Supercomputer Systems Research Articles

Coarse-Grained Molecular Dynamics Simulation on the Impact of the Primary Pores on the Distribution and Morphology of Ionomers within Catalyst Layers during the Drying Process

The catalyst layer (CL) plays a crucial role in polymer electrolyte fuel cells (PEFCs) as it facilitates proton transport by coating the carbon surface with platinum (Pt) particles with the assistance of Nafion ionomers. Given the existence of complex pore structures of supported carbon and their potential impact on the distribution and morphological evolution of the ionomers in the CL of PEFCs, this area is essential while remaining underexplored. Herein, we employed the coarse-grained molecular dynamics (CGMD) simulation method to investigate the impact of the primary pores on the distribution and morphological evolution of ionomers from the Nafion ionomers solution to the Pt/C substrate surface during the drying process of the catalyst layer at the nanoscale. We aim to provide a comprehensive understanding of the behavior of ionomers in this process.To simulate the porous structure of the supported carbon, we employed carbon nanotubes (CNT) structure to represent the primary pores in the MD simulation. By varying the parameters, including the Nafion ionomers to carbon (I/C) ratio, the number and diameter of Pt particles, and the diameter of the CNT, the MD simulation was performed under different conditions. The MD analysis results demonstrated that the structure of the CNT had a significant impact on the distribution and morphology changes of Nafion ionomers during the drying process of the catalyst ink. In the early stage, a small fraction of Nafion ionomers were observed to adsorb directly onto the carbon substrate surface without any interaction with the Pt particles. Meanwhile, the majority of the Nafion ionomers were found to form cylindrical aggregates in the solution, which is attributed to the mutual attraction between the ionomers. During the drying process, the cylindrical Nafion ionomer aggregates were initially attracted to the surface of Pt particles on the carbon substrate due to their hydrophilic sulfonate group shells, undergoing a transformative unfolding upon contacting the carbon substrate. This shift to a lamellar structure is propelled by the hydrophobic attraction of the ionomer backbone and the hydrophilic affinity of the sulfonate groups. Moreover, it is noteworthy that Pt particles with a larger diameter enhance the adsorption of ionomer aggregates by offering a larger surface area during the drying process. Consequently, this leads to the formation of larger bare areas on the carbon substrate surface, even when using a solution with the same I/C ratio. As the drying process further progresses, the Nafion ionomers fully unfold, forming a layered structure on the hydrophobic carbon substrate surface, which finally results in a well-structured ionomer distribution. As the drying process nears completion, ionomer aggregates adhere to the surface of the carbon substrate and form ionomer layers. Specifically, as shown in Fig. 1-c, an adequate I/C ratio will result in the ionomers having a uniform and complete laminar distribution on the carbon substrate along with the edge of CNT. Besides, larger Pt particles resulted in a larger bare area of the carbon substrate (Fig. 1-a). Moreover, the thickness of the ionomer above the primary pore is significantly smaller than that on the surface of the carbon substrate (Fig. 1). These results highlight the significance of taking into account the porous structure when investigating ionomer behavior. The existence of CNT impacts the distribution and morphology of Nafion ionomers during the drying process of the catalyst layer.This study demonstrates the significant influence of the porous structure of primary pores on the distribution and morphological changes of Nafion ionomers in the catalyst layer. These findings offer valuable insights for the development of more effective catalyst inks and optimization of catalyst layer structures. The results also provide a theoretical foundation for enhancing the catalytic efficiency and proton transport performance of PEFCs.AcknowledgementsThe New Energy and Industrial Technology Development Organization (NEDO) of Japan supported this work under Grant number JPNP20003. The simulations were performed on the Supercomputer system 'AFI-NITY' at the Advanced Fluid Information Research Center, Institute of Fluid Science, Tohoku University. We want to express our sincere gratitude to W. Shirakawa for her invaluable assistance. Figure 1

Coarse-Grained Molecular Dynamics Simulation of Ionomer-Mediated Carbon Cluster Bonding in Polymer Electrolyte Fuel Cell Catalyst Layers

