A common problem in numerical optimization and sampling is the detection of relevant states. These could be, for instance, the local minima on a rugged parameter surface or the transition state of a chemical reaction. For most cases, an exhaustive search for the optimal solution is intractable. Here, we focus on parallel sampling and optimization strategies relying on multiple replicas, most prominently, adaptive methods where all simulated replicas use the same propagator and sample the same underlying surface. In these methods, replica intercommunication is used to provide a global assessment as to which replicas are most interesting. This implies, in general, periodic data mining steps across replicas. Furthermore, in order to extract and utilize the gained information in post-processing, data must often be stored, which poses stringent data management and analysis challenges in particular for high-dimensional cases. The minisymposium wishes to discuss the following questions: What are meaningful and easily generalizable tools, strategies, and algorithms to guide the sampling/exploration? How can we maintain scalability and load balance? What types of post-processing algorithms can be applied to the generated data, and are those scalable to provide on-the-fly solutions to direct the exploration?

Selected PASC projects will present the scientific challenge they aim to solve by using high-end supercomputers, and in particular the computational approach adopted. The topics include astrophysics (smooth particle hydrodynamics), numerical weather and climate (stencils and grids), linear algebra for electronic structure (sparse matrix and tensor operations), and biomedical applications (fluid-structure interaction, machine learning).

The focus will be on aspects concerning (i) capability computing: how to scale to several hundreds/thousands of compute nodes, including the use of communication optimal algorithms and asynchronous communication; (ii) performance portability: how to address the growing diversity in hardware on a compute node, including generic software design, and auto-tuning; and (iii) co-design in these projects: how to engage with vendors to optimally exploit current hardware, and to provide feedback that has or will influence next-generation hardware.

Topics include: side by side comparisons of multi-core, many core, and GPU compute nodes; optimization techniques for flops, memory bandwidth, or network performance; JIT compilation of machine specific kernels; programming approaches such as the use of domain specific languages (DSLs), remote memory access (RMA), task based programming.

Discrete-time, infinite-horizon, general equilibrium models are routinely used in macroeconomics and in public finance for exploring the quantitative features of model economies and for counterfactual policy analysis. One important question concerns the importance of household heterogeneity for the amplification and propagation of macroeconomic shocks.

In this session we bring together leading young researchers in the field to present alternative approaches to the computation of equilibria in dynamic stochastic models with heterogeneous agents and/or with financial frictions. Three of the papers directly propose new methods for the solution of models with a continuum of ex post heterogeneous agents.

The power of artificial neural nets at executing challenging tasks learned from data is very attractive to fields of physical science and especially to High Energy Physics. In recent years, there have been a significant number of articles reporting promising results with applying deep learning to HEP challenges. By virtue of the very large number of parameters of artificial neural nets, with deep and wide architecture, trained with stochastic gradient descent, it is mandatory to process a lot of representative data in order to obtain accurate models. Training of such models requires a tremendous amount of computing, and commonly takes days, if not weeks, to converge. GP-GPU technology has enabled a lot of this computation, thanks to the high level of parallelisation of the formalism of artificial neural net, but more however can be gained in parallelised calculation of the stochastic gradient descent. Supercomputing facilities are particularly suited for distributed training of deep neural nets, thanks to their large computation power and excellent connectivity. This minisymposium will address the current state-of-the-art and present performances in training models for high energy physics, with a particular view to software availability and to foster further utilization of supercomputers.

Achieving hardware and energy efficiency is important for current large-scale numerical simulations and will be a key component in the exascale era. In a world of heterogeneous, highly parallel computer architectures with deep memory hierarchies, complex application scenarios, and a broad spectrum of algorithms, a thorough analysis and understanding of the complex interaction of software, data structures, algorithms, and hardware features, a.k.a. performance engineering, is required for implementing codes that allow for portable performance on the computer generations to come. The minisymposium addresses a broad range of topics in performance engineering for modern HPC architectures, ranging from recent advances in performance models and tools supporting a "white-box" performance engineering approach to application performance tuning cases studies and "black-box" solutions. The presentations will point out the potentials and limitations of performance engineering activities and demonstrate the wide spectrum of performance models used in the performance engineering, including simple performance expectations, automatic model parameter selections, and analytic models.

Modern electronic-structure methods provide parameter-free simulations of properties across the whole spectrum of Physics, Chemistry, and Materials Science. They have become a bedrock of advanced analytical methods and are used systematically to interpret advanced experiments and complex interactions, with ever growing perspectives for more "realistic" systems, including defects, thermal and external fields, and transient phenomena. This minisymposium explores the next generation of electronic-structure software, which will lead users to exascale supercomputers, through highly efficient and highly parallel algorithms. We will showcase recent advances in ground- and excited-state calculations, spectroscopic quantities and transport, and adaptive methods which exploit different algorithms for different systems such as periodic/localized or many/few electrons per atom.

