a numerical library based on HDDM that is extended to pre and post processing parts, including mesh generation and visualization of large-scale data, for the Post Petascale simulation.
• Tavernelli Ivano Contributed Talk
  Wednesday, June 3, 2015
  HG F1, 11:10-11:30
  
  Contributed Talk
  
  Solid State Electrolyte Transporter from Ab-Initio, Machine Learning and Data Mining Techniques, Ivano Tavernelli (IBM Research, Switzerland)
  
  Co-Authors: Matthieu Mottet (IBM Research, Switzerland); Teodoro Laino (IBM Research, Switzerland); Alessandro Curioni (IBM Research, Switzerland)
  
  Solid state electrolyte transporter for Lithium ions are of great importance in the design of highly efficient batteries. In this study we investigate the static and dynamical properties of a new class of ceramic electrolytes based on the Garnet structure that exhibits the unprecedented combination of high ionic conductivity and chemical stability. Using data from ab-initio metadynamics combined with thermodynamic integration we parametrized a kinetic Monte Carlo model for the characterization of the transport properties at ambient conditions. This model is the starting point for the design of better performing materials using machine learning and data mining techniques.
• Teixeira Frederico Poster
  
  Poster
  
  LS-09 Novel Semi-Implicit Approach and Platform for Personalized Fluid-Structure Interaction Modeling, Frederico Teixeira (ETH Zurich, Switzerland)
  
  Co-Authors: Phillipp Wissmann (ETH Zurich, Switzerland); Esra Neufeld (ETH Zurich, Switzerland); Dominik Szczerba (ETH Zurich, Switzerland); Niels Kuster (ETH Zurich, Switzerland)
  
  A semi-implicit method is developed to avoid the complexity of monolithic ones and interface degradation issues of common semi-implicit methods. It adapts the fluid mesh at every nonlinear iteration, instead of once per time-step, keeping the monolithic robustness while reducing complexity. It is implemented in a PETSc-based framework which supports HPC and mixed elements (tet/prism/hex/...). The platform incorporates a novel vasculature segmentation technique with tunable interactivity and adaptive surface generation methods to create realistic models from medical image data, as well as modeling, scripting, analysis, and visualization functionality. The methods are applied to real aneurysm.
• Thöle Florian Poster
  
  Poster
  
  PHY-10 The Bulk Monopolization in Diagonal Magnetoelectrics, Florian Thöle (ETH Zurich, Switzerland)
  
  Co-Authors: Michael Fechner (ETH Zurich, Switzerland); Nicola Spaldin (ETH Zurich, Switzerland)
  
  The magnetoelectric monopole - a term appearing in the expansion of the interaction energy of the magnetic field H with the magnetization density - is formally allowed to exist in materials with broken time-reversal and inversion symmetry and is related to the response of diagonal magnetoelectrics. By exploiting similarities to the modern theory of polarization, we develop a theory of bulk monopolization and implement a Wannier function-based calculation in first-principles simulations. We compare the values obtained in the bulk picture to values obtained in a local moment picture and point out the relevance of each description.
• Tkachenko Natalie Poster
  
  Poster
  
  LS-02 Agent-based Modeling of Hunter-gatherer Populations, Natalie Tkachenko (University of Zurich, Switzerland)
  
  Co-Authors: Simone Callegari (University of Zurich, Switzerland); John David Weissmann (University of Zurich, Switzerland); George Lake (University of Zurich, Switzerland); Christoph P. E. Zollikofer (University of Zurich, Switzerland)
  
  Agent-based models (ABMs) are bottom-up computational models of ecological systems and populations built from individuals. In the context of the PASC project HPC-ABGEM, we develop an ABM of human population dynamics in space and time, including life history traits of modern hunter-gatherers such as individual survival probability as a function of age and sex, daily energy requirements, fertility rates, etc. We present preliminarily simulation results and compare the emerging large-scale characteristics of the simulated population with observational data and classical (continuum) models.
• Tuckerman Mark MS Presentation
  Monday, June 1, 2015
  HG E3, 16:30-17:00
  
  MS Presentation
  
  Multi-Scale Modeling Techniques for the Exploration of Complex Systems: Polypeptides, Crystals, and Quantum Systems, Mark Tuckerman (New York University, USA)
  
  Co-Authors:
  
