Multiscale Computing Research Articles

A novel paradigm for solving PDEs: multi-scale neural computing

A novel paradigm for solving PDEs: multi-scale neural computing

Neural-Parareal: Self-improving acceleration of fusion MHD simulations using time-parallelisation and neural operators

The fusion research facility ITER is currently being assembled to demonstrate that fusion can be used for industrial energy production, while several other programmes across the world are also moving forward, such as EU-DEMO, CFETR, SPARC and STEP. The high engineering complexity of a tokamak makes it an extremely challenging device to optimise, and test-based optimisation would be too slow and too costly. Instead, digital design and optimisation must be favoured, which requires strongly-coupled suites of multi-physics, multi-scale High-Performance Computing calculations. Safety regulation, uncertainty quantification, and optimisation of fusion digital twins is undoubtedly an Exascale grand challenge. In this context, having surrogate models to provide quick estimates with uncertainty quantification is essential to explore and optimise new design options. But there lies the dilemma: accurate surrogate training first requires simulation data. Extensive work has explored solver-in-the-loop solutions to maximise the training of such surrogates. Likewise, innovative methods have been proposed to accelerate conventional HPC solvers using surrogates. Here, a novel approach is designed to do both. By bootstrapping neural operators and HPC methods together, a self-improving framework is achieved. As more simulations are being run within the framework, the surrogate improves, while the HPC simulations get accelerated. This idea is demonstrated on fusion-relevant MHD simulations, where Fourier Neural Operator based surrogates are used to create neural coarse-solver for the Parareal (time-parallelisation) method. Parareal is particularly relevant for large HPC simulations where conventional spatial parallelisation has saturated, and the temporal dimension is thus parallelised as well. This Neural-Parareal framework is a step towards exploiting the convergence of HPC and AI, where scientists and engineers can benefit from automated, self-improving, ever faster simulations. All data/codes developed here are made available to the community.

Perspective on machine learning in energy material discovery

Perspective on machine learning in energy material discovery

Special issue on SoC and AI processors

Special issue on SoC and AI processors

Towards Heterogeneous Multi-scale Computing on Large Scale Parallel Supercomputers

New applications that can exploit emerging exascale computing resources efficiently, while providing meaningful scientific results, are eagerly anticipated. Multi-scale models, especially multi-scale applications, will assuredly run at the exascale. We have established that a class of multi-scale applications implementing the heterogeneous multi-scale model follows, a heterogeneous multi-scale computing (HMC) pattern, which typically features a macroscopic model synchronising numerous independent microscopic model simulations. Consequently, communication between microscopic simulations is limited. Furthermore, a surrogate model can often be introduced between macro-scale and micro-scale models to interpolate required data from previously computed micro-scale simulations, thereby substantially reducing the number of micro-scale simulations. Nonetheless, HMC applications, though versatile, remain constrained by load balancing issues. We discuss two main issues: the a priori unknown and variable execution time of microscopic simulations, and the dynamic number of micro-scale simulations required. We tackle execution time variability using a pilot job mechanism to handle internal queuing and multiple sub-model execution on large-scale supercomputers, together with a data-informed execution time prediction model. To dynamically select the number of micro-scale simulations, the HMC pattern automatically detects and identifies three surrogate model phases that help control the available and used core amount. After relevant phase detection and micro-scale simulation scheduling, any idle cores can be used for surrogate model update or for processor release back to the system. We demonstrate HMC performance by testing it on two representative multi-scale applications. We conclude that, considering the subtle interplay between the macroscale model, surrogate models and micro-scale simulations, HMC provides a promising path towards exascale for many multiscale applications.

Multiscale modelling, simulation and computing: from the desktop to the exascale.

This short contribution introduces a theme issue dedicated to 'Multiscale modelling, simulation and computing: from the desktop to the exascale'. It holds a collection of articles presenting cutting-edge research in generic multiscale modelling and multiscale computing, and applications thereof on high-performance computing systems. The special issue starts with a position paper to discuss the paradigm of multiscale computing in the face of the emerging exascale, followed by a review and critical assessment of existing multiscale computing environments. This theme issue provides a state-of-the-art account of generic multiscale computing, as well as exciting examples of applications of such concepts in domains ranging from astrophysics, via material science and fusion, to biomedical sciences. This article is part of the theme issue 'Multiscale modelling, simulation and computing: from the desktop to the exascale'.

Multiscale computing for science and engineering in the era of exascale performance.

