High-performance Supercomputers Research Articles

Quantum-centric supercomputing for materials science: A perspective on challenges and future directions

Computational models are an essential tool for the design, characterization, and discovery of novel materials. Computationally hard tasks in materials science stretch the limits of existing high-performance supercomputing centers, consuming much of their resources for simulation, analysis, and data processing. Quantum computing, on the other hand, is an emerging technology with the potential to accelerate many of the computational tasks needed for materials science. In order to do that, the quantum technology must interact with conventional high-performance computing in several ways: approximate results validation, identification of hard problems, and synergies in quantum-centric supercomputing. In this paper, we provide a perspective on how quantum-centric supercomputing can help address critical computational problems in materials science, the challenges to face in order to solve representative use cases, and new suggested directions.

Making Path Selection Bright: A Routing Algorithm for On-Chip Benes Networks

Making Path Selection Bright: A Routing Algorithm for On-Chip Benes Networks

OpenDroneMap: Multi-Platform Performance Analysis

This paper analyzes the performance of the open-source OpenDroneMap image processing software (ODM) across multiple platforms. We tested desktop and laptop computers as well as high-performance cloud computing and supercomputers. Multiple machine configurations (CPU cores and memory) were used. We used eBee S.O.D.A. drone image datasets from Namibia and northern Finland. For testing, we used the OpenDroneMap command line tool with default settings and the fast orthophoto option, which produced a good quality orthomosaic. We also used the “rerun-all option” to ensure that all jobs started from the same point. Our results show that ODM processing time is dependent upon the number of images, a high number of which can lead to high memory demands, with low memory leading to an excessively long processing time. Adding additional CPU cores is beneficial to ODM up to a certain limit. A 20-core machine seems optimal for a dataset of about 1000 images, although 10 cores will result only in slightly longer processing times. We did not find any indication of improvement when processing larger datasets using 40-core machines. For 1000 images, 64 GB memory seems to be sufficient, but for larger datasets of about 8000 images, higher memory of up to 256 GB is required for efficient processing. ODM can use GPU acceleration, at least in some processing stages, reducing processing time. In comparison to commercial software, ODM seems to be slower, but the created orthomosaics are of equal quality.

Insights from high-fidelity modeling of industrial rotary bell atomization

The global automotive industry sprayed over 2.6 billion liters of paint in 2018, much of which through electrostatic rotary bell atomization, a highly complex process involving the fluid mechanics of rapidly rotating thin films tearing apart into micrometer-thin filaments and droplets. Coating operations account for 65% of the energy usage in a typical automotive assembly plant, representing 10,000s of gigawatt-hours each year in the United States alone. Optimization of these processes would allow for improved robustness, reduced material waste, increased throughput, and significantly reduced energy usage. Here, we introduce a high-fidelity mathematical and algorithmic framework to analyze rotary bell atomization dynamics at industrially relevant conditions. Our approach couples laboratory experiment with the development of robust non-Newtonian fluid models; devises high-order accurate numerical methods to compute the coupled bell, paint, and gas dynamics; and efficiently exploits high-performance supercomputing architectures. These advances have yielded insight into key dynamics, including i) parametric trends in film, sheeting, and filament characteristics as a function of fluid rheology, delivery rates, and bell speed; ii) the impact of nonuniform film thicknesses on atomization performance; and iii) an understanding of spray composition via primary and secondary atomization. These findings result in coating design principles that are poised to improve energy- and cost-efficiency in a wide array of industrial and manufacturing settings.

Artificial Intelligence and Computational Issues in Engineering Applications

High-performance supercomputers and emerging computing clusters created in research and development centres are rapidly increasing available computing power, which scientists are eager to use to implement increasingly advanced computing methods [...].

Machine Learning in Antibacterial Drug Design.

Advances in computer hardware and the availability of high-performance supercomputing platforms and parallel computing, along with artificial intelligence methods are successfully complementing traditional approaches in medicinal chemistry. In particular, machine learning is gaining importance with the growth of the available data collections. One of the critical areas where this methodology can be successfully applied is in the development of new antibacterial agents. The latter is essential because of the high attrition rates in new drug discovery, both in industry and in academic research programs. Scientific involvement in this area is even more urgent as antibacterial drug resistance becomes a public health concern worldwide and pushes us increasingly into the post-antibiotic era. In this review, we focus on the latest machine learning approaches used in the discovery of new antibacterial agents and targets, covering both small molecules and antibacterial peptides. For the benefit of the reader, we summarize all applied machine learning approaches and available databases useful for the design of new antibacterial agents and address the current shortcomings.

