Abstract: Convection and magnetic field of the Earth's outer core are expected to have vast length scales. To resolve these flows at extremely high spatial resolution, high performance computing is required for geodynamo simulations. Calypso has been shown to scale well on computer clusters capable of computing at the order of 10⁵ cores using Message Passing Interface (MPI) and Open Multi-Processing (OpenMP) parallelization for CPUs. Calypso is designed to model magnetohydrodynamics (MHD) of a Boussinesq fluid in a rotating spherical shell, such as the outer core of Earth. Calypso employs a pseudo-spectral method that requires both a forward and backward spherical harmonic transform (SHT) in each iteration of the time integration loop. The time complexity of SHT is dominated by the time complexity of Legendre transform which is O(N^3). We investigate time efficiency of three different algorithms of the SHT using graphics processing units (GPUs). One is to preemptively compute the Legendre polynomials on the CPU before executing Legendre transform on the GPU. In the second approach, the Legendre polynomials are computed “on-the-fly” on a need-by-basis. In the third approach, we employ sum reduction intrinsics offered by the library cuda Unbound. We examine the trade-offs between space and time on Maverick, a Texas Advanced Computing Center (TACC) supercomputer using a GPU enabled Legendre transform. The first algorithm leads to a memory bottleneck that can be alleviated by utilizing the memory hierarchy of Nvidia GPUs. Moreover, coalesced reads and prefetching can further reduce the memory bottleneck. The second algorithm is highly computationally intensive relative to memory bandwidth utilization as a result of repetitive computations. The third algorithm is well-balanced between computational and memory utilization. We conclude that the most-efficient approach is a hybrid model that involves a look-up table as well as on-the-fly computations over the GPU.

Abstract: Adjoint sensitivity computation of parameter estimation problems is a widely used technique in the field of computational science and engineering for retrieving derivatives of a cost func- tional with respect to parameters efficiently. Those derivatives can be used, e.g., for sensitivity analysis, optimization, or robustness analysis. Deriving and implementing adjoint code is an error-prone, non-trivial task which can be avoided by using Algorithmic Differentiation (AD) software. Generating adjoint code by AD software has the downside of usually requiring a huge amount of memory as well as a non-optimal run time. In this article, we couple two approaches for achieving both, a robust and efficient adjoint code: symbolically derived adjoint formula- tions are coupled with AD. Comparisons are carried out for a real-world case study originating from the remote atmospheric sensing simulation software JURASSIC developed at the Institute of Energy and Climate Research – Stratosphere, Research Center Jülich. We show, that the coupled approach outperforms the fully algorithmic approach by AD in terms of run time and memory requirement and argue that this can be achieved while still preserving the desireable feature of AD being automatic.

Abstract: In this paper, we discuss strategies for computing subsets of eigenvectors of matrices corresponding to subdomains of finite element meshes achieving compromise between two contradicting goals. The subset of eigenvectors is required in the construction of coarse spaces used in algebraic multigrid methods (AMG) as well as in certain domain decomposition (DD) methods. The quality of the coarse spaces depends on the number of eigenvectors, which improves the approximation properties of the coarse space and impacts the overall performance and convergence of the associated AMG or DD algorithms. However, a large number of eigenvectors reflects negatively the sparsity of the corresponding coarse matrices, which can become fairly dense. The sparsity of the coarse matrices can be controlled to a certain extent by the size of the subdomains (union of finite elements) referred to as agglomerates. If the size of the agglomerates is too large, then the cost of the eigensolvers increases and eventually can become unacceptable for the purpose of constructing the AMG or DD solvers. This paper investigates strategies to optimize the solution of the partial eigenproblems of interest. In particular, we examine direct and iterative eigensolvers for computing those subsets. Our experiments with synthetic meshes and with a well-known model of an oil-reservoir simulation benchmark indicate that iterative eigensolvers can lead to significant improvements in the overall performance of an AMG solver that exploits such spectral construction of coarse spaces.