  I will discuss our recent advances in the development of multi-scale approaches for studying complex condensed-phase systems. For polypeptide and crystal structure prediction, I will discuss heterogeneous multi-scaling techniques we have introduced for the exploration of high-dimensional free energy surfaces. A two-part strategy will be described wherein landmark features of the surface, i.e., minima and saddles, are first identified and then fed into an enhanced sampling scheme for generating free energy differences. I will then discuss our attempts to combine adaptive resolution techniques with the Feynman path integral for treating systems in which nuclear quantum effects are important
• Tumolo Giovanni MS Presentation
  Monday, June 1, 2015
  HG E1.1, 13:00-13:30
  
  MS Presentation
  
  An Accurate and Efficient Numerical Framework for Adaptive Numerical Weather Prediction, Giovanni Tumolo (The Abdus Salam ICTP, Italy)
  
  Co-Authors: Luca Bonaventura (Politecnico di Milano, Italy)
  
  We present an adaptive discretization approach for model equations typical of NWP, which combines the semi-Lagrangian technique with a TR-BDF2 semi-implicit time discretization and with a DG spatial discretization with (arbitrarily high) variable and dynamically adaptive element degree. The resulting method has full second order accuracy in time, is unconditionally stable and can effectively adapt at runtime the number of degrees of freedom employed in each element, in order to balance accuracy and computational cost. Numerical results of classical 2D benchmarks for shallow water equations on the sphere and Euler equations on a vertical slice confirm the potential of the proposed formulation.

U

• Uekermann Benjamin MS Presentation
  Monday, June 1, 2015
  HG E1.1, 16:30-17:00
  
  MS Presentation
  
  preCICE - a Library for Flexible Surface-Coupling on Massively Parallel Systems, Benjamin Uekermann (Technische Universität München, Germany)
  
  Co-Authors: Hans-Joachim Bungartz (Technische Universität München, Germany); Florian Lindner (Universität Stuttgart, Germany); Miriam Mehl (Universität Stuttgart, Germany)
  
  With increasing compute power, the simulation of multi-physics gains more and more popularity. At the same time, a flexible software development process including existing single-physics codes becomes a necessity to cope with the overall complexity. preCICE is an open-source library to surface-couple single-physics codes. Its high-level API allows to write adapters for single-physics codes in 30 lines. Afterwards various codes can be coupled in a nearly plug-and-play manner. preCICE offers methods for interpolation, means for parallel communication, and sophisticated fix-point acceleration schemes. In this talk, we present our efforts and tests on the parallel efficiency of preCICE.
• Uhrin Martin MS Presentation
  Monday, June 1, 2015
  HG F1, 16:00-16:30
  
  MS Presentation
  
  Challenges in Big Data: When Materials Go Beyond the Periodic Table, Martin Uhrin (EPFL, Switzerland)
  
  Co-Authors: Chris J. Pickard (University College London, United Kingdom)
  
  The properties of any material are, on some level, a function of its internal structure making an understanding of structure key to the discovery process. We use a simple model for particles that attract at long range and repel at short to probe possible structures over a wide range of parameters and stoichiometries. In exploring their energy landscapes we populate a schemaless, MongoDB, database with over 20M entries, many more than current databases populated from experiment or electronic structure calculations. We discuss some challenges of performing local and global analysis over such a large dataset and detail some interesting finds including never before seen structures.
• Unat Didem MS Presentation
  Monday, June 1, 2015
  HG E1.2, 15:30-16:00
  
  MS Presentation
  
  Tiling Abstraction for Data-Centric Computing, Didem Unat (Koc University, Turkey)
  
  Co-Authors: Tan Nguyen (Lawrence Berkeley National Laboratory, USA); Weiqun Zhang (Lawrence Berkeley National Laboratory, USA); John B. Bell (Lawrence Berkeley National Laboratory, USA); John Shalf (Lawrence Berkeley National Laboratory, USA)
  
  Programming models play a crucial role in providing the necessary tools to express locality and minimize data movement, while also abstracting complexity from programmers. Unfortunately, existing compute-centric programming environments provide few abstractions to manage data movement and locality, and rely on a large shared cache to virtualize data movement. We propose three programming abstractions, tiles, layout and loop traversal that address data locality and increased parallelism on emerging parallel computing systems. The TiDA library implements these abstractions in the data structures through domain decomposition and provides performance portable codes.

