High-performance Parallel Computing Research Articles

Leveraging Scratchpad Memory in a Hierarchical Architecture for Multicore

This paper proposes a novel architecture for multi-core processors, tailored for high-performance parallel computing. The architecture is founded on the innovative notion that complex problems can be decomposed into three relatively independent sub-problems: data processing, data management, and data communication. It features a grid of small, programmable processing units intricately connected to their three neighbouring units, forming a physically scalable and fractal environment. With flexibility, modularity, and scalability as focal points, this architecture aims to address the anticipated real-time signal processing demands in future telecommunication and multimedia systems. One notable aspect of the proposed architecture is its direct support for object-oriented features at the hardware level. Adopting a hybrid approach with Scratchpad Memory (SPM) combined with Cache in the on-chip memory hierarchy enhances performance and adaptability for sophisticated multi-core applications. The study highlights SPM management and introduces a dynamic data management framework. Unlike traditional SPM allocation methods relying on compiler or profiling knowledge, the proposed approach leverages random sampling and probability theory to predict hot-access data during runtime. This dynamic memory access pattern guides SPM allocation, ensuring optimal utilization of SPM's advantages in access speed and energy consumption, complemented by hardware support from DataUnit. This paper presents a paradigm shift in multi-core processor architecture, offering advanced features to meet the evolving requirements of parallel computing, making it well-suited for future telecommunication and multimedia systems. This paper introduces a revolutionary multicore processor architecture designed for high-performance parallel computing. The foundation of the proposed architecture lies in the innovative concept that complex problems can be decomposed into three relatively independent sub-problems: data processing, data management, and data communication. The architecture features a grid of small, programmable processing units, each intricately connected to its three neighboring units, forming a physically scalable and fractal environment. With a focus on flexibility, modularity, and scalability, this architecture addresses the real-time signal processing demands anticipated in future telecommunication and multimedia systems. One distinctive aspect of the proposed architecture is its direct support for object-oriented features at the hardware level. The on-chip memory hierarchy adopts a hybrid approach with Scratchpad Memory (SPM) combined with Cache, enhancing performance and adaptability for sophisticated multi-core applications. The study emphasizes SPM management and introduces a dynamic data management framework. In contrast to traditional SPM allocation methods relying on compiler or profiling knowledge, the proposed approach leverages random sampling and probability theory to predict hot-access data during runtime. This dynamic memory access pattern guides SPM allocation, ensuring optimal utilization of SPM's advantages in access speed and energy consumption, complemented by hardware support from DataUnit. This paper presents a paradigm shift in multicore processor architecture, offering advanced features to meet the evolving requirements of parallel computing, making it well-suited for future telecommunication and multimedia systems

An efficient framework for matrix-free SpMV computation on GPU for elastoplastic problems

High computational cost in elastoplastic analysis is often handled by the use of high performance parallel computers. However, the presence of both elastic and plastic states leads to the branching issue, which prevents the realization of true parallel performance. The computational efficiency of an elastoplastic analysis using finite element method is largely determined by the performance of repeated solution of linearized system of equations. In this paper, we propose GPU-based matrix-free strategies to compute sparse matrix–vector multiplication (SpMV) in Conjugate Gradient (CG) iterative solver for the acceleration of solution of linear system of equations. Matrix-free solvers never assemble large sparse global tangent matrix and perform the computation directly with small dense elemental matrices, reducing the storage requirement and preventing the use of problematic sparse storage formats. A uniform treatment of elements in elastic and plastic regions is achieved by the proposed single kernel strategy, which prevents branching, avoids redundant computation and provides efficient memory access. In addition, we propose node-based and degrees-of-freedom (DOF)-based parallel strategies for effective implementation of matrix-free SpMV on a GPU. The proposed strategies use single elemental tangent matrix for all elements in elastic region and individual tangent matrices in plastic region. The computational experiments over three large-scale benchmark examples of elastoplasticity reveal that the performance of the node-based and DOF-based parallelization strategies depends on the amount of plasticity in the body. At low plasticity levels, node-based strategy performs best, achieving 3.2× speedup over an existing GPU-based matrix-free SpMV strategy in the literature. For moderate to high amount of plasticity, the DOF-based strategy outperforms every other strategy and obtains speedups of up to 3.5× over the existing SpMV strategy.

