Simulation Speedup Research Articles

Enhancing 3D planetary atmosphere simulations with a surrogate radiative transfer model

Abstract This work introduces an approach to enhancing the computational efficiency of 3D atmospheric simulations by integrating a machine-learned surrogate model into the OASIS global circulation model (GCM). Traditional GCMs, which are based on repeatedly numerically integrating physical equations governing atmospheric processes across a series of time-steps, are time-intensive, leading to compromises in spatial and temporal resolution of simulations. This research improves upon this limitation, enabling higher resolution simulations within practical timeframes. Speeding up 3D simulations holds significant implications in multiple domains. Firstly, it facilitates the integration of 3D models into exoplanet inference pipelines, allowing for robust characterisation of exoplanets from a previously unseen wealth of data anticipated from JWST and post-JWST instruments. Secondly, acceleration of 3D models will enable higher resolution atmospheric simulations of Earth and Solar System planets, enabling more detailed insights into their atmospheric physics and chemistry. Our method replaces the radiative transfer module in OASIS with a recurrent neural network-based model trained on simulation inputs and outputs. Radiative transfer is typically one of the slowest components of a GCM, thus providing the largest scope for overall model speed-up. The surrogate model was trained and tested on the specific test case of the Venusian atmosphere, to benchmark the utility of this approach in the case of non-terrestrial atmospheres. This approach yields promising results, with the surrogate-integrated GCM demonstrating above 99.0% accuracy and 147 factor GPU speed-up of the entire simulation compared to using the matched original GCM under Venus-like conditions.

Contraction of ZX diagrams with triangles via stabiliser decompositions

Recent advances in classical simulation of Clifford+T circuits make use of the ZX calculus to iteratively decompose and simplify magic states into stabiliser terms We improve on this method by studying stabiliser decompositions of ZX diagrams involving the triangle operation. We show that this technique greatly speeds up the simulation of quantum circuits involving multi-controlled gates which can be naturally represented using triangles. We implement our approach in the QuiZX library (2022 A. Kissinger amd J. van de Wetering Quantum Science and Technology 7, 044001), (2022 A. Kissinger et al F. Le Gall and T. Morimae, ed. 17th Conference on the Theory of Quantum Computation, Communication and Cryptography (TQC 2022), Leibniz International Proceedings in Informatics (LIPIcs) 232, Schloss Dagstuhl—Leibniz-Zentrum für Informatik, Dagstuhl, Germany, pp 5:1–5:13) and demonstrate a significant simulation speed-up (up to multiple orders of magnitude) for random circuits and a variation of previously used benchmarking circuits. Furthermore, we use our software to contract diagrams representing the gradient variance of parametrised quantum circuits, which yields a tool for the automatic numerical detection of the barren plateau phenomenon in ansätze used for quantum machine learning. Compared to traditional statistical approaches, our method yields exact values for gradient variances and only requires contracting a single diagram. The performance of this tool is competitive with tensor network approaches, as demonstrated with benchmarks against the quimb library (2018 J. Gray Journal of Open Source Software 3, 819).

Pac-Sim: Simulation of Multi-threaded Workloads using Intelligent, Live Sampling

High-performance, multi-core processors are the key to accelerating workloads in several application domains. To continue to scale performance at the limit of Moore’s Law and Dennard scaling, software and hardware designers have turned to dynamic solutions that adapt to the needs of applications in a transparent, automatic way. For example, modern hardware improves its performance and power efficiency by changing the hardware configuration, like the frequency and voltage of cores, according to a number of parameters, such as the technology used or the workload running at the time. With this level of dynamism, it is essential to simulate next-generation multi-core processors in a way that can both respond to system changes and accurately determine system performance metrics. Currently, no sampled simulation platform can achieve these goals of dynamic, fast, and accurate simulation of multi-threaded workloads. In this work, we propose a solution that allows for fast, accurate simulation in the presence of both hardware and software dynamism. To accomplish this goal, we present Pac-Sim, a novel sampled simulation methodology for fast, accurate sampled simulation that requires no upfront analysis of the workload. With our proposed methodology, it is now possible to simulate long-running dynamically scheduled multi-threaded programs with significant simulation speedups, even in the presence of dynamic hardware events. We evaluate Pac-Sim using the SPEC CPU2017, NPB, and PARSEC multi-threaded benchmarks with both static and dynamic thread scheduling. The experimental results show that Pac-Sim achieves a very low sampling error of 1.63% and 3.81% on average for statically and dynamically scheduled benchmarks, respectively. Pac-Sim also demonstrates significant simulation speedups as high as 523.5 × (210.3 × on average) for the training input set of SPEC CPU2017 running eight threads.