V

• van 't Wout Elwin Poster
  
  Poster
  
  CSM-11 Simulating Large-Scale Scattering Phenomena with the Open-source Boundary Element Library BEM++, Elwin van 't Wout (University College London, United Kingdom)
  
  Co-Authors: Timo Betcke (University College London, United Kingdom); Pierre Gélat (University College London, United Kingdom); Simon Arridge (University College London, United Kingdom)
  
  The scattering of acoustic and electromagnetic waves are important phenomena in many physical models. The boundary element method naturally reduces the wave propagation in exterior domains into a model on the surface of the object. The potential benefits for simulation of large-scale engineering problems can only be achieved with additional fast algorithms. The open-source library BEM++ provides an advanced computing platform for scattering analysis of general 3D structures. We will present ongoing developments and demonstrate its applicability to large-scale scattering phenomena with simulations of high-intensity focused ultrasound modalities for medical treatment of liver cancer.
• Vander Aa Tom MS Presentation
  Tuesday, June 2, 2015
  HG E21, 13:30-14:00
  
  MS Presentation
  
  ExaShark: A Scalable Hybrid Array Kit for Exascale Simulation, Tom Vander Aa (IMEC, Belgium)
  
  Co-Authors: Imen Chakroun (IMEC, Belgium); Bram Reps (University Antwerpen, Belgium); Wim Vanroose (University Antwerpen, Belgium); Roel Wuyts (IMEC, Belgium)
  
  Design and implementation of solvers that scale to large HPC systems is not only a challenge from a numerical point of view, but also - and even more so - from an implementation and verification point of view. To estimate how a newly developed solver would behave on modern heterogeneous HPC architectures, knowledge of many different programming paradigms and libraries is needed. ExaShark's goal is to reduce the increasing programming burden while still offering good performance.
• Vanroose Wim Minisymposium
  Minisymposium
  
  MS20 Integrating Resilience and Communication Hiding/Avoiding in a Extreme Scale Preconditioned Krylov Solver
  
  Organizer: Wim Vanroose (University Antwerpen, Belgium)
  
  Co-Organizers: Olaf Schenk (Università della Svizzera italiana, Switzerland)
  
  We report the progress of the Exascale Algorithms and Advanced Computational Techniques (Exa2CT) project funded by the European Union. In this project resilience and communication avoiding/hiding techniques are integrated in a preconditioned Krylov solver for exascale applications. We use the pipelining to hide the latencies of the global reductions, stencil compilers to increase the arithmetic intensity and avoid communication at the level of the preconditioner. Resiliency to soft errors is also discussed. We report initial benchmarks of the components and the integrated prototypes on various proto applications ranging from combustion to mechanics.
• Verbosio Fabio Poster
  
  Poster
  
  CSM-02 Computing Entries of Inverse Matrices in Genomic Prediction Problems, Fabio Verbosio (Università della Svizzera italiana, Switzerland)
  
  Co-Authors: Arne De Coninck (Universiteit Gent, Belgium); Drosos Kourounis (Università della Svizzera italiana, Switzerland); Olaf Schenk (Università della Svizzera italiana, Switzerland)
  
  Genomic prediction problems for plant breeding require scalable and efficient techniques for the solution of sparse linear systems where the matrices involved contain both a large sparse and a dense block. A selective inversion process is suggested that consists of two main steps. First, the selective inverse of the sparse block is computed via the PARDISO solver. Secondly, the Schur-complement of the dense block is computed in parallel and it is used to update the entries of the matrix obtained at the first step. This approach allows the efficient evaluation of selected entries of the inverse, hence its trace that is essential for the solution of the associated maximum likelihood problem.
• Verma Siddhartha Contributed Talk
  Wednesday, June 3, 2015
  HG E22, 11:10-11:30
  
  Contributed Talk
  
  Experiment-Based Simulations of Interacting Swimmers, Siddhartha Verma (ETH Zurich, Switzerland)
  
  Co-Authors: Sergey Ivannikov (ETH Zurich, Switzerland); Petros Koumoutsakos (ETH Zurich, Switzerland)
  
  Large groups of fish in the ocean often swim in recognizable patterns. One line of thought identifies 'schooling' as an evolutionary means to dissuade attacks from predators. There is also speculation that such behavior may bestow a hydrodynamic advantage to swimmers. The mechanism of reduction in the required energy input may arise from interactions between the leaders' wake, and the undulatory motion of the following-swimmers. The goal of this study is to probe this possibility, by numerically re-creating swimming behavior observed in experiments. In preliminary studies, both the leader as well as the follower have been observed to experience an increase in swimming efficiency.
• Vilarrasa Victor MS Presentation
  Monday, June 1, 2015
  HG E22, 13:00-13:30
  
  MS Presentation
  
  Modeling of Inelastic Strain Induced by CO2 Injection, Victor Vilarrasa (EPFL, Switzerland)
  