Automatic exhaustive calculations of large material space by Korringa-Kohn-Rostoker coherent potential approximation method applied to equiatomic quaternary high entropy alloys

Automatic exhaustive exploration of a large material space by high-performance supercomputers is crucial for developing new functional materials. We demonstrated the efficiency of high-throughput calculations using the all-electron Korringa-Kohn-Rostoker coherent potential approximation method with the density functional theory for the large material space consisting of quaternary high entropy alloys, which are nonstoichiometric and substitutionally disordered materials. The exhaustive calculations were performed for 147 630 systems based on the AkaiKKR program package and supercomputer Fugaku, where the numerical parameters and self-consistent convergence are automatically controlled. The large material database including the total energies, magnetization, Curie temperature, and residual resistivity was constructed by our calculations. We used frequent itemset mining to identify the characteristics of parcels in magnetization and Curie temperature space. We also identified the elements that enhance the magnetization and Curie temperature and clarified the rough dependence of the elements through regression modeling of the residual resistivity.

NAS-Bench-NLP: Neural Architecture Search Benchmark for Natural Language Processing

Neural Architecture Search (NAS) is a promising and rapidly evolving research area. Training a large number of neural networks requires an exceptional amount of computational power, which makes NAS unreachable for those researchers who have limited or no access to high-performance clusters and supercomputers. A few benchmarks with precomputed neural architectures performances have been recently introduced to overcome this problem and ensure reproducible experiments. However, these benchmarks are only for the computer vision domain and, thus, are built from the image datasets and convolution-derived architectures. In this work, we step outside the computer vision domain by leveraging the language modeling task, which is the core of natural language processing (NLP). Our main contribution is as follows: we have provided search space of recurrent neural networks on the text datasets and trained 14k architectures within it; we have conducted both intrinsic and extrinsic evaluation of the trained models using datasets for semantic relatedness and language understanding evaluation; finally, we have tested several NAS algorithms to demonstrate how the precomputed results can be utilized. We consider that the benchmark will provide more reliable empirical findings in the community and stimulate progress in developing new NAS methods well suited for recurrent architectures.

Accelerating the simulation of annual bifacial illumination of real photovoltaic systems with ray tracing

SummaryAccurate modeling of bifacial illumination is critical to improve the prediction of the energy yield of bifacial solar systems. Monte Carlo ray tracing is the most powerful tool to accomplish this task. In this work, we accelerate Monte Carlo ray tracing of large solar systems by nearly 90%. Our model achieves root-mean-square error values of 7.9% and 37.2% for the front and rear irradiance compared against single-axis tracking field reference data, respectively. The rear irradiance modeling error decreases to 18.9% if suspected snow periods are excluded. Crucially, our full system simulations show that surrounding ground surfaces affect the rear irradiance deep into the system. Therefore, unit system simulations cannot necessarily ignore the influence of the perimeter of large installations to accurately estimate annual yield. Large-scale simulations involving high-performance supercomputing were necessary to investigate these effects accurately, calibrate our simplified models, and validate our results against experimental measurements.

OpenMP application experiences: Porting to accelerated nodes

As recent enhancements to the OpenMP specification become available in its implementations, there is a need to share the results of experimentation in order to better understand the OpenMP implementation’s behavior in practice, to identify pitfalls, and to learn how the implementations can be effectively deployed in scientific codes. We report on experiences gained and practices adopted when using OpenMP to port a variety of ECP applications, mini-apps and libraries based on different computational motifs to accelerator-based leadership-class high-performance supercomputer systems at the United States Department of Energy. Additionally, we identify important challenges and open problems related to the deployment of OpenMP. Through our report of experiences, we find that OpenMP implementations are successful on current supercomputing platforms and that OpenMP is a promising programming model to use for applications to be run on emerging and future platforms with accelerated nodes.