Quantum Computing for High-Energy Physics: State of the Art and Challenges

Quantum computers offer an intriguing path for a paradigmatic change of computing in the natural sciences and beyond, with the potential for achieving a so-called quantum advantage—namely, a significant (in some cases exponential) speedup of numerical simulations. The rapid development of hardware devices with various realizations of qubits enables the execution of small-scale but representative applications on quantum computers. In particular, the high-energy physics community plays a pivotal role in accessing the power of quantum computing, since the field is a driving source for challenging computational problems. This concerns, on the theoretical side, the exploration of models that are very hard or even impossible to address with classical techniques and, on the experimental side, the enormous data challenge of newly emerging experiments, such as the upgrade of the Large Hadron Collider. In this Roadmap paper, led by CERN, DESY, and IBM, we provide the status of high-energy physics quantum computations and give examples of theoretical and experimental target benchmark applications, which can be addressed in the near future. Having in mind hardware with about 100 qubits capable of executing several thousand two-qubit gates, where possible, we also provide resource estimates for the examples given using error-mitigated quantum computing. The ultimate declared goal of this task force is therefore to trigger further research in the high-energy physics community to develop interesting use cases for demonstrations on near-term quantum computers. Published by the American Physical Society 2024

Technology Focus: Reservoir Simulation (July 2024)

The early days of artificial intelligence (AI) and machine learning (ML) were filled with promise, but now we’re witnessing a wave of practical applications transforming reservoir engineering. The integration of AI/ML with traditional workflows marks a turning point, unleashing the immense potential of these proven techniques to address our everyday challenges in reservoir simulation. These advancements lead to more-accurate models and faster simulation turnaround times, allowing us to model various scenarios more efficiently. These techniques enable us to understand and simulate reservoir behavior with greater fidelity and focus on the underlying physics. This deeper understanding translates to more-confident decisions regarding reservoir management and development, with a clearer picture of potential risks. Physics-informed machine learning emerges as a significant step forward, improving our understanding and providing better models while boosting runtimes, convergence, and overall performance. Paper IPTC 23730 highlights its application in tackling computationally intensive tasks such as critical temperature prediction with high accuracy, leading to significant speed-ups in simulations, particularly for complex compositional models undergoing miscible gas injection. Paper IPTC 23935 presents a fundamentally different approach to reservoir simulation studies. Adaptive models use multiple smaller, specialized models instead of one giant model. These models can be developed and run independently, allowing for parallel workflows and significantly reduced turnaround times. Adaptive models foster a culture of rapid experimentation and iteration, aligning perfectly with the fail-fast approach—prioritizing the quick evaluation of ideas and discarding those that don’t show promise early on. Paper SPE 214855 provides a comprehensive overview of a complex problem—wellbore modeling. This paper offers practical insights directly relevant to my work. These advancements represent a significant leap forward, paving the way for more-powerful and -versatile reservoir models. As research and development continue, these techniques are poised to revolutionize how we manage and optimize oil and gas reservoirs. I hope you enjoy reading this selection of papers and find them enlightening. Recommended additional reading at OnePetro: www.onepetro.org. SPE 216722 Developing Consistent Relative Permeability and Capillary Pressure Models for Reservoir Simulation of Carbon Capture and Storage Projects by L.S. Lun, ExxonMobil, et al.