The increasingly large amounts of data being produced by weather and climate simulations and earth system observations is sometimes characterised as a deluge. This deluge of data is both a challenge and an opportunity. The main opportunities are to make use of this wealth of data to 1) improve knowledge by extracting additional knowledge from the data and 2) to improve the quality of the models themselves by analysing the accuracy, or lack thereof, of the resultant simulation data. An example of the former case is improved prediction of large scale phenomena such as El Nino. An example of the latter is the improvement of a Physics parameterisation scheme through detailed analysis of the errors in a large number of datasets.

One way to realise these opportunities is to use machine learning approaches. As machine learning in weather and climate is a relatively new topic this minisymposium introduces the audience to how machine learning could be used in weather and climate and outlines its implications in terms of computing costs. To ground the ideas in concrete examples it also illustrates the use of machine learning in the weather and climate domain with practical examples.

Wall resolved LES is prohibitively expensive, and a first principles based wall modeled LES, that is free of tunable parameters, is far from becoming an alternative to RANS as a predictive tool for design. On the spatial discretization front, most higher-order and spectral methods have had many of their successes in the realm of low to moderate Reynolds number flows, and on relatively less complex geometries. The bulk of the complex flow high Re simulations still employ nominally second-order accurate schemes in unstructured mesh finite volume flow solvers, and finite-element based solvers with modest polynomial order. Even in these fairly mature solvers, evidence suggests that merely running larger cases with an increased number of grid points and on larger computational domains does not always guarantee a better solution (in the LES sense). There is, therefore, a need for a two-level effort whereby benchmark higher-order simulations can serve to inform sub-grid models to improve the predictive capability of existing mature flow solvers. We hope to motivate a discussion along these lines with examples of results of higher-order simulations, as well as their role in assessing and improving the predictive capabilities of sub-grid models with more conventional flow solvers.

A common problem in numerical optimization and sampling is the detection of relevant states. These could be, for instance, the local minima on a rugged parameter surface or the transition state of a chemical reaction. For most cases, an exhaustive search for the optimal solution is intractable. Here, we focus on parallel sampling and optimization strategies relying on multiple replicas, most prominently, adaptive methods where all simulated replicas use the same propagator and sample the same underlying surface. In these methods, replica intercommunication is used to provide a global assessment as to which replicas are most interesting. This implies, in general, periodic data mining steps across replicas. Furthermore, in order to extract and utilize the gained information in post-processing, data must often be stored, which poses stringent data management and analysis challenges in particular for high-dimensional cases. The minisymposium wishes to discuss the following questions: What are meaningful and easily generalizable tools, strategies, and algorithms to guide the sampling/exploration? How can we maintain scalability and load balance? What types of post-processing algorithms can be applied to the generated data, and are those scalable to provide on-the-fly solutions to direct the exploration?

Numerical weather prediction and climate models are large and complex software applications that need to run efficiently on today's and future massively parallel computer systems. The rapid change in these computing architectures and the increase in diversity are seriously affecting the ability to retain a single source code that runs efficiently in different architectures. Several weather models have successfully adapted their codebases to many-core and heterogeneous architectures like GPUs and Xeon Phi using a combination of multiple traditional programming models for parallel architectures like OpenMP, OpenACC and MPI. However porting existing large community codes to multiple architectures is a daunting task and leads to codes that are more complex and difficult to maintain. As a result in the past years numerous new technologies and approaches are emerging in order to provide new programming models, like domain-specific languages (DSLs) or source-to-source translation tools that can increase the productivity of development in weather codes while providing a high degree of performance portability. In this minisymposium we propose a discussion with some of the most prominent novel approaches where the new advances in programming models used for heterogeneous architectures in weather and climate models will be presented.

Many important economic phenomena relate to the notion of risk. Economic actors not only make decisions as a function of their current situation but also depending on their expectation of future developments. Because economic systems are not deterministic, future economic events are usually treated as stochastic phenomena. The general form of the problem at hand is to determine how the probabilistic distribution of future economic events influences current decisions. The wider the distribution, the more risk in today's decisions. The papers in this session present different computation challenges involved in attempting to describe the effect of risk on economic decisions.

The dawn of Web 2.0 applications, smartphones/tablets and the omnipresent yet invisible cloud computing have dramatically changed our perception of the IT landscape in the last decade. At the same time these technologies delivered an abundance of new tools that are more suited to the needs of domain scientists: GitHub/GitLab with their inherent support for agile programming, user space package managers for easy software installations even when facing complex dependencies, and web portals to HPC systems or private clouds so that no prior knowledge is needed to use state-of-the-art compute resources. This minisymposium will showcase all these tools and how they are used in real scientific workflows. The shown best practices are easy to reproduce since they are all based on freely available software packages, so that the audience can use the presentations both as an inspiration but also as a kickstart for their own better science.

The Large Hadron Collider at CERN is smashing high density bunches of protons near the speed of light at a frequency of 40 MHz. Most of the thousands of particles emitted at each bunch crossing are measured and collected with building-sized detectors consisting of multiple sub-detectors each serving its own purpose. The simulation of the signal created by a particle interacting with such a detector is typically done with very detailed simulations and needs to be stepped infinitesimally over meters of material. This simulation is as much computing intensive as the geometry is complex. In current and future detector design, the fine-grained simulation of such a detector is taking a great part of the full computing budget of experiments and poses a computing challenge. While a great deal of effort is being made to parallelise such software, one possible avenue to reduce the computational requirements is with generative models from the field of deep learning. Such generative models have seen success in conditionally generating images and video of various types. We present how such models are built and trained, and how well they can capture the physics of particle interaction and help generate realistic samples for high energy physics analysis.