A Review Study on Security Issues, Benefits, Risks and Various Challenges in Cloud Computing Platform and Proposed Model for Enhancing Security for Sensitive Data

In this recent and technological epoch, Cloud Computing(CC) is the generally promising and upcoming trend in the Universe which aims at the high performance supercomputing. It is the on-request openness of computer system assets, in particular storage space and computing supremacy, with no forthright dynamic supervision by the user. The expression is by and large used to illustrate Data Centers (DC) available to many users over the Internet. Nowadays, most of the organizations move into cloud because it helps the end users to store their personal data into the cloud which can be accessed by the user at anytime and anywhere across the globe. The users data is stored and maintained in DC of the cloud providers viz.. Google, Amazon, Microsoft etc. Despite several advantages of using the cloud computing technology, the end users hesitate to utilize the various cloud services and business organizations are unwilling to deploy/set up their businesses in the cloud because of several security threats and issues involved in cloud computing. The main emphasis of our study shall discuss the cloud computing technology , its characteristics, deployment models, service models and the various security issues, challenges related to cloud computing. We shall also present a schematic idea/algorithm which could be used to ensure the security of sensitive data in cloud storage.

High-Performance Supercomputer Technologies of Simulation and Identification of Nanoporous Systems with Feedback for n-Component Competitive Adsorption

High-performance supercomputer computing technologies have been developed to model and identify the parameters of complex n-component competitive adsorption processes in nanoporous cybersystems with feedback. Using the Laplace transform and the Heaviside operating method with the decomposition of a nonlinear system with Langmuir-type adsorption equilibrium conditions, an efficient parallelization of the vector components of the model solution is proposed. The results of numerical experiments based on parallel computations using multicore computers are presented.

Segment Switching: A New Switching Strategy for Optical HPC Networks

Photonics are becoming realistic technologies for implementing interconnection networks in near future Exascale supercomputer systems. Photonics present key features to design high-performance and scalable supercomputer networks, such as higher bandwidth and lower latencies than their electronic supercomputer networks counterparts. Some research work is focused on conventional network topologies built with photonic technologies, with the aim of taking advantage of photonic characteristics. Nevertheless, these approaches fail in that they keep low the network utilization. We looked into this downside and we found that circuit switching was the main performance limitation. In this article we propose a new switching mechanism, called Segment Switching , to address this constraint and improve the network utilization. Segment Switching splits the circuit in segments of the whole path, and uses buffering on selected nodes on the network. Experimental results show that the devised approach significantly outperforms photonic circuit switching in conventional torus and fat tree networks by 70% and 90%, respectively.

Workflows in AiiDA: Engineering a high-throughput, event-based engine for robust and modular computational workflows

Over the last two decades, the field of computational science has seen a dramatic shift towards incorporating high-throughput computation and big-data analysis as fundamental pillars of the scientific discovery process. This has necessitated the development of tools and techniques to deal with the generation, storage and processing of large amounts of data. In this work we present an in-depth look at the workflow engine powering AiiDA, a widely adopted, highly flexible and database-backed informatics infrastructure with an emphasis on data reproducibility. We detail many of the design choices that were made which were informed by several important goals: the ability to scale from running on individual laptops up to high-performance supercomputers, managing jobs with runtimes spanning from fractions of a second to weeks and scaling up to thousands of jobs concurrently, and all this while maximising robustness. In short, AiiDA aims to be a Swiss army knife for high-throughput computational science. As well as the architecture, we outline important API design choices made to give workflow writers a great deal of liberty whilst guiding them towards writing robust and modular workflows, ultimately enabling them to encode their scientific knowledge to the benefit of the wider scientific community.

Why should we care about FAIR?

For a long while, Open Access to publications was the key focus of the Open Science agenda, but suddenly, everyone is talking about data and other research outputs. What are these? Why should we pay attention to them? What’s the difference between FAIR data and Open data, and why should scholars care about any of this? These are the kinds of questions that have been floating around as funders like the European Commission start to mandate the sharing of data, software, and other outputs that support publications. And while the drive towards FAIR sharing is meant to be applied across all research domains and disciplines, the discussion often focuses on ‘big data’ or data from the natural sciences, collected in laboratories or processed on high performance super computers. Where does this leave humanities scholars? What kinds of data and other research outputs are created here? What considerations are particular to the humanities?
 In February 2020, after an open consultation process that received over 200 comments plus a number of multi-page submissions, ALLEA published “Sustainable and FAIR Data Sharing in the Humanities: Recommendations of the ALLEA E-Humanities Working Group. https://doi.org/10.7486/DRI.tq582c863. The report is structured around the research data lifecycle, providing concrete recommendations for sustainable FAIR sharing. This talk will highlight some of the key challenges faced by humanities scholars in this realm, look to where alliances can be built outside of academia, and dive into the dissemination possibilities that are created through the FAIR sharing of humanities research data.

