Scalable Parallel Research Articles

Scalable implicit solvers with dynamic mesh adaptation for a relativistic drift-kinetic Fokker–Planck–Boltzmann model

In this work we consider a relativistic drift-kinetic model for runaway electrons along with a Fokker–Planck operator for small-angle Coulomb collisions, a radiation damping operator, and a secondary knock-on (Boltzmann) collision source. We develop a new scalable fully implicit solver utilizing finite volume and conservative finite difference schemes and dynamic mesh adaptivity. A new data management framework in the PETSc library based on the p4est library is developed to enable simulations with dynamic adaptive mesh refinement (AMR), distributed memory parallelization, and dynamic load balancing of computational work. This framework and the runaway electron solver building on the framework are able to dynamically capture both bulk Maxwellian at the low-energy region and a runaway tail at the high-energy region. To effectively capture features via the AMR algorithm, a new AMR indicator prediction strategy is proposed that is performed alongside the implicit time evolution of the solution. This strategy is complemented by the introduction of computationally cheap feature-based AMR indicators that are analyzed theoretically. Numerical results quantify the advantages of the prediction strategy in better capturing features compared with nonpredictive strategies; and we demonstrate trade-offs regarding computational costs. The robustness with respect to model parameters, algorithmic scalability, and parallel scalability are demonstrated through several benchmark problems including manufactured solutions and solutions of different physics models. We focus on demonstrating the advantages of using implicit time stepping and AMR for runaway electron simulations.

Database Management Systems: An Examination using NoSQL

Solutions that can accomplish infinite scalability, high availability, and vast parallelism are needed to meet the ever-changing demands of modern data management. Achieving excellent performance standards. When managing massive collections of both organized and unorganized data sets that traditional RDBMS are unable to handle, the new breed of applications such as business intelligence, enterprise analytics, customer relationship management, document processing, social networks, Web 2.0, and cloud computing needs to scale horizontally to thousands of nodes. The pace the point at which data is produced by interactive applications used by numerous concurrent users in distributed processing involving a vast number of servers and managing Big Data applications has surpassed the capacity of relational databases, emphasizing the adoption of databases NoSQL. Reducing has been addressed by NoSQL database systems.

Parallel Algorithms for Hierarchical Nucleus Decomposition

Nucleus decompositions have been shown to be a useful tool for finding dense subgraphs. The coreness value of a clique represents its density based on the number of other cliques it is adjacent to. One useful output of nucleus decomposition is to generate a hierarchy among dense subgraphs at different resolutions. However, existing parallel algorithms for nucleus decomposition do not generate this hierarchy, and only compute the coreness values. This paper presents a scalable parallel algorithm for hierarchy construction, with practical optimizations, such as interleaving the coreness computation with hierarchy construction and using a concurrent union-find data structure in an innovative way to generate the hierarchy. We also introduce a parallel approximation algorithm for nucleus decomposition, which achieves much lower span in theory and better performance in practice. We prove strong theoretical bounds on the work and span (parallel time) of our algorithms. On a 30-core machine with two-way hyper-threading, our parallel hierarchy construction algorithm achieves up to a 58.84x speedup over the state-of-the-art sequential hierarchy construction algorithm by Sariyuce et al. and up to a 30.96x self-relative parallel speedup. On the same machine, our approximation algorithm achieves a 3.3x speedup over our exact algorithm, while generating coreness estimates with a multiplicative error of 1.33x on average.

Diagnosis of glaucoma using multi‐scale attention block in convolution neural network and data augmentation techniques