Abstract: This paper discusses the solution of large-scale linear matrix equations using the Induced Dimension reduction method (IDR(s)). IDR(s) was originally presented to solve system of linear equations, and is based on the IDR(s) theorem. We generalize the IDR(s) theorem to solve linear problems in any finite-dimensional space. This generalization allows us to develop IDR(s) algorithms to approximate the solution of linear matrix equations. The IDR(s) method presented here has two main advantages; firstly, it does not require the computation of inverses of any matrix, and secondly, it allows incorporation of preconditioners. Additionally, we present a simple preconditioner to solve the Sylvester equation based on a fixed point iteration. Several numerical examples illustrate the performance of IDR($s$) for solving linear matrix equations. We also present the software implementation.

Abstract: Numerical solutions of partial differential equations (PDEs) on manifolds continues to generate a lot of interest among scientists in the natural and applied sciences. On the other hand, recent developments of 3D scanning and computer vision technologies have produced a large number of 3D surface models represented as point clouds. Herein, we develop a simple and efficient method for solving PDEs on closed surfaces represented as point clouds. By projecting the radial vector of standard radial basis function(RBF) kernels onto the local tangent plane, we are able to produce a representation of functions that permits the replacement of surface differential operators with their Cartesian equivalent. We demonstrate, numerically, the efficiency of the method in discretizing the Laplace Beltrami operator.

Abstract: The Monte Carlo method is a common and accurate way to model neutron transport with minimal approximations. However, such method is rather time-consuming due to its slow convergence rate. More specifically, the energy lookup process for cross sections can take up to 80% of overall computing time and therefore becomes an important performance hotspot. Several optimization solutions have been already proposed: unionized grid, hashing and fractional cascading methods. In this paper we revisit those algorithms for both CPU and manycore (Intel MIC) architectures and introduce vectorized versions. Tests are performed with the PATMOS Monte Carlo prototype, and algorithms are evaluated and compared in terms of time performance and memory usage. Results show that significant speedup can be achieved over the conventional binary search on both CPU and Intel MIC. Further optimization with vectorization instructions has been proved very efficient on Intel MIC architecture due to its 512-bit Vector Processing Unit (VPU); on CPU this improvement is limited by the smaller VPU width.

Abstract: This article presents a framework for performing ensemble and hybrid data assimilation in a weak-constraint four-dimensional variational data assimilation system (w4D-Var). A practical approach is considered that relies on an ensemble of w4D-Var systems solved by the incremental algorithm to obtain flow-dependent estimates to the model error statistics. A proof-of-concept is presented in an idealized context using the Lorenz multi-scale model. A comparative analysis is performed between the weak- and strong-constraint ensemble-based methods. The importance of the weight coefficients assigned to the static and ensemble-based components of the error covariances is also investigated. Our preliminary numerical experiments indicate that an ensemble-based model error covariance specification may significantly improve the quality of the analysis.

Abstract: Population dynamics illustrates the changes of the size and age composition of populations. Modeling and simulation techniques have been used to model those population dynamics, and the developed models are utilized to design and analyze public polices. One classic method to model population dynamics is microsimulation. The microsimulation describes population dynamics in an individual-level, and an individual acts depending on stochastic process. An emerging method is agent-based model which rather focuses on interactions among individuals and expects to see unexpected situations generated by the interactions. Their different attentions on individuals can make them to complement the weak point of the opponent in the development of population dynamics model. From this perspective, This paper proposes a hybrid model structure combining microsimulation and agent-based model for modeling population dynamics. In the proposed model, the microsimulation model takes a role to depict how an individual chooses its behavior based on stochastic process parameterized by real data; the agent-based model incorporates interactions among individuals considering their own states and rules. The case study introduces Korean population dynamics model developed by the proposed approach, and its simulation results show the population changes triggered by a variance of behavior and interaction factors.

Abstract: The paper presents the idea of a complex model of clinical episode applied, based on data-driven approach for decision support in treatment of ACS (Acute Coronary Syndrome). The idea is aimed towards improvement of predictive capability of a data-driven model by combination of different models within a composite data-driven model. It can implement either hierarchical or alternative combination of models. Three examples of data-driven models are described: simple classifier, outcome prediction based on reanimation time and states-based prediction model to be used as a part of complex model of episodes. To implement the proposed approach a generalized architecture of data-driven clinical decision support systems was developed. The solution is developed as a part of complex clinical decision support system for cardiac diseases for Federal Almazov North-West Medical Research Centre in Saint Petersburg, Russia.