A Parallel Iterative Finite Element Method for the Linear Elliptic Equations

By combining the two-grid discretization with the partition of unity method, a parallel iterative finite element method for the linear elliptic equations is proposed and investigated. Since the construction of the partition of unity is based on the coarse mesh triangulation, the computational domain of each subproblem can be divided automatically and the number of subproblems can be arbitrarily huge as the coarse mesh parameter H tends to zero. That means our method can be easily implemented in high performance supercomputers or cluster of workstations. Theoretical results based on a priori error estimation of the scheme are obtained, which indicate that our method can reach the optimal convergence orders within a few two-grid iterations. Numerical results are reported to assess the theoretical results.

On improving the efficiency of mathematical modeling of the problem of stability of construction

Algorithmic software for mathematical modeling of structural stability is considered, which is reduced to solving a partial generalized eigenvalues problem of sparse matrices, with automatic parallelization of calculations on modern parallel computers with graphics processors. Peculiarities of realization of parallel algorithms for different structures of sparse matrices are presented. The times of solving the problem of stability of composite materialsusing a three-dimensional model of "finite size fibers" on computers of different architectures are given. In mathematical modeling of physical and technical processes in many cases there is a need to solve problems of algebraic problem of eigenvalues (APVZ) with sparse matrices of large volumes. In particular, such problems arise in the analysis of the strength of structures in civil and industrial construction, aircraft construction, electric welding, etc. The solving to these problems is to determine the eigenvalues and eigenvectors of sparse matrices of different structure. The efficiency of solving these problems largely depends on the effectiveness of mathematical modeling of the problem as a whole. Continuous growth of task parameters, calculation of more complete models of objects and processes on computers require an increase in computer productivity. High-performance computing requirements are far ahead of traditional parallel computing, even with multicore processors. High-performance computing requirements are far ahead of traditional parallel computing, even with multicore processors. Today, this problem is solved by using powerful supercomputers of hybrid architecture, such as computers with multicore processors (CPUs) and graphics processors (GPUs), which combine MIMD and SIMD architectures. But the potential of high-performance computers can be used to the fullest only with algorithmic software that takes into account both the properties of the task and the features of the hybrid architecture. Complicating the architecture of modern high-performance supercomputers of hybrid architecture, which are actively used for mathematical modeling (increasing the number of computer processors and cores, different types of computer memory, different programming technologies, etc.) means a significant complication of efficient use of these resources in creating parallel algorithms and programs. here are problems with the creation of algorithmic software with automatic execution of stages of work, which are associated with the efficient use of computing resources, ways to store and process sparse matrices, analysis of the reliability of computer results. This makes it possible to significantly increase the efficiency of mathematical modeling of practical problems on modern high-performance computers, as well as free users from the problems of parallelization of complex problems. he developed algorithmic software automatically implements all stages of parallel computing and processing of sparse matrices on a hybrid computer. It was used at the Institute of Mechanics named after S.P. Tymoshenko NAS of Ukraine in modeling the strength problems of composite material. A significant improvement in the time characteristics of mathematical modeling was obtained. Problems of mathematical modeling of the properties of composite materials has an important role in designing the processes of deformation and destruction of products in various subject areas. Algorithmic software for mathematical modeling of structural stability is considered, which is reduced to solving a partial generalized problem of eigen values of sparse matrices of different structure of large orders, with automatic parallelization of calculations on modern parallel computers with graphics processors. The main methodological principles and features of implementation of parallel algorithms for different structures of sparse matrices are presented, which ensure effective implementation of multilevel parallelism of a hybrid system and reduce data exchange time during the computational process. As an example of these approaches, a hybrid algorithm of the iteration method in subspace for tape and block-diagonal matrices with a frame for computers of hybrid architecture is given. Peculiarities of data decomposition for matrices of profile structure at realization of parallel algorithms are considered. The proposed approach provides automatic determination of the required topology of the hybrid computer and the optimal amount of resources for the organization of an efficient computational process. The results of testing the developed algorithmic software for problems from the collection of the University of Florida, as well as the times of solving the problem of stability of composite materials using a three-dimensional model of "finite size fibers" on computers of different architectures. The results show a significant improvement in the time characteristics of solving problems.

