Running Research and Development Projects
Excellence Initiative: IAS
The Institute for Advanced Study (IAS) of Technische Universität München is the centerpiece of TUM’s institutional strategy to promote top-level research in the so-called Excellence Initiative by the German federal and state governments.
HPC - Tackling the Multi-Challenge
| Project type | IAS focus group |
| Funded by | Excellence Initiative of the German federal and state and governments |
| Begin | 2010 |
| End | 2015 |
| Leader | Univ.-Prof. Dr. Hans-Joachim Bungartz |
| Staff | Dr. rer. nat. habil. Miriam Mehl, Dr. rer. nat. Dirk Pflüger, Christoph Kowitz, M.Sc., Valeriy Khakhutskyy, M.Sc., Dipl.-Math. Benjamin Uekermann, Arash Bakhtiari, M.Sc. (hons) |
| Contact person | Dr. rer. nat. habil. Miriam Mehl |
| Co-operation partner | Prof. George Biros (Georgia, USA), Markus Hegland (Canberra, Australia) |
Brief description
High-performance computing (HPC) is a thriving cross-sectional research field of utmost relevance in science and engineering. Actually, scientific progress is more and more depending on insight gained by computational research. With the increased technological potential, however, the requirements are growing, too – which leads to several computational challenges, which are all
related to some “multi-X” notion: multi-disciplinary, multi-physics, multi-scale, multi-dimensional, multi-level, multi-core. This focus group will
primarily address the three topic multi-physics (mp), multi-dimensional (md), and multi-core (mc).
The interplay of these three subtopics is straightforward: Both mp and md are among the usual suspects that need and, thus, drive HPC technology and mc; mp frequently appears in the context of optimisation or parameter identification or estimation – thriving topics of current md research; and present as well as future mc technology is inspired by algorithmic patterns, as provided by mp and md. Hence, it is not only reasonable to address mp, md, and mc in an integral way, it is essential, and this IAS focus group offers the unique chance of doing this at a very high international level.
Bayern Excellent: MAC@IGSSE
The Munich Centre of Advanced Computing (MAC) is a research consortium which has been established at TUM to bundle research activities related to computational science and engineering (CSE) as well as high-performance computing (HPC) - across disciplines, across departments, and across institutions. In MAC, seven of TUM's departments and other Munich research institutions (Ludwig-Maximilians-Universität, Max-Planck insititutes, the Leibniz Supercomputing Centre of the Bavarian Academy of Sciences and Humanities) as well as TUM's international partners such as KAUST, the King Abdullah University of Science and Technology, join their forces to ensure the sustainable usage of current and future HPC architectures for the most relevant and most challenging CSE applications.
Efficient Parallel Strategies in Computational Modelling of Materials
| Project type | Förderprogramm ”Bayern exzellent”: Munich Centre of Advanced Computing (MAC) |
| Funded by | Bavarian state government, Technische Universität München |
| Begin | 2008 |
| End | 2012 |
| Leader | Prof. Dr. Dr. h.c. Notker Rösch subproject: Univ.-Prof. Dr. Hans-Joachim Bungartz |
| Staff | Martin Roderus |
| Contact person | Martin Roderus |
| Co-operation partner | Prof. Dr. Dr. h.c. Notker Rösch, Prof. Dr. Arndt Bode, Prof. Dr. Michael Gerndt, Prof. Dr. Heinz-Gerd Hegering |
Brief description
The project will develop a new paradigm for the parallelisation of density functional theory (DFT) methods for electronic structure calculations and implement this new strategy. Advanced embedding techniques will account for environment effects (e.g. solvent, support) on a system, which requires a strong modularisation of the DFT approach, facilitating task specific parallelisation, memory management, and low-level optimisation. Efficiency will be further increased by dynamical adaptation of varying resource usage at module level and pooling of applications.