In the past couple years a large number of fintech and more recently insurtech startups have been founded and are challenging established players in the financial services sector. At Baloise – a Swiss company providing insurance services in Switzerland, Belgium, Germany and Luxembourg as well as banking services in Switzerland – we view startups as potential partners on our digital transformation journey rather than competition. This minisymposium aims to demonstrate what problems companies such as Baloise face in terms of digitizing their business and making use of their large amounts of data. To do so, the four sessions will cover the innovation framework Baloise employs in order to rapidly test prototypes, a presentation by Brainalyzed, a startup which aims to optimize and automatize investment decisions at Baloise using AI, a presentation about the challenges arising in the context of data warehouses and legacy systems, and a panel discussion with all the speakers.

Machine Learning is right now a booming field of computer science which finds applications in the development of computer-human interfaces, in the analysis of medical data of huge populations, in the maintenance of cars, planes and elevators, and self-driving cars, to mention a few. During the last twenty years the field has gone through a development and refinement which has been spectacular. For some reason the use of the technology in pure science has been lagging behind; however, we are now starting to see the use of machine learning in the field of quantum chemistry. Here, the approach will enhance the performance of standard quantum chemical calculations – improving convergence, could serve as a tool for post-analysis of huge sets of ab initio results, or could simply replace computationally expensive procedures. Machine learning offers practical alternatives where standard quantum chemical simulations would be prohibitive. During the last few years a small number of quantum chemistry groups have explored the potential of machine learning – the results have been extraordinary and spectacular. Here in this minisymposium we would like to inspire by presenting four different applications in which machine learning is fundamental to success.

Difficult combinatorial problems permeate virtually every area of the sciences, business, and government and many of these problems can be cast as mixed-integer programs (MIPs). A MIP is a mathematical definition of a problem that is comprised of a set of constraints and an objective function. In general, MIPs are NP-hard and require exponential amounts of computation time in the worst case. However, search strategies, such as branch-and-bound, branch-and-cut, and cut-and-solve have evolved to provide optimal solutions for many instances.

Although there has been great progress in this field and computational power has dramatically increased over the years, many important MIPs remain intractable and the use of massive parallelization appears to be a promising means to address this great need. However, many challenges lie ahead. This minisymposium will elucidate some of these challenges, while highlighting progress in this field. It includes a round table discussion with Michael Chan, Sharlee Climer, Daniel Jacobson, Sarah Powers, and Daniel Rehfeldt, and is open to conference participants. The goal of the discussions will be to explore and integrate high-performance expertise with domain-specific insights with an aim to identify strategies that may resolve these pressing challenges.

Computational Fluid Dynamics (CFD) is a natural driver for exascale computing both for academic and industrial cases, and has the potential for substantial societal impact, like reduced energy consumption, alternative sources of energy, improved health care, and improved climate models. This minisymposium focuses on algorithms and methods applicable on the way to exascale for CFD simulations. Application cases were discussed in Part I, whereas in Part II we focus on some of the relevant methodological aspects. The main driver is the EU funded Horizon 2020 project ExaFLOW and will feature presentations showcasing their work on addressing key algorithmic challenges in CFD in order to facilitate simulations at exascale, e.g. accurate and scalable solvers, data reduction methods (compression) and strategies to ensure fault tolerance and resilience. In particular, the talks in this minisymposium will highlight the following topics: Adaptive mesh refinement and adjoint-based error estimators, resilience in transient flow solvers, efficient communication operators using PGAS and mixed CG-HDG formulations for higher-order simulations.

This minisymposium will discuss faults, errors, and failures that occur in extreme-scale computing systems. We want to increase awareness that resilience is a critical topic and that there are efforts and results that offer solutions to scientists and users of extreme-scale computing systems. Hardware level faults fall into two categories: hard faults are a consequence of permanent component failures (requiring repairs), while soft, or transient faults result from single upset events (e.g. a bit-flip in a memory cell) and have impermanent effects on the system. Hard faults typically result in system-wide failures, and in the absence of a fault management mechanism, an executed application is interrupted and its data is lost. The standard approach to cope with failures is to checkpoint, rollback and recover applications. However, it is expected that this approach may no longer be a viable solution on the upcoming Exascale systems. In contrast to hard faults, transient faults cannot always be detected and they can lead to Silent Data Corruptions (SDCs). Significant research efforts have been pursued to develop efficient SDC detectors, but there is not a perfect solution yet. The speakers selected for this minisymposium will give an overview of current solutions and future challenges.