Accelerating particle-in-cell kinetic plasma simulations via reduced-order modeling of space-charge dynamics using dynamic mode decomposition.

We present a data-driven reduced-order modeling of the space-charge dynamics for electromagnetic particle-in-cell (EMPIC) plasma simulations based on dynamic mode decomposition (DMD). The dynamics of the charged particles in kinetic plasma simulations such as EMPIC is manifested through the plasma current density defined along the edges of the spatial mesh. We showcase the efficacy of DMD in modeling the time evolution of current density through a low-dimensional feature space. Not only do such DMD-based predictive reduced-order models help accelerate EMPIC simulations, they also have the potential to facilitate investigative analysis and control applications. We demonstrate the proposed DMD-EMPIC scheme for reduced-order modeling of current density and speedup in EMPIC simulations involving electron beam under the influence of magnetic field, virtual cathode oscillations, and backward wave oscillator.

Wurster fluidised-bed coating: Coarse-graining technique within CFD-DEM in conjunction with heat and mass transfer

In this study, we use a combined CFD-Discrete Element Method to assess predictive capabilities and numerical implications of a coarse-graining technique in Wurster fluidised-bed coaters. We investigated both hydrodynamics and heat and mass transfer and conducted simulations of a full three-phase system for the original and three coarse-grained systems, analysing velocity distributions, macroscopic solid stresses, moisture content and phase temperatures. We achieved a logarithmic simulation speed-up by aggregating up to 64 original particles into each coarse grain. This was accomplished while maintaining fidelity to the original CFD-DEM system in terms of reproducing with high accuracy macroscopic granular flow properties in different regions of a coater (drying, tube and bed regions). By integrating a liquid spray and humid air, we demonstrated that the phase temperatures were accurately predicted within the coarse-grained system, with a high capability of delivering liquid spray distributions with the same uniformity and drying. We also give arguments for choosing a certain degree of coarse-graining as a compromise between a desired reduction of computational costs and a trustworthy reproduction of granular-flow physics encountered in different regions of a Wurster bed. Our findings pave the way to using CFD-DEM to industrially-scaled Wurster-bed systems, which is currently unfeasible due to prohibitive computational costs.

Implicit particle-in-cell development for ion source plasmas

Particle-in-Cell (PIC) codes used to study plasma dynamics within ion sources typically use an explicit scheme. These methods can be slow when simulating regions of high electron density in ion sources, which require resolving the Debye length in space and the plasma frequency in time. Recent developments on fully-implicit PIC models in curvilinear geometries have shown that these spatial/time scales can be significantly decreased/increased respectively, allowing for notable speed-ups in simulation time, and thus making it a potential tool for studying the physics of ion sources. For this purpose, a charge and energy conserving implicit PIC code has been developed in 1D to determine its potential for simulating bounded plasmas. In this paper, we use this model to simulate a 1D benchmark of a bounded plasma with fixed plasma density and electron/ion temperatures. The results are shown to compare well to the benchmark and to the results using an explicit PIC code. It is shown that the total amount of macro-particles used in the simulation is a better figure of merit for accurate results than the standard particles per cell used in literature. Significant speed-ups in computation time can be achieved for high plasma densities if the accuracy requirements are relaxed. In this case, we demonstrate the ability of the implicit PIC code to speed-up simulation time by nearly a factor of 12 compared to explicit PIC.

Hyperparameter Tuning with Gaussian Processes for Optimal Abstraction Control in Simulation-based Optimization of Smart Semiconductor Manufacturing Systems