A High-End Toolbox for Simulation and Optimisation of Multi-Physics PDE Models
| Project type | Förderprogramm ”Bayern exzellent”: Munich Centre of Advanced Computing (MAC) |
| Funded by | Bavarian state government, Technische Universität München |
| Begin | 2008 |
| End | 2012 |
| Leader | Prof. Dr. Michael Ulbrich subproject: Univ.-Prof. Dr. Hans-Joachim Bungartz, Dr. rer. nat. Miriam Mehl |
| Staff | Janos Benk, M.Sc |
| Contact person | Dr. rer. nat. Miriam Mehl |
| Co-operation partner | Prof. Dr. Michael Ulbrich, Prof. Dr. Martin Brokate, Prof. Dr. Ernst Rank, Prof. Dr. Ronald Hoppe (Augsburg) |
Brief description
The project aims at bundling forces to overcome conceptional drawbacks of current simulation software and to make a big step towards a future generation of simulation and optimisation tools for complex systems. The goal is to develop a rapid prototyping HPC software platform for both simulation and optimisation. The design will be hierarchical, with high performance components on all levels, ranging from problem formulation via discretisation to numerics and parallelisation. Work will be interwoven with theoretical investigations of innovative numerical algorithms.
A Scalable Infrastructure for Computational Steering
| Project type | Förderprogramm ”Bayern exzellent”: Munich Centre of Advanced Computing (MAC) |
| Funded by | Bavarian state government, Technische Universität München |
| Begin | 2008 |
| End | 2012 |
| Leader | Prof. Dr. Rüdiger Westermann subproject: Univ.-Prof. Dr. Hans-Joachim Bungartz |
| Staff | Daniel Butnaru, M.Sc |
| Contact person | Univ.-Prof. Dr. Hans-Joachim Bungartz |
| Co-operation partner | Prof. Dr. Rüdiger Westermann, Prof. Bernd Brügge, Ph.D., Prof. Dr. Ernst Rank, Prof. Dr.-Ing. Wolfgang Wall |
Brief description
The goal of this project is to design and prototype a scalable infrastructure for computational steering. It will be targeted for the computational engineering domain, which allows to leverage existing cooperative developments as a starting point and to use real-world data that is representative in size, modality, and structure to what is available in other scientific areas like geology or biology. The infrastructure implements a processing pipeline ranging from scalable data processing workflows to interactive visualisation and human-computer interaction in virtual and augmented reality environments.
TUM-KAUST Strategic Partnership: MAC@KAUST
Simulation of CO2 Sequestration
| Project type | Strategic Partnership with the King Abdullah University of Science and Technology (KAUST)] |
| Funded by | KAUST |
| Begin | 2009 |
| End | 2013 |
| Leader | Univ.-Prof. Dr. Hans-Joachim Bungartz |
| Staff | see Munich Centre of Advanced Computing |
| Contact person | Tobias Weinzierl |
| Co-operation partner | Prof. Dr. Dr.-Ing. habil. Arndt Bode (Computer Architecture), Prof. Dr. Martin Brokate (Numerical Mathematics and Control Theory), Prof. Dr. Drs. h.c.Karl-Heinz Hoffmann (Numerical Mathematics and Control Theory), Prof. Dr.-Ing. Michael Manhart (Hydromechanics), Prof. Dr. Michael Ulbrich (Mathematical Optimisation) |
Brief description
The goal of this project is to design and investigate novel approaches to modelling and simulation of CO2 sequestration processes, in particular in the context of enhanced oil recovery. The project will involve both fine-grain simulations - with all related aspects from multi-phase schemes via numerical algorithmics to high-performance computing issues - and homogenization approaches to efficiently capture the fine-grain effects on the macro-scale. For that, groups with expertise in flow physics, mathematical modelling, numerical analysis, numerical algorithmics, optimisation and inverse problems, and high-performance computing and HPC systems join their forces. Topics addressed will cover multi-scale modelling and homogenisation, fully-resolved pore-scale simulation, constrained optimisation of the sequestration process, enhanced numerics and parallelisation, and HPC implementation.