Recent advances in theory and numerical methods in parallel to the availability of high-quality massive data sets and high-performance computing provide unprecedented opportunities to improve our understanding of Earth’s interior and its mechanism. The goal of this session is to bring computational and Earth scientists together to form a platform to discuss the current status, challenges and future directions in computational geosciences highlighting numerical simulations, the state-of-the-art HPC applications and their scientific outcomes. Contributions include, but are not limited to, the areas of earthquake engineering, passive and active-source seismic imaging, geodynamical modelling, magneto-fluid dynamics, etc. in conjunction with computational approaches such as numerical solvers, large-scale workflow, big data, optimisation strategies, etc. on HPC systems.

Weather and climate models provide society with increasingly reliable weather forecasts and climate projections: critical information that can save both lives and money. With the advent of accelerators in high performance computing, several efforts around the world have begun porting weather and climate models to these emerging hardware architectures using different approaches and programming models. However, little attention has been given to the impact different porting approaches have on the maintainability of the codes. This minisymposium will provide an overview of the porting and maintainability experiences from four different community models in the US and Europe and will include a discussion of the pros/cons of different approaches in view of maintaining performance-portable production atmospheric community models.

Personalized health aims to provide the right treatment at the right time for each individual. A major premise is that empowerment with more knowledge leads to better decision-making. Tailored, predictive interventions have the potential to change from a reactive to a preventative approach, thereby significantly extending the duration of health. To achieve this, highly performant computational environments to store, transfer, analyse and integrate data produced at astonishing rates are prerequisites. Today personalized health becomes a practical reality, and the exploding 'big data' in healthcare provides exciting IT opportunities.

This minisymposium highlights computational approaches to enable the next steps in biomedical discovery. The introductory keynote will outline the current clinical and scientific needs for advanced computational approaches. The second session is about computational methods that already revolutionise medical practice through virtual reality systems for personalized medical surgeries. The third talk discusses the utilisation of HPC for finding new anti-cancer therapies, from discovery to clinical trials. Our final speaker will present the latest developments in workflow environment used by the Swiss Personalized Health Network (SPHN) to support Swiss researchers in exploring the next generation of data-driven health and care innovations.

Theory and experiment have long been two equal pillars of science, and many would hope to add simulation as a third pillar. However, the challenges for those writing the simulation software are immense. The development team must (a) encompass strong domain expertise, so that the simulations are fit for their purpose; (b) develop the code in a way that can be sustained even without the original authors; and (c) demonstrate to their user communities through testing, benchmarking and documentation that the software will be useful in the hands of researchers, who will not be able to read the code.

In this minisymposium, we will hear from the developers of large community codes about approaches they have adopted to unite the teams of people around the development and maintenance of shared codebases. These will cover not just the programming languages and development tools that have been shown to work well, but also how to encourage adoption of good software engineering techniques by professionals and students of other disciplines, and the career-development needs of the research software engineers who will execute the bulk of the work.

Graphs (or networks) are a very powerful abstraction of various phenomena that can be expressed as a relation between entities. For several decades researchers in theoretical computer science and discrete mathematics have been developing a wealth of graph theory and graph algorithms. Recently, however, we see a qualitative change in how graph algorithms are used in practice: The complex structure of graphs in new and emerging applications, the size of typical inputs, and the computer architectures on which graph problems are solved call for novel algorithms and hardware efficient approaches. As computer architectures and memory hierarchies are becoming more complex, increasingly parallel and heterogeneous it is important to develop parallel algorithms and tools with these specific hardware constraints in mind. The minisymposium thus aims to bring together experts in developing, implementing and using modern graph algorithms to address this issue. Existing algorithms and tools will be reviewed in terms of modern HPC architectures and novel hardware efficient approaches will be presented. The application areas for such techniques include sparse matrix partitioning/coloring, and network graph analysis that are traditionally sequential as well as applications that need the extreme performance of emerging hardware architectures.

This minisymposium will address exciting opportunities for plasma simulation in the pre-exascale era, addressing many issues which transcend plasma physics and are of interest to a wide audience. Plasma physics offers many opportunities to explore quintessential complex systems characterized by multi-scale and multi-physics problems. The underlying nonlinear integro-differential equations can only be solved with the help of supercomputers. Therefore, not surprisingly, many computational plasma scientists view the upcoming exascale era as the golden age, finally allowing them for the very first time to attack several long-standing fundamental questions, from turbulence to magnetic reconnection to dynamo action - as well as their self-consistent interactions. The goal of the present minisymposium is to present examples of how the computational plasma physics community is preparing for the exascale era. A particularly fascinating aspect of this theme can be put into the formula "big data meets computation." Handling massive amounts of data - before, during, and after a simulation - and combining experimental or observational data with simulation data are two key challenges in this context. Novel ideas along these lines will be presented.

Heterogeneity has been very evident among HPC systems and the trend only continues to rapidly evolve. Such a trend involves systems equipped with hierarchical processors along with accelerators and memory/storage components that is expected to facilitate migration of an increasingly diverse set of scientific applications thus meeting the demands of a wide user community. However in order to do so, effectively, we need rich programming models and languages that can tap into the massive potential of these large scale systems. This minisymposium addresses how the widely popular parallel programming paradigms such as CUDA, OpenMP 4.5, OpenACC and Alpaka can be adapted for a variety of applications such as atomistic simulation, turbulence combustion, simulation of smoke propagation, and plasma physics. The talks will explain using these real world applications how to balance performance and portability for a minimum of future work.