In the field of sustainable energy technologies, Polymer Electrolyte Fuel Cells (PEFCs) represent a crucial advancement for achieving high-efficiency power generation. This study focuses on the specific behavior of ionomers within PEFCs, particularly how they facilitate the bonding of carbon clusters during the evaporation process, critical for the structural integrity and functionality of the catalyst layer.Amid the global shift toward a hydrogen economy, the development of robust and efficient PEFC technologies is imperative. The Nafion membrane, developed by DuPont, is the standard bearer for proton exchange membranes due to its superior proton conductivity, chemical stability, and durability. Despite these advantages, the performance of PEFCs under high temperature and low humidity conditions remains a challenge due to the reduced efficiency of proton transport. This research concentrates on the micro-mechanics of how ionomers assist in the bonding between carbon clusters during the drying stages of film formation, aiming to overcome this efficiency bottleneck.Employing Coarse-Grained Molecular Dynamics (CGMD) simulations, our study delves into the detailed interactions between ionomers and Pt-carbon clusters during evaporating solvents. The simulations use a refined computational model to capture the dynamic interplay during evaporation conceptualized as micro-differentiated boxes within the system as illustrated in Figure 1. These boxes allow for the analysis of ionomer behavior on a flat, graphene-sheet-modeled cluster surface adorned with Pt particles, assessing how different configurations influence structural and transport properties.This approach not only elucidates the role of ionomers in promoting the aggregation of carbon clusters during evaporation but also provides insight into the nuanced processes that govern the formation and stability of the catalyst layer. Initial simulations pointed out the need for adjustments in the force field settings, which were subsequently optimized to enhance the accuracy of the observations regarding ionomer distribution and bonding effectiveness during evaporation.With the foundational simulations set, future work will extend to exploring how various experimental conditions—such as solvent composition, and the size and density of platinum particles—affect ionomer bonding between carbon clusters. By adjusting these parameters systematically, we aim to define the optimal conditions that enhance the structural cohesion and functional performance of the catalyst layer.This investigation is expected to yield significant insights into the fundamental interactions and mechanisms at play in the bonding process, contributing to a deeper understanding of PEFC technology and leading to advancements in the production methods of these fuel cells. Ultimately, this research will help to refine and optimize the procedures for manufacturing more efficient and durable PEFCs, marking a substantial step toward more sustainable energy systems.AcknowledgmentThe New Energy and Industrial Technology Development Organization (NEDO) of Japan supported this work under Grant number NP20003.A part of the results was obtained by supercomputer system of Institute of Fluid Science, Tohoku University. Figure 1

Low power computation of transoceanic wave propagation for tsunami hazard mitigation

This paper proposes the use of specialized hardware accelerator based on the Field Programmable Gates Array (FPGA) microchip to compute tsunami wave propagation to assess and manage risks of marine natural disasters, namely, tsunami waves caused by underwater earthquakes. After a sufficiently strong seismic event, many countries and research centres launch extensive computations to estimate the tsunami wave parameters in certain parts of the coast to determine if a declaration of a tsunami alarm is warranted. This requires high computating powers which leads to higher energy costs. The paper demonstrates how an FPGA-based special Calculator (architecture of which has been earlier proposed by the authors), installed on a Personal Computer (PC) could be used to calculate the propagation of a tsunami wave over the entire Pacific Ocean, from the subduction zone offshore Kamchatka Peninsula and Kuril Islands to the coast of Chile. Such calculations offer reliable results within a few minutes and make it possible to obtain the distribution of expected tsunami wave heights along the coast. If the obtained results indicate a danger to the population or possible destruction of infrastructure, it becomes paramount to carry out more detailed calculations to accurately estimate the wave parameters at specific locations along the coast where negative consequences are expected. This requires cluster and/or supercomputer systems, which consume significant energy and hence are expensive. In case the modelling results indicate small values of maximum wave heights at populated coastal areas, population of the near-shore regions can be immediately informed about low amplitude tsunami wave; more detailed studies are not needed. This hence leads to noticeable savings in energy consumption. The paper presents a calculation of the propagation of a tsunami wave across the Pacific Ocean on a personal computer using a FPGA-based hardware acceleration of a computer code execution.