Virtual Arabia
| Project type | Strategic Partnership with the King Abdullah University of Science and Technology (KAUST)] |
| Funded by | KAUST |
| Begin | 2009 |
| End | 2013 |
| Leader | Tobias Weinzierl |
| Staff | see Munich Centre of Advanced Computing |
| Contact person | Univ.-Prof. Dr. Hans-Joachim Bungartz |
| Co-operation partner | Prof. Dr. Dr.-Ing. habil. Arndt Bode (Computer Architecture), Prof. Gudrun Klinker, Ph.D. (Augmented Reality), Prof. Dr. Ernst Rank (Computation in Engineering), Prof. Dr. Rüdiger Westermann (Computer Graphics & Visualization) |
Brief description
The goal of this project is to develop a virtual environment for the interactive visual exploration of Saudi Arabia. In contrast to virtual globe viewers like Google Earth, this environment will allow the user to look both above and underneath the earth surface in an integrated way. It will, thus, provide interactive means for the visual exploration of 3D geological structures and dynamic seismic processes as well as atmospheric processes and effects or built or planned infrastructure. The specific techniques required to support such functionality will be integrated into a generic infrastructure for visual computing. The project will cooperate with the KAUST 3D Modelling and Visualisation Centre and the KAUST Computational Earth Sciences Centre.
High Performance Visual Computing
| Project type | Strategic Partnership with the King Abdullah University of Science and Technology (KAUST)] | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Funded by | KAUST | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Begin | 2012 | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| End | 2015 | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Leader | Philipp Neumann, Tobias Weinzierl | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Staff | see Munich Centre of Advanced Computing | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Contact person | Univ.-Prof. Dr. Hans-Joachim Bungartz | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Co-operation partner | }
Brief description The project combines fundamental methodological research in the field of high performance computing (HPC) in a unique way with data exploration on HPC devices and the question how to cross-fertilize seamlessly into applications used at KAUST and TUM to obtain new insight from supercomputing - today and in the upcoming exascale age. It comprises three major goals: First, we ensure the sustainability of some work conducted under the umbrella of “Simulating CO2 Sequestration” (1), as codes stemming from KAUST faculty but extended by TUM project members and associates are prepared for the upcoming generation of supercomputers besides the KAUST facilities. Second, we combine visualization techniques and supercomputing paving the way to interactive, immersive simulation and computational steering. This endeavor both brings together insights from "Virtual Arabia" (2) and researchers with a supercomputing and algorithmic affinity and it uses synergies from both KAUST’s visualization and supercomputing laboratories. Such an endeavor will pay off for future research both at KAUST and TUM, when insight is not obtained in a batch mode as it is today, but problems and phenomena have to be studied interactively. Third, we integrate research results obtained with TUM codes into KAUST applications as well as codes from the “Virtual Arabia” project, i.e. we demonstrate the broader applicability of work done under the umbrella of the KAUST-TUM special partnership projects. DFG - German Research FoundationPriority Program 1648 SPPEXA - Software for Exascale ComputingCoordination Project
Brief description The Priority Programme (SPP) SPPEXA is different from other SPP with respect to its genesis, its volume, its funding via DFG's Strategy Fund, with respect to the range of disciplines involved, and to a clear strategic orientation towards a set of time-critical objectives. Therefore, despite its distributed structure, SPPEXA also resembles a Collaborative Research Centre to a large extent. Its successful implementation and evolution will require both more and more intense structural measures. The Coordination Project comprises all intended SPPEXAwide activities, including steering and coordination, internal and international collaboration and networking, and educational activities. Reference: Priority Program 1648 SPPEXA - Software for Exascale Computing ExaFSA - Exascale Simulation of Fluid-Structure-Acoustics Interaction
Brief description In scientific computing, an increasing need for ever more detailed insights and optimization leads to improved models often including several physical effects described by different types of equations. The complexity of the corresponding solver algorithms and implementations typically leads to coupled simulations reusing existing software codes for different physical phenomena (multiphysics simulations) or for different parts of the simulation pipeline such as grid handling, matrix assembly, system solvers, and visualization. Accuracy requirements can only be met with a high spatial and temporal resolution making exascale computing a necessary technology to address runtime constraints for realistic scenarios. However, running a multicomponent simulation efficiently on massively parallel architectures is far more challenging than the parallelization of a single simulation code. Open questions range from suitable load balancing strategies over bottleneck-avoiding communication, interactive visualization for online analysis of results, synchronization of several components to parallel numerical coupling schemes. We intend to tackle these challenges for fluid-structure-acoustics interactions, which are extremely costly due to the large range of scales. Specifically, this requires innovative surface and volume coupling numerics between the different solvers as well as sophisticated dynamical load balancing and in-situ coupling and visualization methods. Reference: Priority Program 1648 SPPEXA - Software for Exascale Computing EXAHD - An Exa-Scalable Two-Level Sparse Grid Approach for Higher-Dimensional Problems in Plasma Physics and Beyond