Abstract: Crowd behavior and its movement has been an actively studied domain during last three decades. There are microscopic models used for realistic simulation of crowds in different conditions. Such models reproduce pedestrian movement quite well, however, their efficiency can vary depending on the conditions of simulation. For instance, some models show realistic results in high density of pedestrians and vice versa in low density. This work describes an early study aimed at developing an approach to combine several microscopic models using an ensemble approach to overcome individual weaknesses of the models. Possible ways to build hybrid models, as well as the main classes of ensembles are described. A prior calibration procedure was implemented using the evolutionary approach to create an ensemble of the most suitable models using dynamical macro-parameters such as density and speed as the optimization objectives. Several trial experiments and comparisons with single models were carried out for selected types of hybridization.

Abstract: For performance analysis tools to be useful, they need to show the relation of detected bottlenecks to source code. To this end, it often makes sense to use the instruction triggering a problematic event. However for cache line utilization, information on usage is only available at eviction time, but may be better attributed to the instruction which loaded the line. Such attribution is impossible with current processor hardware. Callgrind, a cache simulator part of the open-source Valgrind tool, can do this. However, it only provides Self Costs. In this paper, we extend the cost attribution of cache use metrics to inclusive costs which helps for top-down analysis of complex workloads. The technique can be used for all event types where collected metrics should to be attributed to instructions executing earlier in a program run to be useful.

Abstract: Computational kernels, which are small pieces of software that selectively capture the characteristics of larger applications, have been used successfully for decades. Kernels allow for the testing of a compiler's ability to optimize code, performance of future hardware and reproducing compiler bugs. Unfortunately they can be rather time consuming to create and do not always accurately represent the full complexity of large scientific applications. Furthermore, expert knowledge is often required to create such kernels. In this paper, we present a Python-based tool that greatly simplifies the generation of computational kernels from Fortran based applications. Our tool automatically extracts partial source code of a larger Fortran application into a stand-alone executable kernel. Additionally, our tool also generates state data necessary for proper execution and verification of the extracted kernel. We have utilized our tool to extract more than thirty computational kernels from a million-line climate simulation model. Our extracted kernels have been used for a variety of purposes including: code modernization, identification of limitations in compiler optimizations, numerical algorithm debugging, compiler bug reporting, and for procurement benchmarking.

Abstract: The HPCmatlab framework has been developed for Distributed Memory Programming in Matlab/Octave using the Message Passing Interface (MPI). The communication routines in the MPI library are implemented using MEX wrappers. Point-to-point, collective as well as one-sided communication is supported. Benchmarking results show better performance than the Mathworks Distributed Computing Server. HPCmatlab has been used to successfully parallelize and speed up Matlab applications developed for scientific computing. The application results show good scalability, while preserving the ease of programmability. HPCmatlab also enables shared memory programming using Pthreads and Parallel I/O using the ADIOS package.

Abstract: Runtime verification of large-scale scientific codes is difficult because they often involve thousands of processes, and generate very large data structures. Further, the programs often embody complex algorithms making them difficult for non-experts to follow. Notably, typical scientific codes implement mathematical models that often possess predictable statistical features. Therefore, incorporating statistical analysis techniques in the verification process allows using program’s state to reveal unusual details of the computation at runtime. In our earlier work, we proposed a statistical framework for debugging large-scale applications. In this paper, we argue that such framework can be useful in the runtime verification process of scientific codes. We demonstrate how two production simulation programs are verified using statistics. The system is evaluated on a 20,000-core Cray XE6.

Abstract: In C++, new features and semantics can be added to an existing software package without sweeping code changes by introducing a user-defined type using operator overloading. This approach is used, for example, to add capabilities such as algorithmic differentiation. However, the introduction of operator overloading can cause a multitude of compilation errors. In a previous paper, we identified code constructs that cause a violation of the C++ language standard after a type change, and a tool called OO-Lint based on the Clang compiler that identifies these code constructs with lint-like messages. In this paper, we present an extension of this work that automatically transforms such problematic code constructs in order to make an existing code base compatible with a semantic augmentation through operator overloading. We applied our tool to the CFD software OpenFOAM and detected and transformed 23 instances of problematic code constructs in 160,000 lines of code. A significant amount of these root causes are included up to 425 times in other files causing a tremendous compiler error amplification. In addition, we show the significance of our work with a case study of the evolution of the ice flow modeling software ISSM, comparing a recent version which was manually type changed with a legacy version. The recent version shows no signs of problematic code constructs. In contrast, our tool detected and transformed a remarkable amount of issues in the legacy version that previously had to be manually located and fixed.