Visualization Methods of Vortex Structures Acting on the Wing in High-Speed Flows

This paper considers the use of vortex structures scientific identification and visualization methods in analyzing the effects of these structures on the wing at an angle of attack in supersonic flows. The results of applying the methods of scientific identification and visualization of vortex flows for the analysis and processing of numerical data of supersonic flow are presented. In the considered problem, the generator wing and the main wing are located at an angle of attack of 20° degrees to the incoming flow. Two configurations with different length of the vortex generator are considered. The simulation was carried out using URANS equations. Numerical simulations were performed on the hybrid supercomputer system K-60 at the Keldysh Institute of Applied Mathematics of the Russian Academy of Sciences. The scientific visualization of the simulation results was carried out primarily using the Liutex vortex structures identification criterion, which belongs to the third generation of such methods. And this method was effectively complemented and enhanced by the use of other methods.

Leveraging the Hardware Resources to Accelerate cryo-EM Reconstruction of RELION on the New Sunway Supercomputer

The fast development of biomolecular structure determination has enabled the fine-grained study of objects in the micro-world, such as proteins and RNAs. The world is benefited. However, as the computational algorithms are constantly developed, the enrichment of features increases the algorithmic complexity and brings more computationally unfriendly modules. It calls for efficient solutions to leverage the rich and various hardware resources from the world’s most state-of-the-art supercomputing systems, and to fully accelerate the performance of the applications. In this paper, we present our efforts on porting and optimizing the 3D reconstruction of RELION, one of the most popular cryo-EM software for biomolecular structure determinations, by leveraging different resources of the latest generation of Sunway heterogeneous supercomputer. Several novel approaches are proposed to resolve different challenges faced by the complex algorithm, including a multi-level parallel scheme and operator optimizations to smartly map and scale RELION, efficient strategies to largely address the memory bottlenecks and improve data locality, lock-free writing solutions to minimize write-write conflicts, and pipelining approaches to obtain excellent computation and communication overlap. Combining all proposed optimizations, the computation time is greatly reduced to under 2 hours, achieving 11.9 × and 8.9 × speedups on two different datasets. The overall design scales to 131,072 cores, increasing parallel efficiency from 33% to 61% and from 46% to 70%, respectively. To the best of our knowledge, this is the first work that fully optimized and scaled the 3D reconstruction of RELION using the latest Sunway system.

Implementation of Parallel Applications on Linear Systolic and Star Topologies by Using Multistage Omega Network

Contemporary computer architecture comprises multicomputer settings. Multiple computers provide the facility of high performance to implement multiple threads that are faster and concurrently with different processors that can execute tasks simultaneously. The challenges are as follows: cost and high performance. Certain firms have to provide funding to establish data centers that need locations, hardware, and technicians. After the passage of a period of time, such equipment requires maintenance (updating and upgrading) to take place on it. This paper tackles these challenges by passing the drawbacks of data centers. It provides an algorithm that simulates a virtual machine, as one client is to be connected with eight servers to implement two applications namely convolution and mathematical operations. To represent different topologies, the algorithm simulates supercomputer systems by applying multistage omega network topology and parallel processing. Two types of topologies execute linear systolic and star, and is implemented on transmission control protocol (TCP) and user datagram protocol (UDP) protocols. The goal of this paper is to provide a kernel that draws near a cloud computing system by exploiting the JAVA programming language with its techniques (threading, socket, socket server, Datagram socket, and datagram packet). The results offer the connection between the client and servers to be successfully simulated by using multistage omega network. The linear systolic and star topologies are simulated in virtual parallel processing.