In this position paper, we discuss two relevant topics: (i) generic multiscale computing on emerging exascale high-performing computing environments, and (ii) the scaling of such applications towards the exascale. We will introduce the different phases when developing a multiscale model and simulating it on available computing infrastructure, and argue that we could rely on it both on the conceptual modelling level and also when actually executing the multiscale simulation, and maybe should further develop generic frameworks and software tools to facilitate multiscale computing. Next, we focus on simulating multiscale models on high-end computing resources in the face of emerging exascale performance levels. We will argue that although applications could scale to exascale performance relying on weak scaling and maybe even on strong scaling, there are also clear arguments that such scaling may no longer apply for many applications on these emerging exascale machines and that we need to resort to what we would call multi-scaling.This article is part of the theme issue ‘Multiscale modelling, simulation and computing: from the desktop to the exascale’.

Multi-scale high-performance computing in astrophysics: simulating clusters with stars, binaries and planets.

The demand on simulation software in astrophysics has dramatically increased over the last decades. This increase is driven by improvements in observational data and computer hardware. At the same time, computers have become more complicated to program due to the introduction of more parallelism and hybrid hardware. To keep up with these developments, much of the software has to be redesigned. In order to prevent the future need to rewrite again when new developments present themselves, the main effort should go into making the software maintainable, flexible and scalable. In this paper, we explain our strategy for coupling elementary solvers and how to combine them into a high-performance multi-scale environment in which complex simulations can be performed. The elementary parts can remain succinct while supporting the aggregation to more satisfactory functionality by coupling them on a higher level. The advanced code-coupling strategies we present here allow such a hierarchy and support the development of complex codes. A library of simple elementary solvers subsequently stimulates the rapid development of more complex code that can co-evolve with the latest advances in computer hardware. We demonstrate how to combine several of these elementary solvers in a hierarchical and generic system, and how the resulting complex codes can be applied to multi-scale problems in astrophysics. Our aim is to achieve the best of several worlds with respect to performance, flexibility and maintainability while reducing development time. We succeeded in the development of the hierarchical coupling strategy and the general framework, but a comprehensive library of minimal fundamental-physics solvers is still unavailable. This article is part of the theme issue 'Multiscale modelling, simulation and computing: from the desktop to the exascale'.

ComPat framework for multiscale simulations applied to fusion plasmas

Understanding the dependency between plasma turbulence and overall transport is essential in predicting the performance of fusion devices. This dependency is not fully understood because of the highly disparate spatio-temporal scales involved, which prohibits a fully-resolved turbulence simulation at transport scales. The Computer Patterns for High Performance Multiscale Computing (ComPat) project takes the component based approach for multiscale simulations to connect existing single-scale models into a workflow. In this paper, we present the ComPat’s approach in building a multiscale fusion application: it brings equilibrium, transport and turbulence models together, where the turbulence is described by a 3D gyrofluid code. Initial results and challenges encountered with such approach are also presented and discussed. One of the challenges is to ensure numerical stability, for which adaptive time step size and the search for quasi-steady state are implemented as possible solutions. Another challenge is to improve the overall performance of multiscale simulation, and that is addressed by increasing the level of parallelism to the workflow.

Patterns for High Performance Multiscale Computing

We describe our Multiscale Computing Patterns software for High Performance Multiscale Computing. Following a short review of Multiscale Computing Patterns, this paper introduces the Multiscale Computing Patterns Software, which consists of description, optimisation and execution components. First, the description component translates the task graph, representing a multiscale simulation, to a particular type of multiscale computing pattern. Second, the optimisation component selects and applies algorithms to find the most suitable mapping between submodels and available HPC resources. Third, the execution component which a middleware layer maps submodels to the number and type of physical resources based on the suggestions emanating from the optimisation part together with infrastructure-specific metrics such as queueing time and resource availability. The main purpose of the Multiscale Computing Patterns software is to leverage the Multiscale Computing Patterns to simplify and automate the execution of complex multiscale simulations on high performance computers, and to provide both application-specific and pattern-specific performance optimisation. We test the performance and the resource usage for three multiscale models, which are expressed in terms of two Multiscale Computing Patterns. In doing so, we demonstrate how the software automates resource selection and load balancing, and delivers performance benefits from both the end-user and the HPC system level perspectives.

Detecting and Alleviating Stress with SoDA

This installment of Computer’s series highlighting the work published in IEEE Computer Society journals comes from IEEE Transactions on Multi-Scale Computing Systems.

Community effort endorsing multiscale modelling, multiscale data science and multiscale computing for systems medicine.

Systems medicine holds many promises, but has so far provided only a limited number of proofs of principle. To address this road block, possible barriers and challenges of translating systems medicine into clinical practice need to be identified and addressed. The members of the European Cooperation in Science and Technology (COST) Action CA15120 Open Multiscale Systems Medicine (OpenMultiMed) wish to engage the scientific community of systems medicine and multiscale modelling, data science and computing, to provide their feedback in a structured manner. This will result in follow-up white papers and open access resources to accelerate the clinical translation of systems medicine.