Brief description Higher-dimensional problems (i.e., beyond four dimensions) appear in medicine, finance, and plasma physics, posing a challenge for tomorrow's HPC. As an example application, we consider turbulence simulations for plasma fusion with one of the leading codes, GENE, which promises to advance science on the way to carbon-free energy production. While higher-dimensional applications involve a huge number of degrees of freedom such that exascale computing gets necessary, mere domainde composition approaches for their parallelization are infeasible since the communication explodes with increasing dimensionality. Thus, to ensure high scalability beyond domain decomposition, a second major level of parallelism has to be provided. To this end, we propose to employ the sparse grid combination scheme, a model reduction approach for higher-dimensional problems. It computes the desired solution via a combination of smaller, anisotropic and independent simulations, and thus provides this extra level of parallelization. In its randomized asynchronous and iterative version, it will break the communication bottleneck in exascale computing, achieving full scalability. Our two-level methodology enables novel approaches to scalability (ultra-scalable due to numerically decoupled subtasks), resilience (fault and outlier detection and even compensation without the need of recomputing), and load balancing (high-level compensation for insufficiencies on the application level). Reference: Priority Program 1648 SPPEXA - Software for Exascale Computing SFB-TRR 89: Invasive Computing
Brief description In the proposed CRC/Transregio, we intend to investigate a completely novel paradigm for designing and programming future parallel computing systems called invasive computing. The main idea and novelty of invasive computing is to introduce resource-aware programming support in the sense that a given program gets the ability to explore and dynamically spread its computations to neighbour processors similar to a phase of invasion, then to execute portions of code of high parallelism degree in parallel based on the available (invasible) region on a given multi-processor architecture. Afterwards, once the program terminates or if the degree of parallelism should be lower again, the program may enter a retreat phase, deallocate resources and resume execution again, for example, sequentially on a single processor. In order to support this idea of self-adaptive and resource-aware programming, not only new programming concepts, languages, compilers and operating systems are necessary but also revolutionary architectural changes in the design of MPSoCs (Multi-Processor Systems-on-a-Chip) must be provided so to efficiently support invasion, infection and retreat operations involving concepts for dynamic processor, interconnect and memory reconfiguration. Reference: Transregional Collaborative Research Centre 89 - Invasive Computing A4: Design-Time Characterisation and Analysis of Invasive Algorithmic Patterns
D3: Invasion for High Performance Computing
G8-Initiative: Nuclear Fusion Simulations at Exascale (Nu-FuSe)
Brief description To exploit upcoming exascale systems effectively for fusion modelling creates significant challenges around scaling, resiliency, result validation and programmability. This project will be focusing on meeting these challenges by improving the performance and scaling of community modelling codes to enable simulations orders of magnitude larger than are currently undertaken. BMBF: Federal Ministry of Education and ResearchMultikOSI/MEPKA
Events like "public viewing", festivals or concerts are an important part of urban life and must be managed to ensure safety and security. Yet, event managers and security personal lack scientifically validated and practical instruments. The MultikOSi project combines knowledge and competencies from areas such as crowd management, mathematics, informatics, sociology and civil engineering. The goal is to understand the general processes at urban mass events and, with that knowledge, to develop new methods to improve safety. Part of this are new models of pedestrian dynamics, including sociological and psychological aspects, and new combinations of existing models. The need to consider safety, openness and economics at mass events make the task a multi-criterial optimization process. The holistic and interdisciplinary approach will lead to an empowered planning phase and optimized security concepts for events. The scientific results can be used after the project to develop software tools for event planning. The MEPKA project at Munich University of Applied Sciences investigates mathematical properties of state-of-the-art pedestrian locomotion models to both improve numerical performance of these models and develop new models that are more robust and more consistent with empirical evidence. Based on that, first models incorporating psychological aspects, like self categorization, are devised and validated. Non-Linear Characterization and Analysis of FEM Simulation Results for Motor-Car Components and Crash Tests (SIMDATA-NL)