AbstractGlaucoma is defined as an eye disease leading to vision loss due to the optic nerve damage. It is often asymptomatic, thus, timely diagnosis and treatment is crucial. In this article, we propose a novel approach for diagnosing glaucoma using deep neural networks, trained on fundus images. Our proposed approach involves several key steps, including data sampling, pre‐processing, and classification. To address the data imbalance issue, we employ a combination of suitable image augmentation techniques and Multi‐Scale Attention Block (MAS Block) architecture in our deep neural network model. The MAS Block is a specific architecture design for CNNs that allows multiple convolutional filters of various sizes to capture features at several scales in parallel. This will prevent the over‐fitting problem and increases the detection accuracy. Through extensive experiments with the ACRIMA dataset, we demonstrate that our proposed approach achieves high accuracy in diagnosing glaucoma. Notably, we recorded the highest accuracy (97.18%) among previous studies. The results from this study reveal the potential of our approach to improve early detection of glaucoma and offer more effective treatment strategies for doctors and clinicians in the future. Timely diagnosis plays a crucial role in managing glaucoma since it is often asymptomatic. Our proposed method utilizing deep neural networks shows promise in enhancing diagnostic accuracy and aiding healthcare professionals in making informed decisions.

Scalable parallel ultrafast optical random bit generation based on a single chaotic microcomb

Random bit generators are critical for information security, cryptography, stochastic modeling, and simulations. Speed and scalability are key challenges faced by current physical random bit generation. Herein, we propose a massively parallel scheme for ultrafast random bit generation towards rates of order 100 terabit per second based on a single micro-ring resonator. A modulation-instability-driven chaotic comb in a micro-ring resonator enables the simultaneous generation of hundreds of independent and unbiased random bit streams. A proof-of-concept experiment demonstrates that using our method, random bit streams beyond 2 terabit per second can be successfully generated with only 7 comb lines. This bit rate can be easily enhanced by further increasing the number of comb lines used. Our approach provides a chip-scale solution to random bit generation for secure communication and high-performance computation, and offers superhigh speed and large scalability.

Early experiences on the OLCF Frontier system with AthenaPK and Parthenon‐Hydro

SummaryThe Oak Ridge Leadership Computing Facility (OLCF) has been preparing the nation's first exascale system, Frontier, for production and end users. Frontier is based on HPE Cray's new EX architecture and Slingshot interconnect and features 74 cabinets of optimized 3rd Gen AMD EPYC CPUs for HPC and AI and AMD Instinct 250X accelerators. As a part of this preparation, “real‐world” user codes have been selected to help assess the functionality, performance, and usability of the system. This article describes early experiences using the system in collaboration with the Hamburg Observatory for two selected codes, which have since been adopted in the OLCF test harness. Experiences discussed include efforts to resolve performance variability and per‐cycle slowdowns. Results are shown for a performance portable astrophysical magnetohydronamics code, AthenaPK, and a mini‐application stressing the core functionality of a performance portable block‐structured adaptive mesh refinement framework, Parthenon‐Hydro. These results show good scaling characteristics to the full system. At the largest scale, the Parthenon‐Hydro miniapp reaches a total of zone‐cycles/s on 9216 nodes (73,728 logical GPUs) at 92% weak scaling parallel efficiency (starting from a single node using a second‐order, finite‐volume method).

Numerical investigation of flow past a cylinder using cumulant lattice Boltzmann method

This paper presents simulations of flow past a circular cylinder within the subcritical Reynolds number (Re) range from 3900 to 2 × 105, utilizing the parameterized cumulant lattice Boltzmann model. In this study, a three-dimensional characteristic boundary condition for incompressible flow has been integrated into the lattice Boltzmann method at the outflow boundary to minimize spurious reflection. The flow field, wake statistics, hydrodynamic force, and power spectra results of Re = 3900 from the cumulant lattice Boltzmann model are exhaustively compared with the laboratory data and other numerical models. Relative to other numerical models employing turbulence closure, the cumulant lattice Boltzmann simulations demonstrate enhanced agreement with the experimental data even with relatively coarser grid resolution. The resolution-spanning feature for the cumulant lattice Boltzmann model in turbulent flows, without using explicit turbulence model, aligns with the previous benchmark case studies. The stability-preserving regularization process in the present model is analyzed. Results indicate that the influence of the regularization parameter is mitigated with improved grid resolution. A specific regularization parameter for flow around cylinder simulations is recommended. Variations in flow properties and hydrodynamic forces within the subcritical Reynolds number range of 3900 to 2 × 105 are analyzed. The results confirm that the parameterized cumulant lattice Boltzmann model can accurately simulate practical engineering flows, characterized by complex separation and recirculation, within the subcritical range. Moreover, the computational efficiency and parallel scalability are compared with other numerical methods.