Smart manufacturing utilizes digital twins that are virtual forms of their production plants for analyzing and optimizing decisions. Digital twins have been mainly developed as discrete-event models (DEMs) to represent the detailed and stochastic dynamics of productions in the plants. The optimum decision is achieved after simulating the DEM-based digital twins under various what-if decision candidates; thus, simulation acceleration is crucial for rapid optimum determination for given problems. For the acceleration of discrete-event simulations, adaptive abstraction-level conversion approaches have been previously proposed to switch active models of each machine group between a set of DEM components and a corresponding lookup table-based mean-delay model during runtime. The switching is decided by detecting the machine group’s convergence into (or divergence from) a steady state. However, there is a tradeoff between speedup and accuracy loss in the adaptive abstraction convertible simulation (AACS), and inaccurate simulation can degrade the quality of the optimum (i.e., the distance between the calculated optimum and the actual optimum). In this paper, we propose a simulation-based optimization (SBO) that optimizes the problem based on a genetic algorithm (GA) while tuning specific hyperparameters (related to the tradeoff control) to maximize the speedup of AACS under a specified accuracy constraint. For each individual, the proposed method distributes the overall computing budget for multiple simulation runs (considering the digital twin’s probabilistic property) into the hyperparameter optimization (HPO) and fitness evaluation. We propose an efficient HPO method that manages multiple Gaussian process models (as speedup-estimation models) to acquire promising optimal hyperparameter candidates (that maximize the simulation speedups) with few attempts. The method also reduces each individual’s exploration overhead (as the population evolves) by estimating each hyperparameter’s expected speedup using previous exploration results of neighboring individuals without actual simulation executions. The proposed method was applied to optimize raw-material releases of a large-scale manufacturing system to prove the concept and evaluate the performance under various situations.

Ab initio electron-lattice downfolding: Potential energy landscapes, anharmonicity, and molecular dynamics in charge density wave materials

The interplay of electronic and nuclear degrees of freedom presents an outstanding problem in condensed matter physics and chemistry. Computational challenges arise especially for large systems, long time scales, in nonequilibrium, or in systems with strong correlations. In this work, we show how downfolding approaches facilitate complexity reduction on the electronic side and thereby boost the simulation of electronic properties and nuclear motion-in particular molecular dynamics (MD) simulations. Three different downfolding strategies based on constraining, unscreening, and combinations thereof are benchmarked against full density functional calculations for selected charge density wave (CDW) systems, namely 1H-TaS_22, 1T-TiSe_22, 1H-NbS_22, and a one-dimensional carbon chain. We find that the downfolded models can reproduce potential energy surfaces on supercells accurately and facilitate computational speedup in MD simulations by about five orders of magnitude in comparison to purely ab initio calculations. For monolayer 1H-TaS_22 we report classical and path integral replica exchange MD simulations, revealing the impact of thermal and quantum fluctuations on the CDW transition.

Non-stabilizerness and entanglement from cat-state injection

Recently, cat states have been used to heuristically improve the runtime of a classical simulator of quantum circuits based on the diagrammatic ZX-calculus. Here we investigate the use of cat-state injection within the quantum circuit model. We explore a family of cat states, catm∗ , and describe circuit gadgets using them to concurrently inject non-stabilizerness (also known as magic) and entanglement into any quantum circuit. We provide numerical evidence that cat-state injection does not lead to speed-up in classical simulation. On the other hand, we show that our gadgets can be used to widen the scope of compelling applications of cat states. Specifically, we show how to leverage them to achieve savings in the number of injected qubits, and also to induce scrambling dynamics in otherwise non-entangling Clifford circuits in a controlled manner.

Celeritas: Accelerating Geant4 with GPUs

Celeritas: Accelerating Geant4 with GPUs

Calibration of an expeditious terramechanics model using a higher‐fidelity model, Bayesian inference, and a virtual bevameter test

AbstractThe soil contact model (SCM) is widely used in practice for off‐road wheeled vehicle mobility studies when simulation speed is important and highly accurate results are not a main concern. In practice, the SCM parameters are obtained via a bevameter test, which requires a complex apparatus and experimental procedure. Here, we advance the idea of running a virtual bevameter test using a high‐fidelity terramechanics simulation. The latter employs the “continuous representation model” (CRM), which regards the deformable terrain as an elasto‐plastic continuum that is spatially discretized using the smoothed particle hydrodynamics method. The approach embraced is as follows: a virtual bevameter test is run in simulation using CRM terrain to generate “ground truth” data; in a Bayesian framework, this data is subsequently used to calibrate the SCM terrain. We show that (i) the resulting SCM terrain, while leading to fast terramechanics simulations, serves as a good proxy for the more complex CRM terrain; and (ii) the SCM‐over‐CRM simulation speedup is roughly one order of magnitude. These conclusions are reached in conjunction with two tests: a single wheel test, and a full rover simulation. The SCM and CRM simulations are run in the open‐source software Chrono. The calibration is performed using PyMC, which is a Python package that interactively communicates with Chrono to calibrate SCM. The models and scripts used in this contribution are available as open source for unfettered use and distribution in a public repository.