An efficient eigenvalue bounding method: CFL condition revisited

Direct and large-eddy simulations of turbulence are often solved using explicit temporal schemes. However, this imposes very small time-steps because the eigenvalues of the (linearized) dynamical system, re-scaled by the time-step, must lie inside the stability region. In practice, fast and accurate estimations of the spectral radii of both the discrete convective and diffusive terms are therefore needed. This is virtually always done using the so-called CFL condition. On the other hand, the large heterogeneity and complexity of modern supercomputing systems are nowadays hindering the efficient cross-platform portability of CFD codes. In this regard, our leitmotiv reads: relying on a minimal set of (algebraic) kernels is crucial for code portability and maintenance! In this context, this work focuses on the computation of eigenbounds for the above-mentioned convective and diffusive matrices which are needed to determine the time-step à la CFL. To do so, a new inexpensive method, that does not require to re-construct these time-dependent matrices, is proposed and tested. It just relies on a sparse-matrix vector product where only vectors change on time. Hence, both implementation in existing codes and cross-platform portability are straightforward. The effectiveness and robustness of the method are demonstrated for different test cases on both structured Cartesian and unstructured meshes. Finally, the method is combined with a self-adaptive temporal scheme, leading to significantly larger time-steps compared with other more conventional CFL-based approaches.

Generation and Analysis of a Dataset of Optimal Double-Loop Circulant Networks

Optimal circulant networks are of practical interest as graph models of reliable communication networks for supercomputer systems and networks on a chip. A large dataset of parameters of optimal double-loop circulant networks with a number of nodes of up to 50,000 and 451,000 points has been constructed. The dataset contains, for each number of vertices, all generators corresponding to optimal or suboptimal (in the absence of optimal) descriptions of the graph. This made it possible to find analytical dependencies of the parameters that define families of optimal graphs. Based on the analysis of the constructed dataset, the solution to the problem of determining optimal two-dimensional circulant networks is studied. A process is automated for determining analytical, parametric descriptions of the families of optimal double-loop networks, described polynomials of diameter. Having applied the methods of integrating differential evolution and reduced local search to the analysis of the generated dataset, we rediscovered a number of previously known families and found more than 200 new, different from all those known in literature, families of optimal double-loop networks with generators of linear and quadratic types of polynomials. Our new ap­proach can be applied to studing datasets of other promising classes of graphs and networks.

Chemical simulations of quantum systems using quantum computers - review of algorithms and their experimental verification

Computer simulations using ever-increasing computing power and machine learning techniques allow advanced molecular modelling, molecular dynamics simulations and studies of intermolecular interactions. However, due to the complexity of biological systems and chemical processes at the molecular level, their accurate representation using classical computer models and techniques has faced a number of significant limitations for many years. A new and promising direction for the development of computational science and its potential applications in biochemistry is quantum computing and its integration with classical high-performance supercomputing systems. This article responds to the growing interest in the use of available quantum computers in exemplary applications. In this paper, we aim to provide an overview of the basic notions involved in the development of quantum algorithms and simulations related to issues at the interface of quantum chemistry and biochemistry. In addition, the article introduces the basic principles of performing simulations using the state-of-the-art quantum computers in the era of Noisy Intermediate-Scale Quantum (NISQ). Experimental results of the classical-quantum algorithm Variational Quantum Eigensolver (VQE) for example molecules H2 and CH+ are also presented. Despite the many shortcomings of currently available quantum computers, the analysed VQE algorithm proved to be effective in approximating the ground state of molecules using a minimal functional basis.

Thorough Characterization and Analysis of Large Transformer Model Training At-Scale

Large transformer models have recently achieved great success across various domains. With a growing number of model parameters, a large transformer model training today typically involves model sharding, data parallelism, and model parallelism. Thus, the throughput of large-scale model training depends heavily on the network bandwidth since a combination of model sharding and multiple parallelism strategies incurs various costs. However, prior characterizations of transformer models on high-bandwidth DGX machines that use TFLOPS as a metric may not reflect the performance of a system with lower bandwidth. Furthermore, data and model parallelism reveal significantly distinct training profiles on different system bandwidths at scale and, thus, need a thorough study. In this paper, we provide a bottom-up breakdown of training throughput into compute and communication time, and quantitatively analyze their respective influences on overall end-to-end training scaling. Our evaluation involves an in-depth exploration of data parallelism, scaling up to 512 GPUs with limited bandwidth, and examines three model sharding strategies among six model sizes. We also evaluate three combinations of model parallelism on both high and low bandwidth supercomputing systems. Overall, our work provides a broader perspective on large-scale transformer model training, and our analysis and evaluation yield practical insights for predicting training scaling, shaping the future development of supercomputing system design.