PaSTiLa: Scalable Parallel Algorithm for Unsupervised Labeling of Long Time Series

PaSTiLa: Scalable Parallel Algorithm for Unsupervised Labeling of Long Time Series

An improved CFAR algorithm for multiple environmental conditions

The primary objective of radar digital signal processing is to detect and identify targets in complicated situations, such as those involving clutter or several closely positioned targets. The constant false alarm rate (CFAR) method is more effective for target recognition and has better control over the false alarm rate. Many studies have been conducted on the design of CFAR, but previous CFAR algorithms have not been effective in all or most environmental fields and target scenarios. In this study, an algorithm called Censored Mean Clutter Map CFAR (CM-CM CFAR) has been developed and tested for various environmental conditions. When compared to a fixed false alarm rate, the suggested CFAR algorithm’s Monte Carlo simulation results demonstrated a high detection probability in a variety of environments. This work designs a real-time CM-CM CFAR processor using field-programmable gate array (FPGA) technologies. Xilinx ARTIX 7 FPGA technology is used to develop and map a scalable parallel framework. Consequently, the implementation required 23,741 LUTs and 1825 FF. It is verified that the complexity and operating speed of the suggested CFAR processor are extremely appropriate for real-time implementation when compared to the results of the previously proposed FPGA implementation.

Scalable Multi-Hierarchy Embedded Platform for Neural Population Simulations.

Brain-inspired structured neural circuits are the cornerstones of both computational and perceived intelligence. Real-time simulations of large-scale high-dimensional neural populations with complex nonlinearities pose a significant challenge. Taking advantage of distributed computations using embedded multi-cores, we propose an ARM-based scalable multi-hierarchy parallel computing platform (EmPaas) for neural population simulations. EmPaas is constructed using 340 ARM Cortex-M4 microprocessors to achieve high-speed and high-accuracy parallel computing. The tree-two-dimensional grid-like hybrid topology completes the overall construction, reducing communication strain and power consumption. As an instance of embedded computing, the optimized model for a biologically plausible basal ganglia-thalamus (BG-TH) network is deployed into this platform to verify the performance. At an operating frequency of 168 MHz, the BG-TH network consisting of 4000 Izhikevich neurons is simulated in the platform for 3000 ms with a power consumption of 56.565 mW per core and an actual time of 2748.57 ms, which shows the parallel computing approach significantly improves computational efficiency. EmPaas can meet the requirement of real-time performance with the maximum amount of 2000 Izhikevich neurons loaded in each Extended Community Unit (ECUnit), which provides a new practical method for research in large-scale brain network simulation and brain-inspired computing.

Ant colony optimization for parallel test assembly.

Ant colony optimization (ACO) algorithms have previously been used to compile single short scales of psychological constructs. In the present article, we showcase the versatility of the ACO to construct multiple parallel short scales that adhere to several competing and interacting criteria simultaneously. Based on an initial pool of 120 knowledge items, we assembled three 12-item tests that (a) adequately cover the construct at the domain level, (b) follow a unidimensional measurement model, (c) allow reliable and (d) precise measurement of factual knowledge, and (e) are gender-fair. Moreover, we aligned the test characteristic and test information functions of the three tests to establish the equivalence of the tests. We cross-validated the assembled short scales and investigated their association with the full scale and covariates that were not included in the optimization procedure. Finally, we discuss potential extensions to metaheuristic test assembly and the equivalence of parallel knowledge tests in general.

CPU and GPU parallel efficiency of ARM based single board computing cluster for CFD applications

This paper examines the parallel efficiency of an ARM-based single board computing cluster made of 16 Raspberry Pi 4 and 8 Nvidia Jetson Nano Single Board Computer, considering both CPU and GPU parallel implementation of CFD applications. It is found that the parallelization scales up to 16 Raspberry Pi 4 and 8 Nvidia Jetson Nano (maximum available on the current cluster). Moreover, it is shown that regarding the computation time, about 12 SBC are as fast as a classical computing workstation. Finally, it is shown that such cluster are energy efficient considering CFD applications.