Assessment of the Capabilities of a Simplification Algorithm for Building Energy Modelling for the Evaluation of Control Strategies: a Case Study in Bolzano, Italy

The increase of the energy efficiency of the building stock is a national priority, especially in consideration of the recent energy crisis and of the 2050 decarbonisation goals of the European Union. Public buildings, in particular, are expected to lead the way and give examples of best practices and solutions for energy savings, some of them related to optimized controls of the building systems. In this context, however, many public administrations lack detailed technical competences, time, and resources to define and implement the best energy efficiency measures and controls for public buildings. To address these needs, fast and accurate simplified models for buildings energy simulation appear to be a promising solution. In this context, this research aims at evaluating the applicability and reliability of an approach previously proposed by the authors, i.e., the shoeboxing algorithm, for the study of operation control strategies. The analysis has been conducted on a public kindergarten, in Bolzano, Italy, comparing simplified and detailed building energy models. Results have shown a fairly good level of accuracy of the algorithm and consistency of energy savings, with a remarkably high simulation speedup.

Computational scaling in inverse photonic design through factorization caching

Inverse design coupled with adjoint optimization is a powerful method to design on-chip nanophotonic devices with multi-wavelength and multi-mode optical functionalities. Although only two simulations are required in each iteration of this optimization process, these simulations still make up the vast majority of the necessary computations and render the design of complex devices with large footprints computationally infeasible. Here, we introduce a multi-faceted factorization caching approach to drastically simplify the underlying computations in finite-difference frequency-domain (FDFD) simulations and significantly reduce the time required for device optimization. Specifically, we cache the numerical and symbolic factorizations for the solution of the corresponding system of linear equations in discretized FDFD simulations and re-use them throughout the device design process. As proof-of-concept demonstrations of the resulting computational advantage, we present simulation speedups reaching as high as 9.2× in the design of broadband wavelength and mode multiplexers compared to conventional FDFD methods. We also show that factorization caching scales well over a broad range of footprints independent of the device geometry, from as small as 16μm2 to over 7000μm2. Our results present significant enhancements in the computational efficiency of inverse photonic design and can greatly accelerate the use of machine-optimized devices in future photonic systems.

A Quasi Time-Reversible Scheme Based on Density Matrix Extrapolation on the Grassmann Manifold for Born-Oppenheimer Molecular Dynamics.

This Letter introduces the so-called Quasi Time-Reversible scheme based on Grassmann extrapolation (QTR G-Ext) of density matrices for an accurate calculation of initial guesses in Born-Oppenheimer Molecular Dynamics (BOMD) simulations. The method shows excellent results on four large molecular systems that are representative of real-life production applications, ranging from 21 to 94 atoms simulated with Kohn-Sham (KS) density functional theory surrounded with a classical environment with 6k to 16k atoms. Namely, it clearly reduces the number of self-consistent field iterations while at the same time achieving energy-conserving simulations, resulting in a considerable speed-up of BOMD simulations even when tight convergence of the KS equations is required.