Electronic effect on the L3,2 edges generated by Cu 3d and O 1s orbitals in CuO nanofibers, pure and doped with Mn.

The present work reports a comparative study on CuO and 3.0% Mn-doped CuO nanofibers (NFs) produced by the electrospinning method. State-of-the-art characterization techniques such as Electron Energy Loss Spectroscopy (EELS) and X-Ray photoelectron Spectroscopy (XPS) have been employed to determine the multivalent states and elucidate their electronic structure at the Cu-L3,2 (2p) edge. The CASTEP code has been used to model CuO (P1) and Mn-doped (C 2/c) with the GGA-PBE functional on the supercomputing system with the Xeon10 module (10 nuclei). Structural characterization was done by using a CM 200 Philips Transmission Electron Microscope (TEM) and the chemical state of surfaces with ESCALAB 250Xi XPS Thermoscientific TM. The NFs present an indirect energy bandgap, Eg = 2.03 eV and 2.85 eV, respectively. The results show that the Cu-L3,2 edge presents degenerate states associated with Cu 2p - O and Cu 3d - O hybridization, indicated as Cu3+.

H5bench: A unified benchmark suite for evaluating HDF5 I/O performance on pre‐exascale platforms

SummaryParallel I/O is a critical technique for moving data between compute and storage subsystems of supercomputers. With massive amounts of data produced or consumed by compute nodes, high‐performant parallel I/O is essential. I/O benchmarks play an important role in this process; however, there is a scarcity of I/O benchmarks representative of current workloads on HPC systems. Toward creating representative I/O kernels from real‐world applications, we have created h5bench , a set of I/O kernels that exercise hierarchical data format version 5 (HDF5) I/O on parallel file systems in numerous dimensions. Our focus on HDF5 is due to the parallel I/O library's heavy usage in various scientific applications running on supercomputing systems. The various tests benchmarked in the h5bench suite include I/O operations (read and write), data locality (arrays of basic data types and arrays of structures), array dimensionality (one‐dimensional arrays, two‐dimensional meshes, three‐dimensional cubes), I/O modes (synchronous and asynchronous). In this paper, we present the observed performance of h5bench executed along several of these dimensions on existing supercomputers (Cori and Summit) and pre‐exascale platforms (Perlmutter, Theta, and Polaris). h5bench measurements can be used to identify performance bottlenecks and their root causes and evaluate I/O optimizations. As the I/O patterns of h5bench are diverse and capture the I/O behaviors of various HPC applications, this study will be helpful to the broader supercomputing and I/O community.

Thorough Characterization and Analysis of Large Transformer Model Training At-Scale

Large transformer models have recently achieved great success across various domains. With a growing number of model parameters, a large transformer model training today typically involves model sharding, data parallelism, and model parallelism. Thus, the throughput of large-scale model training depends heavily on the network bandwidth since a combination of model sharding and multiple parallelism strategies incurs various costs. However, prior characterizations of transformer models on high-bandwidth DGX machines that use TFLOPS as a metric may not reflect the performance of a system with lower bandwidth. Furthermore, data and model parallelism reveal significantly distinct training profiles on different system bandwidths at scale and, thus, need a thorough study. In this paper, we provide a bottom-up breakdown of training throughput into compute and communication time, and quantitatively analyze their respective influences on overall end-to-end training scaling. Our evaluation involves an in-depth exploration of data parallelism, scaling up to 512 GPUs with limited bandwidth, and examines three model sharding strategies among six model sizes. We also evaluate three combinations of model parallelism on both high and low bandwidth supercomputing systems. Overall, our work provides a broader perspective on large-scale transformer model training, and our analysis and evaluation yield practical insights for predicting training scaling, shaping the future development of supercomputing system design.