Accelerated training of bootstrap aggregation-based deep information extraction systems from cancer pathology reports.

In machine learning, it is evident that the classification of the task performance increases if bootstrap aggregation (bagging) is applied. However, the bagging of deep neural networks takes tremendous amounts of computational resources and training time. The research question that we aimed to answer in this research is whether we could achieve higher task performance scores and accelerate the training by dividing a problem into sub-problems. The data used in this study consist of free text from electronic cancer pathology reports. We applied bagging and partitioned data training using Multi-Task Convolutional Neural Network (MT-CNN) and Multi-Task Hierarchical Convolutional Attention Network (MT-HCAN) classifiers. We split a big problem into 20 sub-problems, resampled the training cases 2,000 times, and trained the deep learning model for each bootstrap sample and each sub-problem-thus, generating up to 40,000 models. We performed the training of many models concurrently in a high-performance computing environment at Oak Ridge National Laboratory (ORNL). We demonstrated that aggregation of the models improves task performance compared with the single-model approach, which is consistent with other research studies; and we demonstrated that the two proposed partitioned bagging methods achieved higher classification accuracy scores on four tasks. Notably, the improvements were significant for the extraction of cancer histology data, which had more than 500 class labels in the task; these results show that data partition may alleviate the complexity of the task. On the contrary, the methods did not achieve superior scores for the tasks of site and subsite classification. Intrinsically, since data partitioning was based on the primary cancer site, the accuracy depended on the determination of the partitions, which needs further investigation and improvement. Results in this research demonstrate that 1. The data partitioning and bagging strategy achieved higher performance scores. 2. We achieved faster training leveraged by the high-performance Summit supercomputer at ORNL.

High-Performance Supercomputer Technologies of Simulation of Nanoporous Feedback Systems for Adsorption Gas Purification

High-performance methods and computer technologies for modeling of non-isothermal gas adsorption in nanoporous cyber-feedback systems based on the Heaviside operating method and Landau’s linearization approach to Langmuir’s nonlinear adsorption equilibrium using decomposition in a convergent series at the temperature transition point with subsequent efficient parallelization of the model over a small parameter are developed. The results of numerical experiments based on high-speed parallel computations on multicore computers are presented.

Optimizing BCPNN Learning Rule for Memory Access.

Simulation of large scale biologically plausible spiking neural networks, e.g., Bayesian Confidence Propagation Neural Network (BCPNN), usually requires high-performance supercomputers with dedicated accelerators, such as GPUs, FPGAs, or even Application-Specific Integrated Circuits (ASICs). Almost all of these computers are based on the von Neumann architecture that separates storage and computation. In all these solutions, memory access is the dominant cost even for highly customized computation and memory architecture, such as ASICs. In this paper, we propose an optimization technique that can make the BCPNN simulation memory access friendly by avoiding a dual-access pattern. The BCPNN synaptic traces and weights are organized as matrices accessed both row-wise and column-wise. Accessing data stored in DRAM with a dual-access pattern is extremely expensive. A post-synaptic history buffer and an approximation function thus are introduced to eliminate the troublesome column update. The error analysis combining theoretical analysis and experiments suggests that the probability of introducing intolerable errors by such optimization can be bounded to a very small number, which makes it almost negligible. Derivation and validation of such a bound is the core contribution of this paper. Experiments on a GPU platform shows that compared to the previously reported baseline simulation strategy, the proposed optimization technique reduces the storage requirement by 33%, the global memory access demand by more than 27% and DRAM access rate by more than 5%; the latency of updating synaptic traces decreases by roughly 50%. Compared with the other similar optimization technique reported in the literature, our method clearly shows considerably better results. Although the BCPNN is used as the targeted neural network model, the proposed optimization method can be applied to other artificial neural network models based on a Hebbian learning rule.