Tensor algebra operations are ubiquitous in domains including data analytics, machine learning, engineering, and science. Moreover, the use of tensor methods in these domains has grown tremendously in the last ten years. High-performance implementations and parallelization of tensor algebra operations underlying these methods require considerations beyond standard techniques used in linear algebra. A further key challenge is the development of effective abstractions and libraries for tensor algebra, as no standard interface or library (like LAPACK) has been established in the scientific community. The presentations in this minisymposium will cover challenges faced in the development of four distinct major tensor software library efforts. These libraries focus on sparse and dense tensor contractions, with three targeting distributed memory: Cyclops, NWChem, and DBCSR, and a fourth, TACO, targeting shared memory. These efforts have been application-driven, in particular by the tensor problems prevalent in electronic structure calculations. The minisymposium will serve as a discussion on the current state of the art of the tensor libraries in order to understand possible synergy between the projects and with the possibility to build a common interface for tensor algebra operations.

The healthcare sector is clearly experiencing a data revolution. With advances in digital health records, genomic sequencing, burgeoning growth of social networks and media for community-health, and the emerging "App" market for health-related mobile and web-enabled applications – there is tremendous access and availability of private and public data for advancing precision medicine and improving population health. This ability to leverage these datasets for translational value, across the continuum of basic, preclinical, and clinical science will be critical for addressing in an effective and timely manner emerging personalized and population healthcare challenges. At the same time, artificial intelligence is making continuing advances in biomedicine. However, there are outstanding questions of how AI can provide actionable clinical insights. We will bring together a community of biomedical researchers and computer scientists to present the latest advances as well as discuss successes, challenges, and next frontiers in health intelligence. The overarching goal of the symposium is to highlight health informatics applications and related methodological advances that involve heterogeneous biomedical data (e.g., imaging, genomic, text, and sensor data) while emphasizing the current and emerging challenges of ensuring their translational value and broad population impact.

Life Sciences have become crucially dependent on software for analysis of experimental data, systems modelling and simulation, data integration across various repositories and databases, etc. The dramatic increase of available tools has enabled scientists to perform ever more complex studies while taking advantage of modern high-end (HPC and HTC) compute facilities. Experimental facilities are producing staggering amounts of data which led to the rapid development of novel data analytics and machine learning techniques. We are at a stage where there is a need for additional focus on improving the interoperability of software applications, enabling better coupling of tools with data sources and devising efficient workflows and libraries for the upcoming Exascale era in HPC. Such advances will considerably improve the productivity of researchers and allow them to address novel scientific problems. This minisymposium brings together invited experts from two leading institutions in the field – BioExcel, the European Center of Excellence for Computational Biomolecular Research (www.bioexcel.eu), and MolSSI, Molecular Sciences Software Institute (www.molssi.org) in the US to discuss advances in this important field. After the talks we will hold a round-table discussion on "Simulations at Exascale - myth or reality?".

Recent advances in theory and numerical methods in parallel to the availability of high-quality massive data sets and high-performance computing provide unprecedented opportunities to improve our understanding of Earth’s interior and its mechanism. The goal of this session is to bring computational and Earth scientists together to form a platform to discuss the current status, challenges and future directions in computational geosciences highlighting numerical simulations, the state-of-the-art HPC applications and their scientific outcomes. Contributions include, but are not limited to, the areas of earthquake engineering, passive and active-source seismic imaging, geodynamical modelling, magneto-fluid dynamics, etc. in conjunction with computational approaches such as numerical solvers, large-scale workflow, big data, optimisation strategies, etc. on HPC systems.

The aim of this minisymposium is to discuss research at the intersection of high-dimensional approximation and parallel computing. High-dimensional approximation drives e.g. uncertainty quantification, optimization, machine learning and big data as well as simulations of complex physics models. It is well-known that approximation of functions of growing dimension has the challenge of the curse of dimensionality. Over the last decades, many powerful mathematical tools have been developed to do weaken or overcome this. These include, but are not limited to (quasi) Monte Carlo, multi-level / multi-fidelity techniques, sparse grids, low-rank and tensor product approximations, hierarchical matrices, compressed sensing and meshfree methods.

There is a growing interest to solve high-dimensional approximation problems at large scale. While many of the discussed methods have good or even optimal approximation properties and complexities for larger dimensions, some have been primarily developed for sequential execution. However, to solve large scale approximation problems, it becomes necessary to develop fast, scalable and parallel numerical methods. This minisymposium invites contributions in high-dimensional approximation ranging from initial studies for the use of parallel techniques up to full scale parallel methods that run on large HPC clusters. Speakers will showcase both algorithmic-oriented and application-centered research.

Increasing ensemble size, grid resolution and complexity of Earth system models have driven data volumes to a point where scientists are forced to give some of it up. Estimates for total data volumes of the Climate Model Intercomparison Project 6 (CMIP6, Eyring et al. 2016) range from 15 to 30 PBytes. Traditional disk or tape storage bandwidth and capacity is not keeping pace with increases in computational capacity, and data storage and movement constraints are already limiting scientific analysis of these enormous simulation efforts.