Brief description The project aims at the extraction of the (few) effective dimensions in high-dimensional simulation data in the context of automotive design. Linear methods, like the principal component analysis, alone are not sufficient for many of those applications due to significant non-linear effects. Therefore, they will be complemented by methods that are able to resolve nonlinear relationships, especially by means of sparse grid discretizations. SkaSim: Scalable HPC-Software for molecular simulation in the chemical industry
Brief description Molecular dynamics (MD) and Monte-Carlo (MC) simulations form the basis for investigating many relevant application scenarios in science and engineering. At the heart of these simulations lie physically meaningful and quantitative models of molecular interactions, requiring precise validation through state of the art ab initio calculations and experimental data. The extreme spatial and temporal resolution (individual molecules, femtoseconds) of such simulations allow for very reliable predictions of material properties, even where experiments are impossible or dangerous. However, this extreme resolution also implies substantial computational demands in order to investigate scenarios in a timely manner. The same holds true for nanofluidics: realistic insights, not obtainable experimentally, can be captured through simulation. Complex phenomena as for instance phase transitions (e.g. condensation) can be investigated on the molecular level, allowing new and more fundamental insights. However, as the dynamics of every molecule is evaluated explicitly, the number of simulated molecules needs to be considerable in order to capture the phenomena in question. Determining experimentally elusive properties of matter is attracting increasing attention from industry. Be it in process engineering, where the already highly optimized procedures can only be improved through better and more detailed data and understanding. The computational power required to generate the quantity and quality of data required is significant. Thus, only through the efficient use of cutting-edge hardware can these demands be met. However, many relevant scenarios are far from trivial to simulate at scale, e.g. coinciding fluid and gaseous phases in a highly dynamic environment as in condensation or evaporation. However, the industrial development of new products and processes will experience a fundamental change in the coming years. Expensive and oftentimes dangerous experiments can be replaced with safe and increasingly efficient and affordable simulations. For this transition to take place, simulations need to be performed with accuracies comparable to highquality experiments. Besides the computational requirements, this calls for extremely accurate molecular models and, for complex scenarios, reliable new methodologies. The challenges to simulate such scenarios efficiently are huge and will be addressed in SkaSim. HEPP: International Helmholtz Graduate School for Plasma Physics
Brief description Given the multiscale nature of most problems of interest, advanced algorithms and efficient implementations on massively parallel platforms are usually required in order to tackle them. In this context, a close collaboration of theoretical plasma physicists with applied mathematicians and computer scientists can be of great benefit. Thus, state-of-the-art numerical techniques, hardware-aware implementation strategies, and scalable parallelization approaches are explored in terms of their potential to minimize the overall computational requirements and to maximize the reliability and robustness of the simulations.
Volkswagen Stiftung: ASCETE, ASCETE-II (Advanced Simulation of Coupled Tsunami-Earthquake Events)
Brief description Earthquakes and tsunamis represent the most dangerous natural catastrophes and can cause large numbers of fatalities and severe economic loss in a single and unexpected extreme event as shown in Sumatra in 2004, Samoa in 2009, Haiti in 2010, or Japan in 2011. Both phenomena are consequences of the complex system of interactions of tectonic stress, fracture mechanics, rock friction, rupture dynamics, fault geometry, ocean bathymetry, and coast line geometry. The ASCETE project forms an interdisciplinary research consortium that – for the first time – will couple the most advanced simulation technologies for earthquake rupture dynamics and tsunami propagation to understand the fundamental conditions of tsunami generation. To our knowledge, tsunami models that consider the fully dynamic rupture process coupled to hydrodynamic models have not been investigated yet. Therefore, the proposed project is original and unique in its character, and has the potential to gain insight into the underlying physics of earthquakes capable to generate devastating tsunamis. Intel Parallel Computing Center: Extreme Scaling on x86/MIC (ExScaMIC)