High-definition simulation of packed-bed liquid chromatography

Numerical simulations of chromatography are conventionally performed using reduced-order models that homogenize aspects of flow and transport in the radial and angular dimensions. This enables much faster simulations at the expense of lumping the effects of inhomogeneities into a column dispersion coefficient, which requires calibration via empirical correlations or experimental results. We present a high-definition model with spatially resolved geometry. A stabilized space–time finite element method is used to solve the model on massively parallel high-performance computers. We simulate packings with up to 10,000 particles. The impact of particle size distribution on velocity and concentration profiles as well as breakthrough curves is studied. Our high-definition simulations provide unique insight into the process. The high-definition data can also be used as a source of ground truth to identify and calibrate appropriate reduced-order models that can then be applied for process design and optimization.

Research progress of computer vision tasks based on deep learning and SAE network

Abstract In recent years, artificial intelligence has gradually become the core driving force of a new round of scientific and technological revolution and industrial transformation, and is exerting a profound impact on all aspects of human life. With the rapid development of Internet big data and high-performance parallel computing, relevant research in computer vision has made significant progress in the past few years, becoming one of the important application branches in the field of artificial intelligence. The exercise of image classification forming part of computer vision tasks involves a large amount of computation, and training based on traditional deep learning (DL) classification models typically involves slow training and low accuracy in many parameters. Thus, in order to solve these problems, an image classification model based on DL and SAE network was proposed. Firstly, the main research of computer vision task-image classification is introduced in detail. Then, the combination framework of deep neural network and SAE network is built. At the same time, the deep neural network was used to carry out convolution operation of the parameters learned by SAE and extract each feature of the image with neurons, so as to improve the training accuracy of the deep neural network. Finally, the traditional deep neural network and SAE network were used for comparative experiment and analysis. Experimental results show that the proposed method has a certain degree of improvement in image classification accuracy compared with traditional deep neural network and SAE network, and the accuracy reaches 97.13%.

Box dimension of the border of Kingdom of Saudi Arabia

Fractal dimension unlike topological dimension is (usually) a non-integer number which measures complexity, roughness, or irregularity of an object with respect to the space in which the set lies. It is used to characterize highly irregular objects in nature containing statistical self-similarity such as mountains, snowflakes, clouds, coastlines, borders etc.In this article, box dimension (a version of fractal dimension) of the border of Kingdom of Saudi Arabia (KSA) is computed using a multicore parallel processing algorithm based on the classical box-counting method. A power law relation is obtained from numerical simulations which relates the length of the border with the scale size and provides a very close estimate of the actual length of the KSA border within the scaling regions and scaling effects on the length of KSA border are considered. The algorithm presented in the article is shown to be highly scalable and efficient and the speedup of the algorithm is computed using Amdahl's and Gustafson's laws. For simulations, a high performance parallel computer is employed using Python codes and QGIS software.

Large-scale simulation of high-intensity focused ultrasound with Sierra/SD

High-frequency simulations of acoustic wave propagation are known to be computationally challenging, and the difficulty is compounded for large domains. For problems like high-intensity focused ultrasound (HIFU) with coupling between piezoelectric and acoustic media, both challenges are present. The larger domain sizes, higher-order elements, and greater levels of refinement necessary in such simulations result in millions of degrees of freedom, large system matrices, and substantial memory requirements, raising the need for parallel high-performance computing (HPC). Sierra/SD is a massively parallel HPC application developed for finite element method simulations in structural dynamics and acoustics. In this work, Sierra/SD is used to simulate an acoustic pulse from a piezoelectric transducer focused on an elastic scatterer in a fluid medium. Three-dimensional simulation results are presented for the acoustic field in the fluid and the stress field in the scatterer, and performance is compared between Sierra/SD and COMSOL Multiphysics for smaller geometries. Finally, to showcase the expanded analysis possibilities afforded by HPC for HIFU, an example is presented using a support vector machine to determine a decision boundary for maximum stress in the scatterer. [SNL is managed and operated by NTESS under DOE NNSA contract DE-NA0003525.]