  Co-Authors: Lyesse Laloui (EPFL, Switzerland)
  
  Inelastic strain, which is associated to the induced microseismicity observed in CO2 injection sites like Weyburn or In Salah, is likely to occur in CO2 storage sites. Furthermore, CO2 will reach the storage formation at a colder temperature than the rock, which will induce a thermal stress reduction. We model non-isothermal injection of CO2 in a deep saline formation using the fully coupled finite element code CODE_BRIGHT. We find that despite inelastic strain is likely to occur in the cooled region around the injection well, fracture propagation does not extend into the caprock in the normal faulting stress regime simulated in this study.
• Vlimant Jean-Roch MS Presentation
  Tuesday, June 2, 2015
  HG F30, 11:30-12:00
  
  MS Presentation
  
  Big & Smart, High Energy Data, Jean-Roch Vlimant (California Institut of Technology, USA)
  
  Co-Authors: Maria Spiropulu (Caltech, USA)
  
  The raw data rate at the LHC is 1 Petabyte/sec. In terms of production, capturing, communicating, aggregating, storing and analyzing it this is arguably the biggest and most challenging science data frontier and one that offers solutions for other fields and affords innovation as we move to the next era of the High Luminosity LHC when the needs and requirements can be an order magnitude bigger. I will discuss Big and Smart Data and intelligent data handling systems in high energy physics as we move forward towards the High Luminosity LHC and beyond and highlight the importance of validation and verification that we can afford in this science domain.

W

• Wadsley James MS Presentation
  Tuesday, June 2, 2015
  HG E3, 13:30-14:00
  
  MS Presentation
  
  A New Method for Fast Approximate Radiative Transfer, James Wadsley (McMaster University, Canada)
  
  Co-Authors: Rory Woods (McMaster University, Canada); Hugh Couchman (McMaster University, Canada)
  
  We present a new scheme for radiative transfer. The goals are a fast approach (low cost, no timestep constraints) for an approximate radiation field to drive chemistry and heating for many potential applications. The method combines ray tracing and the tree method. It is flexible in terms of geometry and can use multiple wave-bands. It has been implemented in the GASOLINE2 code but is modular and could be added to different code types. The result scales well with source numbers and is cheap for sub-regions (i.e. multiple timesteps). We demonstrate test results for ionizing radiation (HII regions) and shadowing tests. We show first results on some astrophysical applications.
• Weber Valery Contributed Talk
  Wednesday, June 3, 2015
  HG E22, 10:30-10:50
  
  Contributed Talk
  
  Off-Limits Modeling for Biological Systems, Valery Weber (IBM Research, Switzerland)
  
  Co-Authors: Teodoro Laino (IBM Research, Switzerland); Alessandro Curioni (IBM Research, Switzerland)
  
  Reliable atomistic modeling of complex life-science systems has so far been limited by the sizes and time scales accessible by molecular simulations. In this talk, we will present a radical and innovative approach to enable the application of quantum simulations to biological domains typical of classical Hamiltonians. Our approach will merge the extensive knowledge accumulated in the field of semi-empirical quantum simulation and the latest development in low-complexity and scalable algorithms, allowing us to exploit modern supercomputers efficiently and to achieve high scalability and low time to solution. Details of the methodology and efficiency will be discussed.
• Wellein Gerhard MS Presentation
  Tuesday, June 2, 2015
  HG F3, 12:00-12:30
  
  MS Presentation
  
  Performance Engineering of the Kernel Polynomial Method on Large-Scale CPU-GPU Systems, Gerhard Wellein (University of Erlangen-Nuremberg, Germany)
  
  Co-Authors: Moritz Kreutzer (University of Erlangen-Nuremberg, Germany); Georg Hager (University of Erlangen-Nuremberg, Germany); Andreas Pieper (University of Greifswald, Germany); Andreas Alvermann (University of Greifswald, Germany); Holger Fehske (University of Greifswald, Germany)
  
  The Kernel Polynomial Method (KPM) is a well-established scheme in quantum physics to determine eigenvalue density and spectral properties of large sparse matrices. We demonstrate the high performance potential of KPM on heterogeneous CPU-GPU supercomputers. At the node level we perform optimizations which transform our sparse matrix problem towards pure data streaming with high computational intensity. All optimizations are guided by a performance modelling process that indicates how computational bottlenecks change. The optimized KPM code is embedded in a hybrid-parallel framework to perform large scale KPM calculations on Piz Daint (CRAY XC30) achieving 0.5 PF on 4096 nodes.
• Wensch Joerg MS Presentation
  Tuesday, June 2, 2015
  HG E22, 10:30-11:00
  