A Generalized Block-Matrix Circuit for Closed-Loop Analog In-Memory Computing

Matrix-based computing is ubiquitous in an increasing number of present-day machine learning applications such as neural networks, regression and 5G communications. Conventional systems based on von-Neumann architecture are limited by the energy and latency bottleneck induced by the physical separation of the processing and memory units. Inmemory computing (IMC) is a novel paradigm where computation is performed directly within the memory, thus eliminating the need for constant data transfer. IMC has shown exceptional throughput and energy efficiency when coupled with crosspoint arrays of resistive memory devices in open-loop matrix-vector-multiplication and closed-loop inverse-matrix-vector multiplication (IMVM) accelerators. However, each application result in a different circuit topology, thus complicating the development of reconfigurable, general-purpose IMC systems. In this paper, we present a generalized closed-loop IMVM circuit capable of performing any linear matrix operation by proper memory remapping. We derive closed-form equations for the ideal input/ output transfer functions, static error, and dynamic behavior, introducing a novel continuous-time analytical model allowing for orders-of-magnitude simulation speedup with respect to SPICE-based solvers. The proposed circuit represents an ideal candidate for general-purpose accelerators of machine learning.

A machine-learning digital-twin for rapid large-scale solar-thermal energy system design

In many industrialized regions of the world, large-scale photovoltaic systems now contribute a significant part to the energy portfolio during daylight operation. However, as energy demands peak shortly before sunset and persist for several hours afterwards, the integration of solar-thermal systems is extremely advantageous as a green “bridge” energy source. Accordingly, this work develops a digital-twin model to track and optimize the flow of incoming solar power through a complex solar-thermal storage system, consisting of a large array of adaptable mirrors, an optical-receiver and a power distribution system for customers to extract energy. Specifically, the solar power flow is rapidly computed with a reduced order model of Maxwell’s equations, based on a high-frequency decomposition of the irradiance into multiple rays that experience mirror reflections, losses and ultimately receiver absorption and customer delivery. The method allows for rapid testing (in microseconds) of the performance of large numbers of mirror-receiver layout configurations in design space, over extremely long time periods, such as weeks, months and years, using a genetic-based machine-learning digital-twin framework, which integrates submodels for: •optics and tracking of the Fresnel multi-mirror system,•thermal absorption of the optical energy by the receiver and•optimal operating temperatures balancing radiative losses with heat storage. The overall machine-learning digital-twin optimizes the configuration layout to balance meeting customer demands and operational efficiency. Numerical examples are provided to illustrate the approach. Finally, a deep-learning algorithm is developed and applied to the create an Artificial Neural-Net representation, which allows for even further simulation speedup.

Machine learning interatomic potential for silicon-nitride (Si3N4) by active learning.

Silicon nitride (Si3N4) is an extensively used material in the automotive, aerospace, and semiconductor industries. However, its widespread use is in contrast to the scarce availability of reliable interatomic potentials that can be employed to study various aspects of this material on an atomistic scale, particularly its amorphous phase. In this work, we developed a machine learning interatomic potential, using an efficient active learning technique, combined with the Gaussian approximation potential (GAP) method. Our strategy is based on using an inexpensive empirical potential to generate an initial dataset of atomic configurations, for which energies and forces were recalculated with density functional theory (DFT); thereafter, a GAP was trained on these data and an iterative re-training algorithm was used to improve it by learning on-the-fly. When compared to DFT, our potential yielded a mean absolute error of 8 meV/atom in energy calculations for a variety of liquid and amorphous structures and a speed-up of molecular dynamics simulations by 3-4 orders of magnitude, while achieving a first-rate agreement with experimental results. Our potential is publicly available in an open-access repository.

Modeling for Complex Modular Power Electronic Transformers Using Parallel Computing

The modular power electronic transformer (PET) faces difficulty carrying out microsecond-level electromagnetic transient simulations. This article provides a high-speed and high-precision simulation method capable of eliminating the internal nodes and reducing the order of the nodal admittance matrix. Meanwhile, the parallel computing is integrated into the whole solution process, which achieves a significant simulation speedup. A physical prototype is established to prove that the detailed model (DM) is sufficient to reflect the dynamics of physical devices. Moreover, simulations in PSCAD/EMTDC are carried out to compare the proposed method with the DM in terms of accuracy and time efficiency. Simulation results show that the proposed method is accurate to simulate the external and internal dynamics of PET with hundreds of times simulation speed acceleration.