Abstract: In this paper, we discuss a simple and efficient technique for multi-objective design optimization of multi-parameter microwave and antenna structures. Our method exploits a stencil-based approach for identification of the Pareto front that does not rely on population-based metaheuristic algorithms, typically used for this purpose. The optimization procedure is realized in two steps. Initially, the initial Pareto-optimal set representing the best possible trade-offs between conflicting objectives is obtained using low-fidelity representation (coarsely-discretized EM model simulations) of the structure at hand. This is realized by sequential construction and relocation of small design space segments (patches) in order to create a path connecting the extreme Pareto front designs identified beforehand. In the second step, the Pareto set is refined to yield the optimal designs at the level of the high-fidelity electromagnetic (EM) model. The appropriate number of patches is determined automatically. The approach is validated by means of two multi-parameter design examples: a compact impedance transformer, and an ultra-wideband monopole antenna. Superiority of the patching method over the state-of-the-art multi-objective optimization techniques is demonstrated in terms of the computational cost of the design process.

Abstract: Supersonic vehicles are an important type of potential transports. Analysis of these vehicles requires the use of accurate models, which are also computationally expensive, to capture the highly nonlinear physics. This paper presents results of numerical investigations of using physics-based surrogate models to design supersonic airfoil shapes. Variable-fidelity models are generated using inviscid computational fluid dynamics simulations and analytical models. By using response correction techniques, in particular, the manifold mapping technique, fast surrogate models are constructed. The effectiveness of the approach is investigated using lift-constrained drag minimization problems of supersonic airfoil shapes. Compared with direct optimization, the results show that an order of magnitude speed up can be obtained. Furthermore, we investigate the effectiveness of the variable-fidelity technique in terms of speed and design quality using several combinations of medium-fidelity and low-fidelity models.

Abstract: Ultrasonic testing (UT) is used to detect internal flaws in materials or to characterize material properties. Computational simulations are an important part of the UT process. Fast models are essential for UT applications such as inverse design or model-assisted probability of detection. This paper presents applications of surrogate modeling techniques to create fast approximate models of UT simulator responses. In particular, we use data-driven surrogate modeling techniques (kriging interpolation), and physics-based surrogate modeling techniques (space mapping), as well a mixture of the two approaches. These techniques are demonstrated on metal components immersed in a water bath during the inspection process.

Abstract: A framework, PhaseLift, was recently proposed to solve the phase retrieval problem. In this framework, the problem is solved by optimizing a cost function over the set of complex Hermitian positive semidefinite matrices. This paper considers an approach based on an alternative cost function defined on a union of appropriate manifolds. It is related to the original cost function in a manner that preserves the ability to find a global minimizer and is significantly more efficient computationally. A rank-based optimality condition for stationary points is given and optimization algorithms based on state-of-the-art Riemannian optimization and dynamically reducing rank are proposed. Empirical evaluations are performed using the PhaseLift problem. The new approach is shown to be an effective method of phase retrieval with computational efficiency increased substantially compared to the algorithm used in original PhaseLift paper.

Abstract: The usage of heterogeneous multicore platforms is appealing for applications, e.g. hard real-time systems, due to the potential reduced energy consumption offered by such platforms. However, the power wall is still a barrier to improving the processor design process due to the power consumption of components. Hard real-time systems are part of life critical environments and reducing the energy consumption on such systems is an onerous and complex process. This paper reassesses the problem of finding assignments of hard real-time tasks among heterogeneous processors respecting time constraints and targeting low power consumption. We also propose models based on a well-established literature formulation of the Multilevel Generalized Assignment Problem (MGAP). We tackle the problem from the perspective of different models and their interplay on the search for optimal solutions. Experimentation show that using strict schedulability tests as restrictions of 0/1 integer linear programming results in faster solvers capable of finding optimum solutions with lower power consumption.