A computational framework for pharmaco‐mechanical interactions in arterial walls using parallel monolithic domain decomposition methods

AbstractA computational framework is presented to numerically simulate the effects of antihypertensive drugs, in particular calcium channel blockers, on the mechanical response of arterial walls. A stretch‐dependent smooth muscle model by Uhlmann and Balzani is modified to describe the interaction of pharmacological drugs and the inhibition of smooth muscle activation. The coupled deformation‐diffusion problem is then solved using the finite element software FEDDLib and overlapping Schwarz preconditioners from the Trilinos package FROSch. These preconditioners include highly scalable parallel GDSW (generalized Dryja–Smith–Widlund) and RGDSW (reduced GDSW) preconditioners. Simulation results show the expected increase in the lumen diameter of an idealized artery due to the drug‐induced reduction of smooth muscle contraction, as well as a decrease in the rate of arterial contraction in the presence of calcium channel blockers. Strong and weak parallel scalability of the resulting computational implementation are also analyzed.

Development of a Scalable Parallel Compositional Simulator for Thermo-Hydromechanical Coupling in Fractured Rocks Using an Embedded Discrete Fracture Model

Summary Numerical simulation of thermo-hydromechanical (THM) coupling in practical complex fractured rocks is an essential but challenging issue for the evaluation and optimization of underground energy production. In this study, we present our work on a scalable parallel compositional simulator for THM coupling, which is suitable for massive 3D polygonal fractures. In addition, we also present the improvements, parallel implementation, and optimization of an embedded discrete fracture model (EDFM). A unified cell-centered grid system based on the finite volume method (FVM) is used for all governing equations, and an extended stencil is adopted for mechanical equations to resolve the low-resolution defect of the traditional FVM. The deformation of both matrix rock and fractures is considered. A sequential fully implicit (SFI) method is adopted to solve THM coupling. This simulator is validated against three analytical solution models. Finally, we apply the simulator to two cases including a multilayered shale gas reservoir with massive natural fractures and a fractured geothermal model using CO2 as a working fluid. We also test the performance and parallel scalability on 1,024 CPU cores with up to 50 million matrix gridblocks and 5.5 million fracture gridblocks. The results show that this simulator can efficiently solve the THM coupling problem in practical massive fractures.

Top-ii-vol: Massively Parallel Scalable Meshing for Seismic Risk Assessment of Nuclear Sites

High performance computing is widely used for conducting high-resolution geophysics Finite Element Method (FEM) simulations. These simulations are crucial for risk assessment at nuclear sites and help gain deeper insights into underlying physics. However, generating FEM-compatible meshes from point-cloud data – which often serves as input for mesh generation from geomodelling tools – poses significant challenges due to the lack of associated CAD data and extremely large-scale nature. This article introduces a workflow for automatically converting Digital Elevation Model (DEM) point-cloud data into three-dimensional unstructured volumetric meshes. Leveraging domaindecomposition method via the parallel MPI I/O programming, our meshing algorithm efficiently constructs meshes, accommodating an arbitrary number of CPUs and enabling either single or distributed mesh generation. Demonstrating parallel efficiency and scalability, we produced a 68 billion element tetrahedral mesh from real DEM point-cloud data of the Cadarache region in France in approximately 1 second using over 24000 processing units. Meshes produced by this technique are utilized for seismic hazard FEM simulations, one of which is presented in this study. Scalability tests on massively parallel computers show high scalability, with quasi-linear strong scaling characteristics up to 24000 processing units.

A Scalable Parallel Coevolutionary Algorithm With Overlapping Cooperation for Large-Scale Network-Based Combinatorial Optimization

A Scalable Parallel Coevolutionary Algorithm With Overlapping Cooperation for Large-Scale Network-Based Combinatorial Optimization

Fabrication of one-dimensional nanostructures standing vertically on a substrate through layer-by-layer deposition.