  MS Presentation
  
  Multirate Infinitesimal Step Methods for the Compressible Euler Equations, Joerg Wensch (Technische Universität Dresden, Germany)
  
  Co-Authors: Oswald Knoth (Leibniz Institute of tropospheric research, Germany)
  
  The simulation of atmospheric dynamics relies on the numerical solution of the Euler equations. We consider here sound-advection systems for stratified fluids. These systems exhibit wave type solutions which impose CFL-based restrictions on the time step size. Buoyancy, sound and advection cause separate restrictions each on their own scale. We develop Finite Volume Methods where a splitting approach allows different treatment of fast and slow waves. The methods are optimized with respect to order and maximum stable CFL numbers. Different scenarios for the scales of sound and advection are discussed.
• Werder Mauro MS Presentation
  Monday, June 1, 2015
  HG E21, 15:30-16:00
  
  MS Presentation
  
  A Subglacial Hydrology Model Embedding a 1D Channel Network in a 2D Model-Domain, Mauro Werder (University of Zurich, Switzerland)
  
  Co-Authors:
  
  I present the Julia implementation of GlaDS, a Glacier Drainage System model. It simulates the water flow at the ice-bedrock interface underneath a glacier or ice sheet, both in a channelized and a diffusive macro-porous system. Its most innovative feature is that the strictly 1D channel equations are embedded into the 2D model-domain by simulating them on the edges of the numerical grid. The equations are a non-linear reaction-diffusion equation coupled to two local ODEs. The model uses finite elements combined with a Rosenbrock-W time stepper. I will review some of the code in detail and highlight Julia specific implementation strategies, such as algebraic datatypes and multiple dispatch.
• Wittum Gabriel MS Presentation
  Tuesday, June 2, 2015
  HG F3, 11:00-11:30
  
  MS Presentation
  
  Highly Scalable and Robust Multi-Grid Solvers for Extreme Computing, Gabriel Wittum (Universität Frankfurt, Germany)
  
  Co-Authors: Ingo Heppner (G-CSC, Germany); Michael Hoffer (G-CSC, Germany); Arne Nägel (G-CSC, Germany); Raphael Prohl (G-CSC, Germany); Sebastian Reiter (G-CSC, Germany); Martin Rupp (G-CSC, Germany); Andreas Vogel (G-CSC, Germany)
  
  Projected from today, exascale computers are characterized by billion way parallelism. Computing on such extreme scale needs methods with perfect scaling and optimal complexity. The solver itself must be of optimal numerical complexity - a requirement becoming more and more severe with increasing problem size - and scale efficiently on extreme scales of parallelism. We will present a multigrid approach scaling efficiently unto the full size of the largest computers available and looks promising for even larger scales. We further show that robustness can be maintained during the scaling process for relevant application problems while still maintaining optimal complexity.

X

• Xenarios Ioannis MS Presentation
  Monday, June 1, 2015
  HG E1.2, 13:00-13:30
  
  MS Presentation
  
  From Genomes to Systems Biology, Ioannis Xenarios (Swiss Institute of Bioinformatics, Switzerland)
  
  Co-Authors: Team members (Vital-IT and Swiss-Prot, Switzerland)
  
  From the part list genes, proteins and chemical compounds to the complex dynamics of biological systems, computational approaches are critical. In this presentation I'll describe the challenges linking the genotypes to the phenotypes and the ability/inability to reconstruct what we understand of the biological system. I'll take examples in the field of immunology and more 'simple' model organism to explain the methodologies used to both characterize a system and also to predict novel regulations or unexpected combination of perturbation.

Y

• Yakutovich Aliaksandr Poster
  
  Poster
  
  MAT-21 Understanding Enantioselectivity of PdGa High-Symmetry Surfaces, Aliaksandr Yakutovich (EMPA, Switzerland)
  
  Co-Authors: Carlo A. Pignedoli (EMPA, Switzerland); Jan Prinz (EMPA, Switzerland); Roland Widmer (EMPA, Switzerland); Oliver Groening (EMPA, Switzerland); Daniele Passerone (EMPA, Switzerland)
  