With significant momentum towards global convection-resolving climate simulations, traditional compute-store-analyze workflows still common in climate science today will break down and research into alternative data analysis workflows is urgently needed. This minisymposium will highlight and critically discuss the potential and challenges of several pathways to alleviate the data bottleneck in climate science: data compression, in-situ/in-transit data analysis and processing, recomputation.

Recent years have seen a rise in both scrutiny and attention being paid to scientific software because of the corresponding rise in scientific discovery through simulations and data analysis. To produce high quality science it is critical to employ high quality tools and processes. The vast majority of projects engaged in using computations for advancing scientific understanding have yet to attain sufficient robustness in either their software or their process. What has changed is the realization by these projects that they must strive for robustness if they want credibility. However, given the fundamentally multidisciplinary nature of computational science, and scarcity of resources, training and experience, research teams struggle. One of the recently launched community efforts for addressing this gap is the Better Scientific Software website (BSSw.io), a resource for scientific software developers and users which includes curated and contributed content, discussion forums and many other helpful features. Through this minisymposium we seek to introduce the scientific software community to BSSw, highlighting some of the content offered. We also seek to broaden our reach with presentations from practitioners in several scientific communities highlighting what they do and their plans for the sustainability of their software.

Materials discovery requires one to explore many possible chemical compositions as well as many possible structures for a given composition. Even though highly powerful computers allow us to perform density functional calculations quite rapidly for a limited number of configurations of moderate size, the huge number of force evaluations required for structural and stochiometric explorations of a very large number of materials is too expensive on the density functional level.

Machine learning schemes may come to our assistance in this context. They have been shown to deliver density functional accuracy at a greatly reduced numerical cost for limited test sets. Since all machine learning schemes are intrinsically interpolation schemes it is however not clear how well these schemes can extrapolate to discover for instance entirely new materials that were not contained in the fitting database.

The minisymposium will therefore focus on problems and systems that are difficult to treat with present day machine learning schemes and present methods that might overcome some of the current limitations.

Turbulence in plasmas confined by a background magnetic field stand out as one of today’s great challenges in physics. To describe such systems, the most appropriate approach is based on gyrokinetic theory, which takes advantage of the small characteristic frequency of turbulence as compared to the cyclotron frequency. It results in a reduction of phase space from 6D to 5D, and eliminates the fastest time scales from the description. Despite this reduction in complexity, solving the set of gyrokinetic equations, consistently with the electromagnetic field equations, remains a formidable task. It is an intrinsically nonlinear, multi-scale problem and thus requires major HPC resources, for which advanced numerical techniques are a necessity in order to obtain a reasonable time-to-solution. In this minisymposium, progress on state-of-the-art codes based on all three major computational approaches (Eulerian, semi-Lagrangian and Lagrangian-PIC) will be presented, as well as possible innovative alternatives.

In the last two years, the field of gravitational wave astronomy has seen a breakthrough: In February 2016, the ground-based Laser Interferometer Gravitational-Wave Observatory (LIGO) announced the first direct detection of gravitational waves and the first observation of a binary black hole merger. Very recently, the LIGO-Virgo collaborations and their partner observatories announced the first joint observation of a neutron star merger in gravitational waves and the electromagnetic spectrum, marking a new era in multi-messenger astronomy. This minisymposium is intended to give an overview over certain important aspects of gravitational-wave data analysis with the current generation of ground-based detectors, highlighting applications of high-performance computing, machine learning and citizen science. The four presentations of the minisymposium will address the following topics: searches for gravitational waves in the detector data, computational aspects of source parameter estimation and physics implications of the LIGO-Virgo data, numerical relativity and its applications for the modelling of gravitational waves, aspects of data quality and data cleaning for the LIGO-Virgo data.

This minisymposium will present the latest advances in using HPC systems worldwide for physics results from large experiments in High Energy Physics and Particle Astrophysics. While HPC usage worldwide will be described, the experiences from the use of leadership class computing facilities in the US by the Large Hadron Collider experiments will be highlighted. LHC experiments have traditionally used the infrastructure provided by the Worldwide LHC Computing Grid. This has been supplemented in recent years by incorporating traditional HPC systems into the production and analysis computing systems at the LHC. Primarily CPU intensive simulation workflows are executed at HPCs - though other workflows are also being tested. The experiences gained by the LHC experiments over the past few years have opened HPC usage for other experimental and data intensive sciences. Four talks at this minisymposium will summarize the state of the art and the future wishlist for HPC usage for current and future experiments.

The HPUQ minisymposium focuses on uncertainty quantification (UQ) of mechanistic models for natural sciences (eg. Engineering, Life and Aquatic Sciences) using high performance computing (HPC). The statistical inference (e.g. calibration) of models for complex mechanistic models, in the abundance of data arriving from heterogeneous sources, poses a methodological and computational challenge for scientists. In the first session, the minisymposium highlights cutting edge frameworks for rigorous and robust UQ as ABCpy, Π4U, PyMLMC, SPUX to address these issues, with a focus towards optimal algorithmic performance and efficient utilization of HPC resources. In the second session of the minisymposium, we shift the focus to the applications of UQ methodologies in several important scientific domains spanning from Biomedicine and Biomechanics to Aerospace Engineering and Fluid Dynamics.