Simulating the Impact of Dynamic Rerouting on Metropolitan-scale Traffic Systems

The rapid introduction of mobile navigation aides that use real-time road network information to suggest alternate routes to drivers is making it more difficult for researchers and government transportation agencies to understand and predict the dynamics of congested transportation systems. Computer simulation is a key capability for these organizations to analyze hypothetical scenarios; however, the complexity of transportation systems makes it challenging for them to simulate very large geographical regions, such as multi-city metropolitan areas. In this article, we describe enhancements to the Mobiliti parallel traffic simulator to model dynamic rerouting behavior with the addition of vehicle controller actors and vehicle-to-controller reroute requests. The simulator is designed to support distributed-memory parallel execution using discrete event simulation and be scalable on high-performance computing platforms. We demonstrate the potential of the simulator by analyzing the impact of varying the population penetration rate of dynamic rerouting on the San Francisco Bay Area road network. Using high-performance parallel computing, we can simulate a day in the San Francisco Bay Area with 19 million vehicle trips with 50 percent dynamic rerouting penetration over a road network with 0.5 million nodes and 1 million links in less than three minutes. We present a sensitivity study on the dynamic rerouting parameters, discuss the simulator’s parallel scalability, and analyze system-level impacts of changing the dynamic rerouting penetration. Furthermore, we examine the varying effects on different functional classes and geographical regions and present a validation of the simulation results compared to real-world data.

SERGHEI (SERGHEI-SWE) v1.0: a performance-portable high-performance parallel-computing shallow-water solver for hydrology and environmental hydraulics

Abstract. The Simulation EnviRonment for Geomorphology, Hydrodynamics, and Ecohydrology in Integrated form (SERGHEI) is a multi-dimensional, multi-domain, and multi-physics model framework for environmental and landscape simulation, designed with an outlook towards Earth system modelling. At the core of SERGHEI's innovation is its performance-portable high-performance parallel-computing (HPC) implementation, built from scratch on the Kokkos portability layer, allowing SERGHEI to be deployed, in a performance-portable fashion, in graphics processing unit (GPU)-based heterogeneous systems. In this work, we explore combinations of MPI and Kokkos using OpenMP and CUDA backends. In this contribution, we introduce the SERGHEI model framework and present with detail its first operational module for solving shallow-water equations (SERGHEI-SWE) and its HPC implementation. This module is designed to be applicable to hydrological and environmental problems including flooding and runoff generation, with an outlook towards Earth system modelling. Its applicability is demonstrated by testing several well-known benchmarks and large-scale problems, for which SERGHEI-SWE achieves excellent results for the different types of shallow-water problems. Finally, SERGHEI-SWE scalability and performance portability is demonstrated and evaluated on several TOP500 HPC systems, with very good scaling in the range of over 20 000 CPUs and up to 256 state-of-the art GPUs.

Сандық әдістер есептерін шешуде параллель есептеулер қолдануды оқу процесіне ендірудің теориялық негіздері

Currently, the development of high-performance parallel computing is one of the modern requirements. There are many unsolved problems in the world. Many of these tasks require highperformance calculations. Therefore, in the field of education, there is a need to use parallel computing in solving problems of numerical methods. This requires professionals who can use parallel computing. This article makes a theoretical analysis of the learning conditions using parallel computing in solving problems of numerical methods in universities around the world and the Republic. The necessity of using parallel calculations in solving problems of numerical methods in higher educational institutions of the Republic of Kazakhstan is substantiated.