Abstract: TBD

Abstract: This paper extends FARSITE (a software used for wildfire modeling and simulation) to incorporate data assimilation techniques based on noisy and limited spatial resolution observations of the fire perimeter to improve the accuracy of wildfire spread predictions. To include data assimilation in FARSITE, uncertainty on both the simulated wildfire perimeter and the measured wildfire perimeter is used to formulate optimal updates for the prediction of the spread of the wildfire. For data assimilation, Wildfire perimeter measurements with limited spatial resolution and a known uncertainty are used to to formulate an optimal adjustment in the fire perimeter prediction. The adjustment is calculated from the Kalman filter gain in an Ensemble Kalman filter that exploit the uncertainty information on both the simulated wildfire perimeter and the measured wildfire perimeter. The approach is illustated on a wildfire simulation representing the 2014 Cocos fire and presents comparison results for hourly data assimilation results.

Abstract: The accurate prediction of forest fire propagation is a crucial issue to minimize its effects. So, several models have been developed to determine the forest fire propagation beforehand. Such models require several input parameters that in some cases cannot be known precisely in a real emergency. So, a two stage methodology was developed to calibrate the input parameters to improve the quality of the prediction. This methodology was based on Genetic Algorithms which require the execution of many simulations. Moreover, when the fire front is large some input parameters are not uniform along the whole front and complementary models must be introduced to determine the values of those parameters along the whole area involved. One of these non-uniform parameters is wind. So, in this work a wind field model is introduced. This model implies more computation time and response time is the main constraint. The prediction must be provided as fast as possible to be useful, so it is necessary to exploit all available computing resources. So a Hybrid MPI-OpenMP application has been developed to reach a response in the shortest possible time. This work is focused on reduce the execution time of a worker in a MPI Master/Worker structure analyzing the simulation software parts which compose the Fire Simulator System for large scale fores fires that runs on each worker.

Abstract: Online state and parameter estimation of atmospheric dispersion processes using multiple mobile sensor platforms is a prominent example of Dynamic Data-Driven Application Systems (DDDAS). Based on repeated predictions of a partial differential equation (PDE) model and measurements of the sensor network, estimates are updated and sensor trajectories are adapted to obtain more informative measurements. While most of the monitoring strategies require a central supercomputer, a novel decentralized plume monitoring approach is proposed in this paper. It combines the benefits of distributed approaches like scalability and robustness with the prediction ability of PDE process models. The strategy comprises model order reduction to keep calculations computationally tractable and a joint Kalman Filter with Covariance Intersection for incorporating measurements and propagating state information in the sensor network. Moreover, a cooperative vehicle controller is employed to guide the sensor vehicles to dynamically updated target locations that are based on the current estimated error variance.

Abstract: This paper shows the use of optimal filter estimation for real-time data processing to automatically detect dynamic transient effects in phasor data produced a synchrophasor vector processing systems. The optimal filters are estimated on the basis of phasor data where no disturbances are present and the estimation problem is formulated as a least squares optimization. Event detection bounds are computed from variance estimates and events are detected by formulating conditions on the number of consecutive samples that filtered phasor signals are outside of the bounds. Event detection is illustrated on the phasor data obtained from a microPMU system developed by Power Standards Lab.

Abstract: As more and more public and private Cloud resources are becoming available, it is common for a user to have access to multiple Cloud resources at the same time. Cloud federation dynamically aggregates multiple Cloud resources into a federated one. This paper explores how to build and run scientific workflows on top of a federated Cloud by integrating Kepler scientific workflow platform with CometCloud platform. Our integration can leverage capabilities of the two plat- forms: 1) dynamic resource federation, provisioning and allocation from CometCloud; 2) Easy workflow composition from Kepler; 3) Dynamic workflow scheduling and execution from the integration. We apply our integration to a bioinformatics workflow with three Cloud resources to evaluate its capabilities. We also discuss possible future directions from the integration.

Abstract: Scientific community across many disciplines is exploring new ways to extract knowledge from all available sources. Historically, written manuscripts have been the media of choice for recording experimental findings. Many disciplines such as social science, medical science are exploring ways to automate knowledge discovery from a vast repository of published scientific work. This work attempts to accelerate the process of information extraction by extending Kepler, a graphical workflow management tool. Kepler provides a simple way of designing and executing complex workflows in the form of directed graphs. This work presents a scalable approach to convert published research as PDF documents into indexable XML documents using Kepler. This conversion is a critical step in the Natural Language Processing pipeline. Kepler's distributed data processing capability enables scientists to scale this critical computation by simply adding more computing resources over the cloud.