Brief description The project is optimizing four different established or upcoming CSE community codes for Intel-based supercomputers. We assume a target platform that will offer several hundred PetaFlop/s based on Intel's x86 (including Intel® Xeon Phi™ coprocessors) architecture. To prepare simulation software for such platforms, we tackle two expected major challenges: achieving a high fraction of the available node-level performance on (shared-memory) compute nodes and scaling this performance up to the range of 10,000 to 100,000 compute nodes. We examine four applications from different areas of science and engineering: earthquake simulation and seismic wave propagation with the ADER-DG code SeisSol, simulation of cosmological structure formation using GADGET, the molecular dynamics code ls1 mardyn developed for applications in chemical engineering, and the software framework SG++ to tackle high-dimensional problems in data mining or financial mathematics (using sparse grids). While addressing the Xeon Phi™ coprocessor, in particular, the project tackles fundamental challenges that are relevant for most supercomputing architectures – such as parallelism on multiple levels (nodes, cores, hardware threads per core, data parallelism) or compute cores that offer strong SIMD capabilities with increasing vector width. Elite Network of Bavaria (ENB):Bavarian Graduate School of Computational Engineering (BGCE)
Brief description The Bavarian Graduate School of Computational Engineering is an association of the three Master programs: Computational Engineering (CE) at the University of Erlangen-Nürnberg, Computational Mechanics (COME), and Computational Science and Engineering (CSE), both at TUM. Funded by the Elitenetzwerk Bayern, the Bavarian Graduate School offers an Honours program for gifted and highly motivated students. The Honours program extends the regular Master's programs by several academic offers:
Students who master the regular program with an above-average grade, and successfully finish the Honours program, as well, earn the academic degree "Master of Science with Honours". Numerical Aspects of the Simulation of Quantum Many-body Systems
Brief description In the last years a growing attention has been dedicated to many body quantum systems from the point of view of quantum information. Indeed, after the initial investigation of simple systems as single or two qubits, the needs of understanding the characteristics of a realistic quantum information device leads necessary to the study of many body quantum systems. These studies are also driven by the very fast development of experiments which in the last years reach the goal of coherent control of a few qubits (ion traps, charge qubits, etc...) with a roadmap for further scaling and improvement of coherent control and manipulation techniques. Also, new paradigm of performing quantum information tasks, such as quantum information transfer, quantum cloning and others, without direct control of the whole quantum system but using our knowledge of it has increased the need of tools to understand in details the behaviour of many body quantum system as we find them in nature. These new goals of the quantum information community lead to an unavoidable exchange of knowledge with other communities that already have the know-how and the insight to address such problems; for example the condensed matter, computational physics or quantum chaos communities. Applying known techniques and developing new ones from a quantum information perspective have already produced fast and unexpected developments in these fields. The comprehension of many body quantum systems ranging from few qubits to the thermodynamical limit is thus needed and welcome not only to develop useful quantum information devices, but it will lead us to a better understanding of the quantum world. Reference: Computations in Quantum Tensor Networks KONWIHR (Bavarian Competence Network for Technical and Scientific High Performance Computing):Optimization of Dense and Sparse Matrix Kernels for SeisSol on SuperMUC
Brief description SeisSol is one of the leading simulation codes for earthquake scenarios, in particular for accurate simulation of dynamic rupture processes. In the proposed project, we optimize the performance of SeisSol via a code generation approach. In a two-step procedure, the set-up of element matrices (which are used to express SeisSol's innermost kernel operations) are extracted, and optimized kernel implementations are generated (exploiting SIMD operations, register blocking, etc.) and integrated into SeisSol. Performance evaluation will be done on the SuperMUC platform. MISTI MIT-TUM Project: Combining Model Reduction with Sparse Grids into a Multifidelity Framework for Design, Control and Optimization
Brief description Many engineering problems require repeated simulations in order to model and optimize a real life system. Such models are typically quite complex and a single solution usually involves a huge computational effort. If a large number of such expensive solutions is needed, the models become impractical and alternatives are sought, with the goal of enabling interactive and highly reliable high-accuracy simulations. Surrogate models mimic the behavior of the simulation model as closely as possible and are at the same time computationally much cheaper to evaluate. While certain surrogate methods exist and perform well for specific problems, their acceptance is slowed by their complex and intrusive manner. They need to be reconsidered for each problem class and are sensitive to the characteristics of the underlying simulation. In this project we open a collaboration between MIT and TUM in the area of model reduction with an initial focus on non-intrusive methods. These treat the simulation as a black box and, based only on a number of snapshots, deliver an approximation which can than be efficiently queried. The joint work will combine MIT’s model-reduction techniques with TUM’s sparse grid methods with the goal of delivering a novel non-intrusive model reduction technique. |