Optimal siting of rainwater harvesting systems for reducing combined sewer overflows at city scale

The installation of green infrastructure (GI) is an effective approach to manage urban stormwater and combined sewer overflow (CSO) by restoring pre-development conditions in urban areas. Research on simulation-optimization techniques to aid with GI planning decision-making is expanding. However, due to high computational expense, the simulation-optimization methods are often based on design storm events, and it is unclear how much different rainfall scenarios (i.e., design storm events vs. long-term historical rainfall data) impact the optimal siting of GI. The Parallel Pareto Archived Dynamically Dimensioned Search (ParaPADDS) algorithm in a novel simulation-optimization tool OSTRICH-SWMM was used to leverage distributed computing resources. A case study was conducted to optimally site rainwater harvesting cisterns within 897 potential subcatchments throughout the City of Buffalo, New York. Seven design storm events with different return periods and rainfall durations and a one-month historical rainfall time series were considered. The results showed that the optimal solutions of siting cisterns using event-based scenarios, though less computationally expensive, may not perform well under continuous rainfall scenarios, suggesting design rainfall scenarios should be carefully considered for optimizing GI planning. The impact of rainfall scenarios was particularly significant in the middle region of the Pareto front of multi-objective optimization. Utilizing high-performance parallel computing, OSTRICH-SWMM is a promising tool to optimize GI at large spatial and temporal scales.

Run-Time Remapping Algorithm of Dataflow Actors on NoC-Based Heterogeneous MPSoCs

Multiprocessor system-on-chip (MPSoC) platforms have been emerging as the main solution to cope with processor frequency ceiling and power density issues while still improving performances. Then, network-on-chip (NoC) has been adopted to provide the increasing number of processors with the required communication bandwidth as well as with the necessary flexibility. Video processing and streaming applications are adopting dynamic dataflow model of computation as the need for high performance parallel computing is growing. Dataflow applications executed on modern MPSoC-based architectures are becoming increasingly dynamic and more data-dependent. Different tasks execute concurrently with significant modifications in their workloads and resource demanding over time depending on the input data. Hence, adopting any static or offline dynamic scheduling for mapping tasks will not cope with the computation variations. This article introduces an original run-time mapping algorithm based on the Move Based (MB) method targeting a dedicated heterogeneous NoC-based MPSoC architecture to achieve workload balancing and optimized communication traffic. The performance of the proposed algorithm is verified by conducting cycle-accurate SystemC simulations of the adopted NoC implementing a real MPEG4-SP decoder. The obtained results reveal the effectiveness of our proposed algorithm. For various real-life videos, the proposed algorithm systematically succeeded to enhance significantly the performance.

Development and Design of ADPSS Co-simulation Technology for Large-Scale New Energy Electromagnetic Transient Simulation in High Performance Calculation

As large-scale wind and photovoltaic power generation incorporating into the AC power grid, the impact of new energy fluctuation and uncertainty on the transient stability of power grid have attracted widespread attention among researchers and engineers in the power industry. It is necessary to improve the transient simulation and modelling ability of new energy to provide a more efficient simulation platform for the stability research of new energy generation connecting to large power grid. In this paper, the ADPSS technology of large-scale new energy electromagnetic transient simulation is mainly studied in depth, including the following contents: the design and development of the interface and program of ADPSS-Matlab co-simulation, the improvement and optimization of the parallel high performance computing, and the verification of the accuracy of co-simulation for the modelling and experimental research with large-scale new energy simulation.

High-Performance and Parallel Computing Techniques Review: Applications, Challenges and Potentials to Support Net-Zero Transition of Future Grids

The transition towards net-zero emissions is inevitable for humanity’s future. Of all the sectors, electrical energy systems emit the most emissions. This urgently requires the witnessed accelerating technological landscape to transition towards an emission-free smart grid. It involves massive integration of intermittent wind and solar-powered resources into future power grids. Additionally, new paradigms such as large-scale integration of distributed resources into the grid, proliferation of Internet of Things (IoT) technologies, and electrification of different sectors are envisioned as essential enablers for a net-zero future. However, these changes will lead to unprecedented size, complexity and data of the planning and operation problems of future grids. It is thus important to discuss and consider High Performance Computing (HPC), parallel computing, and cloud computing prospects in any future electrical energy studies. This article recounts the dawn of parallel computation in power system studies, providing a thorough history and paradigm background for the reader, leading to the most impactful recent contributions. The reviews are split into Central Processing Unit (CPU) based, Graphical Processing Unit (GPU) based, and Cloud-based studies and smart grid applications. The state-of-the-art is also discussed, highlighting the issue of standardization and the future of the field. The reviewed papers are predominantly focused on classical imperishable electrical system problems. This indicates the need for further research on parallel and HPC approaches applied to future smarter grid challenges, particularly to the integration of renewable energy into the smart grid.