Layer-by-layer deposition has been widely used to prepare heterogeneous nanocomposites with controllable properties, where the controllability of nanocomposite properties can be further enhanced by combining the self-assembly of hard materials, such as nanocrystals, and soft materials, such as polymer matrices. However, the self-assembled structure of nanocrystals in a polymer matrix is limited by thermodynamics. Herein, we introduced the strong interaction between nanocrystals and a substrate to generate a thermodynamically unfavorable one-dimensional nanostructure standing vertically on the substrate. Two different shapes-"wire-like" and "steeple-like" or "antenna-like"-of one-dimensional nanostructures standing vertically on the substrate can be obtained using cube-like nanocrystals as building blocks and carefully controlling their size in each deposited layer. This low-cost and massively parallel scale synthesis method to generate one-dimensional nanostructures standing vertically on a substrate can be used to replace the expensive and time-consuming "lithography" synthesis method. This synthesis method also provides a simple way to design and fabricate one-dimensional nanostructures with desired properties standing vertically on a substrate by controlling nanocrystal properties in each deposited layer.

Parallel scale de-blur net for sharpening video images for remote clinical assessment of hand movements

Parallel scale de-blur net for sharpening video images for remote clinical assessment of hand movements

Two classes of action-stabilizing synergies reflecting spinal and supraspinal circuitry.

We explored force-stabilizing synergies during accurate four-finger constant force production tasks in spaces of finger modes (commands to fingers computed to account for the finger interdependence) and of motor unit (MU) firing frequencies. The main specific hypothesis was that the multifinger synergies would disappear during unintentional force drifts without visual feedback on the force magnitude, whereas MU-based synergies would be robust to such drifts. Healthy participants performed four-finger accurate cyclical force production trials followed by trials of constant force production. Individual MUs were identified in the flexor digitorum superficialis (FDS) and extensor digitorum communis (EDC). Principal component analysis was applied to motor unit frequencies to identify robust MU groups (MU-modes) with parallel scaling of the firing frequencies in FDS, in EDC, and the combined MUs of FDS + EDC. The framework of the uncontrolled manifold hypothesis was used to quantify force-stabilizing synergies when visual feedback on the force magnitude was available and 15 s after turning the visual feedback off. Removing visual feedback led to a force drift toward lower magnitudes, accompanied by the disappearance of multifinger synergies. In contrast, MU-mode synergies were minimally affected by removing visual feedback off and continued to be robust for the FDS and for the EDC, while being absent for the (FDS + EDC) analysis. We interpret the findings within the theory of hierarchical control of action with spatial referent coordinates. The qualitatively different behavior of the multifinger and MU-mode-based synergies likely reflects the difference in the involved neural circuitry, supraspinal for the former and spinal for the latter.NEW & NOTEWORTHY Two types of synergies, in the space of commands to individual fingers and in the space of motor unit groups, show qualitatively different behaviors during accurate multifinger force-production tasks. After removing visual feedback, finger force synergies disappear, whereas motor unit-based synergies persist. These results point at different neural circuitry involved in these two basic classes of synergies: supraspinal for multieffector synergies, and spinal for motor unit-based synergies.

A scalable parallel computing method for autonomous platoons

This paper developed a scalable parallel computing method that can be used for platoon simulations and controller validations. A scalable adaptive platooning control law was firstly designed, which accommodates a variety of vehicle-to-vehicle communication topologies. A road vehicle dynamics model that considered the Magic Formula tyre model and suspension dynamics was then derived and validated. The parallel computing method adopted the Message Passing Interface technique to allow fast and scalable simulations. Platoon length changes do not require controller and algorithm changes. An 11-vehicle platoon on a real-world 10 km long road section was simulated. Different localisation sensor errors, communication delays, heterogenous vehicle masses and driving modes were considered. Results show that localisation errors have negligible influences on space errors. Aggressive driving and heterogeneous vehicle masses slightly increase space errors (increases less than 0.23 m). Communication delays are the greatest influencer for space errors. Increases for 15, 45 and 75 ms delays were 0.43, 1.41 and 2.41 m, respectively. It is further shown that parallel computing can improve the computing speed by three times on personal computers and seven to 12 times on workstations.