High-Performance Computing Training by Skills for Advanced Computing Ecosystems

High-Performance Computing (HPC) is one of the pillars of developing modern science and disruptive technologies, uniting computer architectures and parallel programming into multidisciplinary interactions to face domain-specific problems. That is why different areas of knowledge require their future professionals (scientists or not) to acquire skills in using HPC. The Super Computing and Distributed Systems Camping School, SC-Camp, is a non-profit activity that proposes a series of courses about HPC with an important focus on practical sessions (more than half of the time) addressed to undergraduate and graduate students who could benefit from HPC by demand. It is an itinerant school, bringing the HPC knowledge to a different place every year, focusing on diversity, sustainability, and humanity.

Stress, strain, neutron transport and radiation effects in a full fusion tokamak device: A virtual MAST-U study

A finite-element method (FEM) model for the Mega-Ampere Spherical Tokamak - Upgrade (MAST-U) fusion tokamak has been developed to evaluate stress and deformations in the full device structure and to assess the stability of the whole tokamak with respect to its simulated exposure to an artificial level of neutron irradiation. Here, we use MAST-U as a proxy for a fusion power plant to explore the level of fidelity made possible by modern supercomputing systems. Gravity and atmospheric pressure were used to test the high-resolution FEM model, involving in excess of 122 million elements. Taking the MASTU fusion plasma as a neutron source, we perform full-scale neutron transport calculations to quantify spatial variations in the neutron flux and assess the neutron radiation exposure across the structure. This is a first step towards applying recently developed multiscale computational tools to evaluate the spectrum of stress in the tokamak, identifying the location of stress concentrations as well as their magnitude. This study provides an example of full fusion device neutronics and FEM simulations which are enabling UKAEA to define computational requirements for modelling a whole fusion power plant as well as for specifying operating conditions for the relevant materials.

GPU Implementation of the Improved CEEMDAN Algorithm for Fast and Efficient EEG Time-Frequency Analysis.

Time-frequency analysis of EEG data is a key step in exploring the internal activities of the human brain. Studying oscillations is an important part of the analysis, as they are thought to provide the underlying mechanism for communication between neural assemblies. Traditional methods of analysis, such as Short-Time FFT and Wavelet Transforms, are not ideal for this task due to the time-frequency uncertainty principle and their reliance on predefined basis functions. Empirical Mode Decomposition and its variants are more suited to this task as they are able to extract the instantaneous frequency and phase information but are too time consuming for practical use. Our aim was to design and develop a massively parallel and performance-optimized GPU implementation of the Improved Complete Ensemble EMD with the Adaptive Noise (CEEMDAN) algorithm that significantly reduces the computational time (from hours to seconds) of such analysis. The resulting GPU program, which is publicly available, was validated against a MATLAB reference implementation and reached over a 260× speedup for actual EEG measurement data, and provided predicted speedups in the range of 3000-8300× for longer measurements when sufficient memory was available. The significance of our research is that this implementation can enable researchers to perform EMD-based EEG analysis routinely, even for high-density EEG measurements. The program is suitable for execution on desktop, cloud, and supercomputer systems and can be the starting point for future large-scale multi-GPU implementations.

Application of multivariate time-series model for high performance computing (HPC) fault prediction.

Aiming at the high reliability demand of increasingly large and complex supercomputing systems, this paper proposes a multidimensional fusion CBA-net (CNN-BiLSTAM-Attention) fault prediction model based on HDBSCAN clustering preprocessing classification data, which can effectively extract and learn the spatial and temporal features in the predecessor fault log. The model can effectively extract and learn the spatial and temporal features from the predecessor fault logs, and has the advantages of high sensitivity to time series features and sufficient extraction of local features, etc. The RMSE of the model for fault occurrence time prediction is 0.031, and the prediction accuracy of node location for fault occurrence is 93% on average, as demonstrated by experiments. The model can achieve fast convergence and improve the fine-grained and accurate fault prediction of large supercomputers.