Visualization and modeling of forest fire propagation in Patagonia

Fire propagation is a big concern all over the world. Visualization is a valuable tool to test possible different scenarios for fire spread, specially for designing strategies for fire control, mitigation and management. We present a parallel High-Performance Computing (HPC) forest fire simulator with an interactive and intuitive user interface that offers several functionalities to the user. The visualization interface allows to choose the propagation model of preference, the scenario of interest, as well as numerous simulation features including firebreaks and ignition points. We show some of the outputs for two different mathematical models for fire spreading. The simulator was developed with an open source philosophy in the framework of Faster Than Real Time (FTRT) applications thinking on its possible use in the field during a forest fire propagation. It can be run in Linux (Ubuntu) and Windows Operating Systems and for portability purposes the simulator was also implemented on a NVIDIA Jetson Nano.

(Retracted) Support vector machine parallelized remote sensing image classification algorithm based on big data

With the development of big data technology, machine learning classification methods have been widely used in the classification and recognition of remote sensing images. For remote sensing big data, how to quickly and efficiently use machine learning classification algorithms to classify remote sensing images is an urgent problem. It is a general term for the theory, method, technology, and activities of obtaining valuable information based on massive remote sensing data sets, synthesizing auxiliary data from other sources, and using big data thinking and means. The purpose of this paper is to study the support vector machine (SVM) parallelized remote sensing image classification algorithm based on big data. We propose a parallel nesting of GPU in MPI multiprocesses based on the big data framework, which can more effectively improve the calculation processing speed and build a high-performance SVM parallel computing framework based on the big data framework. The optimization problem of SVM considers both empirical risk and structural risk minimization and requires maximum edge distance when constructing hyperplane decision boundaries, so there is ample space between the interval boundaries to accommodate the test samples. Based on this framework, we improve the machine learning SVM algorithm and realize the high-performance parallel computing of SVM classification algorithm on this platform. It is an efficient hybrid parallel mode to nest GPU in MPI multiprocess in parallel. When the number of nodes is 2, 4, and 6, the speedup of the SVM classification algorithm is 1.52, 2.24, and 2.55.

A Two-Stage Multi-Objective Genetic Algorithm for a Flexible Job Shop Scheduling Problem with Lot Streaming

The work in this paper is motivated by a recently published article in which the authors developed an efficient two-stage genetic algorithm for a comprehensive model of a flexible job-shop scheduling problem (FJSP). In this paper, we extend the application of the algorithm to solve a lot streaming problem in FJSP while at the same time expanding the model to incorporate multiple objectives. The objective function terms included in our current work are the minimization of the (1) makespan, (2) maximum sublot flowtime, (3) total sublot flow time, (4) maximum job flowtime, (5) total job flow time, (6) maximum sublot finish-time separation, (7) total sublot finish-time separation, (8) maximum machine load, (9) total machine load, and (10) maximum machine load difference. Numerical examples are presented to illustrate the greater need for multi-objective optimization in larger problems, the interaction of the various objective function terms, and their relevance in providing better solution quality. The ability of the two-stage genetic algorithm to jointly optimize all the objective function terms is also investigated. The results show that the algorithm can generate initial solutions that are highly improved in all of the objective function terms. It also outperforms the regular genetic algorithm in convergence speed and final solution quality in solving the multi-objective FJSP lot streaming. We also demonstrate that high-performance parallel computation can further improve the performance of the two-stage genetic algorithm. Nevertheless, the sequential two-stage genetic algorithm with a single CPU outperforms the parallel regular genetic algorithm that uses many CPUs, asserting the superiority of the two-stage genetic algorithm in solving the proposed multi-objective FJSP lot streaming.