Abstract: In this paper we present the approach proposed by EU H2020 INDIGO-DataCloud project to orchestrate dynamic workflows over a cloud environment. The main focus of the project is on the development of open source Platform as a Service solutions targeted at scientific communities, deployable on multiple hardware platforms, and provisioned over hybrid e-Infrastructures. The project is addressing many challenging gaps in current cloud solutions, responding to specific requirements coming from scientific communities including Life Sciences, Physical Sciences and Astronomy, Social Sciences and Humanities, and Environmental Sciences. We are presenting the ongoing work on implementing the whole software chain on the Infrastructure as a Service, PaaS and Software as a Service layers, focusing on the scenarios involving scientific workflows and big data analytics frameworks. INDIGO module for Kepler worflow system has been introduced along with the INDIGO underlying services exploited by the workflow components. A climate change data analytics experiment use case regarding the precipitation trend analysis on CMIP5 data is described, that makes use of Kepler and big data analytics services.

Abstract: Our work addresses the influence of the growing adoption of information and communication technologies (ICT) for campaigning purposes on the evolving dynamics of information flows from the eminently social beings that candidates are. Our approach combines an analysis of contents to technological and methodological concerns. The breadth of these objectives as well as the amount of data to be considered pointed to a need for collaboration between several researchers for sharing out tasks and bringing together expertise from various disciplines. This paper presents results concerning three of the data collections life cycle phases: collection, cleaning, and storage. The result is a data collection ready to be analysed for different purposes. In particular, in our experimental validation it has been used for comparing political campaigns behaviour in France and the United Kingdom during the European elections in 2015.

Abstract: We study a special bilevel programming problem that arises in transactions between a Natural Gas Shipping Company and a Pipeline Operator. Because of the business relationships between these two actors, the timing and objectives of their decision-making process are different. In order to model that, bilevel programming was traditionally used in previous works. The problem theoretically studied to facilitate its solution; this included linear reformulation, heuristic approaches, and branch-and-bound techniques. We present a linear programming reformulation of the latest version of the model, which is easier and faster to solve numerically. This reformulation makes it easier to theoretically analyze the problem, allowing us to draw some conclusions about the nature of the solution. Since elements of uncertainty are definitely present in the bilevel natural gas cash-out problem, its stochastic formulation is developed in the form of a bilevel multi-stage stochastic programming model with recourse. After reducing the original formulation to a bilevel linear problem, a stochastic scenario tree is defined by its node events, and time series forecasting is used to produce stochastic values for data of natural gas price and demand. Numerical experiments were run to compare the stochastic solution with the perfect information solution and the expected value solutions.

Abstract: This paper considers a real life case study that determines the minimum number of sellers required to attend a set of customers located in a certain region taking into account the weekly schedule plan of the visits, as well as the optimal route. The problem is formulated as a combination of assignment, scheduling and routing problems. In the new formulation, case studies of small size subset of customers of the above type can be solved optimally. However, this subset of customers is not representative within the business plan of the company. To overcome this limitation, the problem is divided into three phases. A greedy algorithm is used in Phase I in order to identify a set of cost-effective feasible clusters of customers assigned to a seller. Phase II and III are then used to solve the problem of a weekly program for visiting the customers as well as to determine the route plan using MILP formulation. Several real life instances of different sizes have been solved demonstrating the efficiency of the proposed approach.

Abstract: In the last years development of exascale computing technology lead to need of obtaining evaluation for the energy consumption when large-scale problems are solved with different high-performance computing (HPC) systems. In this paper we study the energy efficiency of a class of Monte Carlo (MC) and Quasi-Monte Carlo(QMC) algorithms for given integral equation using hybrid HPC systems. The algorithms are applied to solve quantum kinetic integral equations describing ultra-fast transport in quantum wire. We compare the energy performance for both algorithms using a GPU-based computer platform and CPU-based computer platform with and without hyper threading (HT) technology. We use SPRNG library and CURAND generator to produce parallel pseudo-random (PPR) sequences in case of MC algorithms over CPU-based and GPU -based platforms, respectively. In the case of QMC algorithms Sobol and Halton sequences are used to produce parallel quasi-random (PQR) sequences. We compare the obtained results by the tested algorithms with respect to given energy metrics. The results of our study demonstrate the importance to take into account not only scalability of the HPC intensive algorithms but also their energy efficiency. They also show the need of further optimisation of the QMC algorithms when a GPU-based computing platform are used.