An Adaptive Generalized Leaky Integrate-and-Fire Model for Hippocampal CA1 Pyramidal Neurons and Interneurons

Full-scale morphologically and biophysically realistic model networks, aiming at modeling multiple brain areas, provide an invaluable tool to make significant scientific advances from in-silico experiments on cognitive functions to digital twin implementations. Due to the current technical limitations of supercomputer systems in terms of computational power and memory requirements, these networks must be implemented using (at least) simplified neurons. A class of models which achieve a reasonable compromise between accuracy and computational efficiency is given by generalized leaky integrate-and fire models complemented by suitable initial and update conditions. However, we found that these models cannot reproduce the complex and highly variable firing dynamics exhibited by neurons in several brain regions, such as the hippocampus. In this work, we propose an adaptive generalized leaky integrate-and-fire model for hippocampal CA1 neurons and interneurons, in which the nonlinear nature of the firing dynamics is successfully reproduced by linear ordinary differential equations equipped with nonlinear and more realistic initial and update conditions after each spike event, which strictly depends on the external stimulation current. A mathematical analysis of the equilibria stability as well as the monotonicity properties of the analytical solution for the membrane potential allowed (i) to determine general constraints on model parameters, reducing the computational cost of an optimization procedure based on spike times in response to a set of constant currents injections; (ii) to identify additional constraints to quantitatively reproduce and predict the experimental traces from 85 neurons and interneurons in response to any stimulation protocol using constant and piecewise constant current injections. Finally, this approach allows to easily implement a procedure to create infinite copies of neurons with mathematically controlled firing properties, statistically indistinguishable from experiments, to better reproduce the full range and variability of the firing scenarios observed in a real network.

Three-dimensional inversion of corona structure and simulation of solar wind parameters based on the photospheric magnetic field deduced from the Global Oscillation Network Group

In this research, the Potential Field Source Surface–Wang–Sheeley–Arge (PFSS–WSA) solar wind model is used. This model consists of the Potential Field Source Surface (PFSS) coronal magnetic field extrapolation module and the Wang–Sheeley–Arge (WSA) solar wind velocity module. PFSS is implemented by the POT3D package deployed on Tianhe 1A supercomputer system. In order to obtain the three–dimensional (3D) distribution of the coronal magnetic field at different source surface radii (Rss), the model utilizes the Global Oscillation Network Group (GONG) photospheric magnetic field profiles for two Carrington rotations (CRs), CR2069 (in 2008) and CR2217 (in 2019), as the input data, with the source surface at Rss = 2Rs, Rss = 2.5Rs and Rss = 3Rs, respectively. Then the solar wind velocity, the coronal magnetic field expansion factor, and the minimum angular distance of the open magnetic field lines from the coronal hole boundary are estimated within the WSA module. The simulated solar wind speed is compared with the value for the corona extrapolated from the data observed near 1 AU, through the calculations of the mean square error (MSE), root mean square error (RMSE) and correlation coefficient (CC). Here we extrapolate the solar wind velocity at 1 AU back to the source surface via the Parker spiral. By comparing the evaluation metrics of the three source surface heights, we concluded that the solar source surface should be properly decreased with respect to Rss = 2.5Rs during the low solar activity phase of solar cycle 23.

Challenges in High-Performance Computing

High-Performance Computing, HPC, has become one of the most active computer science fields. Driven mainly by the need for high processing capabilities required by algorithms from many areas, such as Big Data, Artificial Intelligence, Data Science, and subjects related to chemistry, physics, and biology, the state-of-art algorithms from these fields are notoriously demanding computer resources. Therefore, choosing the right computer system to optimize their performance is paramount. This article presents the main challenges of future supercomputer systems, highlighting the areas that demand the most of HPC servers; the new architectures, including heterogeneous processors composed of artificial intelligence chips, quantum processors, the adoption of HPC on cloud servers; and the challenges of software developers when facing parallelizing applications. We also discuss challenges regarding non-functional requirements, such as energy consumption and resilience.