Modeling energy and mass transport phenomena in the solid state accurately and efficiently in computer simulations is of critical importance to the discovery and design of new materials in important applications, such as lithium-ion solid-state electrolytes, proton-conducting fuel cells and better thermoelectrics. Progress in recent years enabled the modeling community to elucidate transport mechanisms in several materials of importance, but to achieve further progress major obstacles need to be tackled.

Firstly, the computational cost of the current methods limits the time and lengths scales of the systems that can be modeled. We will present several efforts that are targeted either at reaching larger system sizes and time scales in simulations, or to improve the estimates of transport coefficients from finite simulation lengths using more advanced statistical tools.

Secondly, with increases in computer power and novel methods for modeling and analysis it has become possible to push for large-scale screening of structural databases for desired transport properties. Such high-throughput approaches require new concepts of automatization, data reproducibility and results dissemination in the field of materials' simulation, and we will present the developed concepts, applications and results.

Nearly all fields in geophysics combine numerical models and data measurements to predict the future state of a dynamical system and/or infer unknown parameters. Such models may produce highly nonlinear systems with extremely large numbers of unknowns.

The ever-increasing power and widespread availability of massively parallel supercomputers offers researchers the opportunity to continually increase both the spatio-temporal resolution and the physical complexity within their numerical models. However, this requires access to solvers that can harness the resources of high-performance computing clusters efficiently and which scale for problem sizes with billions of degrees of freedom.

In this minisymposium, we discuss numerical and algorithmic approaches, scientific libraries and coding practices to develop and maintain scalable solvers for various applications in geophysics. Examples include, but are not limited to, seismic wave propagation and imaging, geodynamics, and hydro-mechanical processes.

Reliable weather predictions and climate projections are of vital importance for society and for the creation and preservation of prosperity. An increase in horizontal resolution in global simulations of the atmosphere to ~1 km would enable us to represent large cloud-systems and deep convection explicitly within simulations. This would reduce much of the uncertainty in predictions of weather and climate. Unfortunately, it is not possible yet to run operational simulations with state-of-the-art global weather and climate models at this level of resolution using today’s supercomputing facilities. This minisymposium will provide an update on the progress of global weather and climate models when running at 1 km horizontal resolution. The most significant challenges towards cloud-resolving simulations will be assessed (including scalability, simulations far away from peak-performance, I/O volume and efficient time-stepping schemes) and possible ways to overcome these barriers will be discussed.

The cost of generating biological data is dropping exponentially, a decrease that has far outstripped predictions based on Moore’s Law. This has ushered in a new era of systems biology in which there are unprecedented opportunities to gain insights into complex biological systems. Integrated biological models need to capture the higher order complexity of the interactions among cellular components. Solving such complex combinatorial problems will give us unprecedented levels of understanding of biological systems. However, this leads to a combinatorial explosion in the search space of biological data. These exponentially increasing volumes of data, combined with the desire to model more and more sophisticated sets of relationships within a cell and across an organism (or in some cases even ecosystems), have led to an unmet need for computational resources and sophisticated algorithms that can make use of them. Thus the bottleneck in biological science is no longer data generation but is in fact computational analysis. A full model of all of the higher order interactions of cellular and organismal components is one of the ultimate grand challenges of systems biology. The use of machine and deep learning algorithms provide some of the methodologies with which to achieve this goal.

The algorithms underlying numerical weather prediction (NWP) and climate models that have been developed in the past few decades face an increasing challenge to adapt to paradigm shifts imposed by new hardware developments. The emerging diverse and complex hardware solutions have a large impact on the programming models traditionally used in NWP and climate modelling software, triggering a rethink of design choices for future software frameworks and how Earth system model (ESM) components interact. Furthermore there is a drive to increase the model complexity to include ever more processes of the whole Earth system. These complexities will inevitably break NWP modelling infrastructures that did not take these aspects into consideration 30+ years ago. As upcoming hardware solutions evermore push the boundaries of parallel execution, each ESM component may be reaching a limit in parallel scaling efficiency.

Already, coupling infrastructure exists that enables concurrent execution of model components. In order to make use of increasing parallelism, it may be required to take a step backwards, and redesign the ESM components to become more modular and enable more concurrent execution. This minisymposium will provide an update on progress both in infrastructure developments and developments in concurrent model component execution.

In the push towards exascale computing the current paradigm of bulk-synchronous distributed computing is giving way to a more asynchronous paradigm, where it is becoming increasingly important to have the inherent asynchrony in an application be exploited for performance and scalability. Whether the asynchrony is enabled algorithmically, expressed explicitly in the user program, discovered by a programming model and/or runtime there are critical research questions that need to be addressed. In this paradigm, asynchronous many task (AMT) programming models have made great progress in demonstrating the concept and paving the way forward. While asynchronous task based programming models for shared-memory systems have been around for a long time serious challenges remain in extending them to a distributed memory setting. This minisymposium is targeted at providing a platform to convey the progress and assess the challenges in distributed asynchronous computing. The talks included in the minisymposium will cover the range of relevant topics; resilience and fault tolerance for distributed AMT programming models, asynchrony-tolerant numerical schemes and algorithms for stencil-based applications, task-based parallel simulations of multi-phase flows, and task-based runtimes utilizing directed acyclic graphs and a domain specific language for multi-physics simulations of turbulent reacting flows.