Abstract: Multiscale Modelling and Simulation (MMS) is a cornerstone in the today’s research in computational science. Simulations containing multiple models, with each model operating at a different temporal or spatial scale, are a challenging setting that frequently require innovative approaches in areas such as scale bridging, code deployment, error quantification, and scientific analysis. The aim of the MMS workshop is to encourage and consolidate the progress in this multidisciplinary research field, both in the areas of the scientific applications and the underlying infrastructures that enable these applications. Here we briefly introduce the scope of the workshop and highlight some of the key aspects of this this year’s submissions.

Abstract: Simulation of charge transport in disordered organic materials requires a huge number of quantum mechanical calculations of charge hopping parameters which are then used to compute important macroscopic properties such as the charge mobility. We present the realization of the quantum patch approach to solve a tightly coupled multiscale model for charge transport in organic materials. In contrast to previously used models, this model includes the effect of the electrostatic environment of the molecules on the energy disorder (the so-called polaron effect) explicitly and self-consistently on the quantum mechanics level. This gives rise to tasks of very different resource footprints and, on the other hand, to dependencies between very large number of tasks, representing a considerable computational challenge. Our solution concept is based on embedding the quantum mechanics tasks into a workflow, which accounts for the dependencies arising from the self-consistency loops, and applies a specific scheduling strategy based on the computational characteristics of the different task types. We have implemented the model as part of the software package Shredder and show how the implementation exploits the inherent parallelism of the multiscale model and effectively alleviate the effects of load imbalance and dependencies. The model can be used to virtually explore properties of numerous organic materials using high performance computing and so to optimize material composition, morphology and manufacturing processes.

Abstract: The ever increasing demand for higher levels of detail and accuracy in modeling of complex systems has led researchers in recent years to turn to multiscale modeling (MSM) as a mechanism to extend traditional models. The creation of multiscale models for a complex system largely involves identification of the individual scales relevant to the system and their integration into a single encompassing model. Our work is focused on the development of an adaptive computational framework for MSM that allows for rapid construction of multiscale models through composition of individual at-scale models. Our primary focus is on new scalable numerical algorithms applicable to a wide range of MSM applications. These algorithms include: i) adaptive computational strategies for MSM, ii) algorithms for scale-bridging in MSM, and iii) algorithms for development of surrogate models to reduce the computational cost associated with MSM. We will describe our computational framework for MSM and highlight its use in developing a multiscale model of an energetic material and a high-throughput capability for battery research.

Abstract: We present FabSim, a toolkit developed to simplify a range of computational tasks for researchers in diverse disciplines. FabSim is flexible, adaptable, and allows users to perform a wide range of tasks with ease. It also provides a systematic way to automate the use of resourcess, including HPC and distributed resources, and to make tasks easier to repeat by recording contextual information. To demonstrate this, we present three use cases where FabSim has enhanced our research productivity. These include simulating cerebrovascular bloodflow, modelling clay-polymer nanocomposites across multiple scales, and calculating ligand-protein binding affinities.

Abstract: Coupled multi-model simulation strategies used in various scientific such as climate, MHD, nuclear reactors and subsurface physics implementations that are tuned to achieve high efficiency while satisfying problem-dependent approximations. Often, highly specialized, domain specific tools are preferred for accurate resolution of characteristic scales in these models due to inherent difficulties in generalizing the workflow in complex multi-scale applications. In this talk, we conduct a survey of several successful frameworks and categorize them based on scalable and flexible algorithms exposed under four specific operators applied on coupled solution data: specification, transformation, transfer and orchestration. We also identify essential attributes to improve scientific productivity in terms of feature extensibility, supported numerical algorithms, computational efficiency, and software abstractions that would make an ideal coupling tool for a variety of applications.