Multiscale computing in the exascale era

We expect that multiscale simulations will be one of the main high performance computing workloads in the exascale era. We propose multiscale computing patterns as a generic vehicle to realise load balanced, fault tolerant and energy aware high performance multiscale computing. Multiscale computing patterns should lead to a separation of concerns, whereby application developers can compose multiscale models and execute multiscale simulations, while pattern software realises optimized, fault tolerant and energy aware multiscale computing. We introduce three multiscale computing patterns, present an example of the extreme scaling pattern, and discuss our vision of how this may shape multiscale computing in the exascale era.

MultiGrain/MAPPER: A distributed multiscale computing approach to modeling and simulating gene regulation networks

Modeling and simulation of gene-regulatory networks (GRNs) has become an important aspect of modern systems biology investigations into mechanisms underlying gene regulation. A key task in this area is the automated inference or reverse-engineering of dynamic mechanistic GRN models from gene expression time-course data. Besides a lack of suitable data (in particular multi-condition data from the same system), one of the key challenges of this task is the computational complexity involved. The more genes in the GRN system and the more parameters a GRN model has, the higher the computational load. The computational challenge is likely to increase substantially in the near future when we tackle larger GRN systems. The goal of this study was to develop a distributed computing framework and system for reverse-engineering of GRN models. We present the resulting software called MultiGrain/MAPPER. This software is based on a new architecture and tools supporting multiscale computing in a distributed computing environment. A key feature of MultiGrain/MAPPER is the realization of GRN reverse-engineering based on the underlying distributed computing framework and multi-swarm particle swarm optimization. We demonstrate some of the features of MultiGrain/MAPPER and evaluate its performance using both real and artificial gene expression data.

IEEE Transactions on Multi-Scale Computing Systems House Advertisement

IEEE Transactions on Multi-Scale Computing Systems House Advertisement

IEEE Transactions on Multi-Scale Computing Systems House Advertisement

IEEE Transactions on Multi-Scale Computing Systems House Advertisement

Advancing a distributed multi-scale computing framework for large-scale high-throughput discovery in materials science

We describe the development of a large-scale high-throughput application for discovery in materials science. Our point of departure is a computational framework for distributed multi-scale computation. We augment the original framework with a specialized module whose role is to route evaluation requests needed by the high-throughput application to a collection of available computational resources. We evaluate the feasibility and performance of the resulting high-throughput computational framework by carrying out a high-throughput study of battery solvents. Our results indicate that distributed multi-scale computing, by virtue of its adaptive nature, is particularly well-suited for building high-throughput applications.

Guest Editorial for Special Issue on Emerging Memory Technologies—Modeling, Design, and Applications for Multi-Scale Computing

Guest Editorial for Special Issue on Emerging Memory Technologies—Modeling, Design, and Applications for Multi-Scale Computing

Introduction to IEEE Transactions on Multiscale Computing Systems (TMSCS)

Reports on the introduction of the IEEE Transactions on Multiscale Computing Systems (TMSCS). IEEE TMSCS is a new journal devoted to computing systems that exploit multiscale and multi-functionality. Multiscale and multi-functionality computing are very rich fields. The traditional computing domain is transitioning in to exciting new directions due to several emerging technologies and advanced applications such as: massive many core processing platforms, nano architectures, big data driven applications, mapping of the brain, and molecular-scale computing. Consequently, multiscale and multi-functional computing have become a topic of increasing importance and emphasis in both research and graduate/undergraduate computer science and computer engineering education.

A High Performance Computing Approach to the Simulation of Fluid-Solid interaction Problems with Rigid and Flexible Components

Abstract This work outlines a unified multi-threaded, multi-scale High Performance Computing (HPC) approach for the direct numerical simulation of Fluid-Solid Interaction (FSI) problems. The simulation algorithm relies on the extended Smoothed Particle Hydrodynamics (XSPH) method, which approaches the fluid flow in a La-grangian framework consistent with the Lagrangian tracking of the solid phase. A general 3D rigid body dynamics and an Absolute Nodal Coordinate Formulation (ANCF) are implemented to model rigid and flexible multibody dynamics. The two-way coupling of the fluid and solid phases is supported through use of Boundary Condition Enforcing (BCE) markers that capture the fluid-solid coupling forces by enforcing a no-slip boundary condition. The solid-solid short range interaction, which has a crucial impact on the small-scale behavior of fluid-solid mixtures, is resolved via a lubrication force model. The collective system states are integrated in time using an explicit, multi-rate scheme. To alleviate the heavy computational load, the overall algorithm leverages parallel computing on Graphics Processing Unit (GPU) cards. Performance and scaling analysis are provided for simulations scenarios involving one or multiple phases with up to tens of thousands of solid objects. The software implementation of the approach, called Chrono:Fluid, is part of the Chrono project and available as an open-source software.