  The PdGa intermetallic compound emerged recently for its remarkable catalytic properties in the purification of the feedstock for polyethylene production. PdGa has an intrinsically chiral structure and can be grown in either enantiomeric forms. Recent striking experimental observations reveal a subtle difference in the enantioselectivity of the PdGa(111) and the PdGa(-1-1-1) termination with respect to adsorption of 9-ethynylphenanthrene. This poster presents the challenges recently encountered by atomistic modeling (based on density functional theory and on Monte Carlo simulations) in unravelling the origin of the experimental findings.
• Yamazaki Hiroe MS Presentation
  Monday, June 1, 2015
  HG E1.1, 15:30-16:00
  
  MS Presentation
  
  Development of a Vertical Slice Model Using Mixed FEM Discretizations, Hiroe Yamazaki (Imperial College London, United Kingdom)
  
  Co-Authors: Colin Cotter (Imperial College London, United Kingdom); Jemma Shipton (Imperial College London, United Kingdom); Lawrence Mitchell (Imperial College London, United Kingdom); David A Ham (Imperial College London, United Kingdom); Andrew McRae (Imperial College London, United Kingdom)
  
  A vertical slice model is developed for the Euler-Boussinesq equations with a constant temperature gradient in the y-direction (the Eady-Boussinesq model). The model is a solution of the full 3D equations with no variation normal to the slice, which is an idealized problem used to study the formation and subsequent evolution of weather fronts. Mixed FEM discretizations on an extruded mesh are performed using Firedrake, a high-level Python framework for the portable solution of PDEs on unstructured meshes. I will discuss the details of the vertical slice model implementations and the preliminary results obtained through test simulations developed to implement the modelling of fronts.
• Yeltekin Sevin MS Presentation
  Tuesday, June 2, 2015
  HG F1, 11:00-11:30
  
  MS Presentation
  
  Dynamic Oligopoly with Uncertain Demand, Sevin Yeltekin (Carnegie Mellon University, United States)
  
  Co-Authors:
  
  The evolution of strategic interaction, competition, and market power have been some of the central issues in the field of industrial organization. However, most of the analysis has been confined to simplified models since solving dynamic models with heterogenous firms, market entry and exit, investment are quite difficult. In this paper we introduce a numerical method for computing equilibria of dynamic multi-firm competition game with endogenous productivity, uncertain demand and entry/exit.

Z

• Zerzion Daniel Poster
  
  Poster
  
  LS-10 Standardization of Efficient Genomic Data Representation and Processing, Daniel Zerzion (Swiss Institute of Bioinformatics, Switzerland)
  
  Co-Authors: Yann Thoma (HEIG-VD, Switzerland); Enrico Petraglio (HEIG-VD, Switzerland); Claudio Alberti (EPFL, Switzerland); Marco Mattavelli (EPFL, Switzerland); Pascal Kahlem (Swiss Institute of Bioinformatics, Switzerland); Dmitry Kuznetsov (Swiss Institute of Bioinformatics, Switzerland); Nicolas Guex (Swiss Institute of Bioinformatics, Switzerland); Christian Iseli (Swiss Institute of Bioinformatics, Switzerland); Heinz Stockinger (Swiss Institute of Bioinformatics, Switzerland); Daniel Zerzion (Swiss Institute of Bioinformatics, Switzerland); Ioannis Xenarios (Swiss Institute of Bioinformatics, Switzerland)
  
  The PoSeNoGap project focuses on the design and implementation of efficient genomic data formats and tools for genome analysis applications to scale to large datasets. The proposed solutions shall exploit the available parallelism of the platforms operated by PASC partners. The project promotes investigations on current issues in genomic data representation, compression and transmission within MPEG, an ISO/IEC working group specialized in digital media representation and compression. This activity has led to the definition of reference genomic datasets used to assess the performance of existing bioinformatic tools and to identify requirements for the standard representation of genome data.
• Zipoli Federico Contributed Talk
  Wednesday, June 3, 2015
  HG E1.2, 11:10-11:30
  
  Contributed Talk
  
  Molecular-Dynamics Simulations of Resistance Switch in Amorphous Carbon, Federico Zipoli (IBM Research, Switzerland)
  
  Co-Authors: Alessandro Curioni (IBM Research, Switzerland)
  
  Diamond-like carbon films are considered very promising materials for resistive random access memories. The information is stored by setting different level of electric resistance in amorphous carbon (a-C). An electric pulse is used to switch between the low and the high resistance states. The current challenge is to ensure reversible switch in a-C. We use atomistic simulations at the realistic device sizes to find which structural changes produce the resistance switch and its associated optimal process conditions. This work is funded by the EU research & innovation project CareRAMM, N. 309980.