A GPU-accelerated light-field super-resolution framework based on mixed noise model and weighted regularization

Light-field (LF) super-resolution (SR) plays an essential role in alleviating the current technology challenge in the acquisition of a 4D LF, which assembles both high-density angular and spatial information. Due to the algorithm complexity and data-intensive property of LF images, LFSR demands a significant computational effort and results in a long CPU processing time. This paper presents a GPU-accelerated computational framework for reconstructing high-resolution (HR) LF images under a mixed Gaussian-Impulse noise condition. The main focus is on developing a high-performance approach considering processing speed and reconstruction quality. From a statistical perspective, we derive a joint ell ^1-ell ^2 data fidelity term for penalizing the HR reconstruction error taking into account the mixed noise situation. For regularization, we employ the weighted non-local total variation approach, which allows us to effectively realize LF image prior through a proper weighting scheme. We show that the alternating direction method of the multipliers algorithm (ADMM) can be used to simplify the computation complexity and results in a high-performance parallel computation on the GPU Platform. An extensive experiment is conducted on both synthetic 4D LF dataset and natural image dataset to validate the proposed SR model’s robustness and evaluate the accelerated optimizer’s performance. The experimental results show that our approach achieves better reconstruction quality under severe mixed-noise conditions as compared to the state-of-the-art approaches. In addition, the proposed approach overcomes the limitation of the previous work in handling large-scale SR tasks. While fitting within a single off-the-shelf GPU, the proposed accelerator provides an average speedup of 2.46{times } and 1.57{times } for {times }2 and {times }3 SR tasks, respectively. In addition, a speedup of 77{times } is achieved as compared to CPU execution.

Multiscale numerical tool for studying nonlinear dynamics in solids induced by strong laser pulses.

Strong-field phenomena in solids exhibit extreme high-order nonlinear optical effects, which have triggered many theoretical and experimental investigations. However, there is still a lack of highly efficient numerical tools to simulate the relevant phenomena. In this paper, a versatile multiscale numerical tool set is developed for studying high-order nonlinear optical effects in solids, generated by ultrafast strong laser pulses. This tool is based on the tight-binding model approximation of the crystal structure, the related parameters of which are obtained from the density functional theory calculations. And the nonlinear effects are explored by solving the Maxwell equationscoupled with the semiconductor Bloch equations. Our numerical tool can provide not only basic electronic structures and optical responses of the crystal, but also the real-time evolution of the macroscopic electromagnetic fields and the current density. The high-performance parallel computing and the interpolation method in our tool make it possible to study the strong-field nonlinear responses and propagation effects on a large spatial and temporal scale. Finally, three theoretical or experimental results published recently are satisfactorily reproduced, showing a good performance of the current toolbox.

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.

Enhancing scalability of a matrix-free eigensolver for studying many-body localization

We propose several techniques to enhance the parallel scalability of a matrix-free eigensolver designed for studying many-body localization (MBL) of quantum spin chain models with nearest-neighbor interactions and on-site disorder. This type of problem is computationally challenging because the dimension of the associated Hamiltonian matrix grows exponentially with respect to the number of spins L, and we need to average over different realizations of the random disorder to obtain relevant statistical behavior. For each disorder realization, we need to compute eigenvalues from different regions of the spectrum and their corresponding eigenvectors. In previous work, the interior eigenstates for a single eigenvalue problem are computed via the shift-and-invert Lanczos algorithm. Due to the extremely high memory footprint of the LU factorizations, this technique is not well suited for large L’s. For example, we need thousands of compute nodes on modern high performance computing infrastructures to go beyond L = 24. The matrix-free approach does not suffer from this memory bottleneck, however, its scalability is limited by a computation and communication load imbalance. To reduce this imbalance and to significantly enhance the scalability of the matrix-free eigensolver, we reorder the matrix and leverage the consistent space runtime, CSPACER. We also show its efficiency in managing irregular communication patterns at scale compared to optimized MPI non-blocking two-sided and one-sided RMA implementation variants. This effort enables us to study MBL for spin chains with a larger number of spins. The efficiency and effectiveness of the proposed algorithm is demonstrated by computing eigenstates on a massively parallel many-core high performance computer.