The miniaturisation of devices towards the molecular scale is a dynamic field of research that has evolved dramatically in recent years. Its applications span a broad field, from the fabrication of smaller, more powerful computer chips, to charge storage, robotics, solar cells or biosensors. Complex functional materials are being envisioned and developed with these applications in mind, in an unprecedented effort towards rational design, through a combination of theory, computation and experiment.

In this context, the challenge for theory is to formulate laws and uncover patterns when the interactions that drive the relevant processes are as varied as the systems being investigated. Potential energy surfaces in nanoscale materials are complex, with multiple competing minima and steep barriers. The properties of the complex materials are also a challenge for computation: quantum effects in nanoscale interactions are reflected in the behaviour of the material as a whole, a statistical entity. This minisymposium will bring together researchers who have made significant contributions to the two essential challenges in materials simulations: sampling complex potential energy surfaces in structural prediction and bridging the relevant length scales in materials fabrication and properties. Current challenges and perspectives in method development and new applications will be discussed.

The volume of data generated by astrophysics observatories, and the complexity of knowledge management, has increased constantly and is now on the verge to reach even higher levels with observatories planned to generate 10-100s PB per year. This increase in volume, mostly driven by ground based telescopes with high resolution detectors and/or extreme time sampling, requires new computing infrastructures and new ways for scientists to interact with data and computing.

Observatory interfaces will need to move from data to analysis, interpretation and knowledge and from analysis to synthesis. Data mining, driven by the needs of science and education will require an integration of archive, pipeline and interpretation, which are still largely conceived as disconnected services. This evolution will benefit from the increased computing power and artificial intelligence which will help interpreting data flow exceeding human insight and will likely transform the astronomy business model.

The goal of this minisymposium is to clarify this evolution by comparing the needs of current and planned observatories and study how this implementation could take place both at the technical level and in the way scientists collaborate. The specific opportunities for Switzerland will also be discussed.

The ATLAS experiment at CERN’s Large Hadron Collider depends on the Worldwide LHC Computing Grid, the WLCG, for its remote computing infrastructure. PanDA, the workload management system used by ATLAS, annually processes over an exabyte of data using an average of 250,000 distributed batch slots, to enable hundreds of new scientific results. An effort was launched to extend PanDA, called BigPanDA, to access HPC resources, funded by the US Department of Energy (DOE-ASCR). Through this successful effort, ATLAS today uses about 20 million hours monthly on the Titan supercomputer at Oak Ridge National Laboratory. Many other supercomputers have also been integrated into ATLAS computing. This minisymposium will explore the software and operational lessons learned in integrating HPCs with traditional grid computing, and describe recent efforts to use BigPanDA for many other scientific domains. Three talks will summarize the state of the art and the future wishlist for HPC usage for current and future experiments, while a concluding expert panel discussion will focus on the future.

The HPUQ minisymposium focuses on uncertainty quantification (UQ) of mechanistic models for natural sciences (eg. Engineering, Life and Aquatic Sciences) using high performance computing (HPC). The statistical inference (e.g. calibration) of models for complex mechanistic models, in the abundance of data arriving from heterogeneous sources, poses a methodological and computational challenge for scientists. In the first session, the minisymposium highlights cutting edge frameworks for rigorous and robust UQ as ABCpy, Π4U, PyMLMC, SPUX to address these issues, with a focus towards optimal algorithmic performance and efficient utilization of HPC resources. In the second session of the minisymposium, we shift the focus to the applications of UQ methodologies in several important scientific domains spanning from Biomedicine and Biomechanics to Aerospace Engineering and Fluid Dynamics.

This minisymposium will explore some unconventional methods for numerical simulations of partial differential equations. Often, discretizations of PDEs produce regular error patterns, some of which can be both quantified and smoothed by adding small stochastic terms. In fact, stochastic processes can be intimately connected with PDEs. For example, the heat equation is just a PDE form for the distribution of Brownian random motion. More general frameworks can be built around the Feynman-Kac formula (e.g. Mark Freidlin's book on functional integrals, Princeton Univ. Press, 1985). Such formulations are both flexible and intrinsically parallelizable (e.g. W. Petersen and P. Arbenz, Oxford Univ. Press, 2004). In high spacial dimensions, such techniques, when formulated as Monte-Carlo methods, are much more accurate than might be expected and make many difficult high dimensional simulations even possible. In, say, 3-D, certain interactions (e.g. foams, bubbles) can be modelled phenomenologically where details about these interactions are not well understood. In addition, we will discuss balancing methods or systems in external fields, and the connections between particle methods and the PDEs which describe the distributions.