JournalTOCs

The Journal of Supercomputing
Journal Prestige (SJR): 0.407

Citation Impact (citeScore): 2
Number of Followers: 1

Hybrid journal (It can contain Open Access articles)
ISSN (Print) 1573-0484 - ISSN (Online) 0920-8542
Published by Springer-Verlag

[2468 journals]

Addressing the class imbalance problem in network intrusion detection
systems using data resampling and deep learning
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: Network intrusion detection systems (NIDS) are the most common tool used to detect malicious attacks on a network. They help prevent the ever-increasing different attacks and provide better security for the network. NIDS are classified into signature-based and anomaly-based detection. The most common type of NIDS is the anomaly-based NIDS which is based on machine learning models and is able to detect attacks with high accuracy. However, in recent years, NIDS has achieved even better results in detecting already known and novel attacks with the adoption of deep learning models. Benchmark datasets in intrusion detection try to simulate real-network traffic by including more normal traffic samples than the attack samples. This causes the training data to be imbalanced and causes difficulties in detecting certain types of attacks for the NIDS. In this paper, a data resampling technique is proposed based on Adaptive Synthetic (ADASYN) and Tomek Links algorithms in combination with different deep learning models to mitigate the class imbalance problem. The proposed model is evaluated on the benchmark NSL-KDD dataset using accuracy, precision, recall and F-score metrics. The experimental results show that in binary classification, the proposed method improves the performance of the NIDS and outperforms state-of-the-art models with an achieved accuracy of 99.8%. In multi-class classification, the results were also improved, outperforming state-of-the-art models with an achieved accuracy of 99.98%.
  PubDate: 2023-07-01
A prefetch control strategy based on improved hill-climbing method in
asymmetric multi-core architecture
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: Cache prefetching is a traditional way to reduce memory access latency. In multi-core systems, aggressive prefetching may harm the system. In the past, prefetching throttling strategies usually set thresholds through certain factors. When the threshold is exceeded, prefetch throttling strategies will control the aggressive prefetcher. However, these strategies usually work well in homogeneous multi-core systems and do not work well in heterogeneous multi-core systems. This paper considers the performance difference between cores under the asymmetric multi-core architecture. Through the improved hill-climbing method, the aggressiveness of prefetching for different cores is controlled, and the IPC of the core is improved. Through experiments, it is found that compared with the previous strategy, the average performance of big core is improved by more than 3%, and the average performance of little cores is improved by more than 24%.
  PubDate: 2023-07-01
Improved YOLOv5 for real-time traffic signs recognition in bad weather
conditions
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: One of significant tasks in autonomous vehicle technology is traffic signs recognizing. It helps to avoid traffic violations on the road. However, recognition of traffic signs becomes more complicated in bad weather such as lack of light, rain, fog. Those bad weather conditions cause low accuracy of detecting and recognizing. In this paper, we aim to build a model to recognize and classify the traffic signs in different bad weather conditions by applying deep learning technique. Weather data are collected from variety types as well as generated from different techniques. Collected data are trained on the YOLOv5s, YOLOv7 model. In order to increase the accuracy, those YOLOv5s are improved on different models by replacing Squeeze-and-Excitation (SE) attention module or Global Context(GC) block. On the test set: the accuracy of YOLOv5s is 76.8%, the accuracy of YOLOv7 is 78% the accuracy of YOLOv5s+SE attention module is 78.4% and the accuracy of YOLOv5s+C3GC is 79.2%. The results show that YOLOv5s+C3GC model significantly improves the accuracy in recognition of blurred-distant-objects.
  PubDate: 2023-07-01
Mapping the national HPC ecosystem and training needs: The Greek paradigm
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: HPC is a key tool for processing and analyzing the constantly growing volume of data, from 64.2 zettabytes in 2020 to an expected 180 zettabytes in 2025 (1 zettabyte is equal to 1 trillion gigabytes). As such, HPC has a large number of application areas that range from climate change, monitoring and mitigating planning to the production of safer and greener vehicles and treating COVID-19 pandemic to the advancement of knowledge in almost every scientific field and industrial domain. The current work presents an HPC Training Mapping Framework and the relevant findings and processed data of an online Training Needs Analysis (TNA) survey. The latter was used to map the training demands and gaps of existing skills and future ones. The participants consist of academia and industry and the data were utilized to find the profile of HPC user alongside the best training practices that are in need. It is found that in Greece during the year 2021, the stakeholder segment with the highest number of respondents was from academia and research with a total of 74%. The vast majority appear to have basic information accounting for 37% of the respondents. In terms of familiarity, users with intermediate familiarity with HPC represented 21% of respondents, followed by non-familiar users that accounted in total for 16.1. Advanced and highly advanced user segments account only for 8.6% and 7.4% accordingly. Overall, it is found that a: (1) fast-pace, (2) entry level, (3) applied HPC training but (4) not focused only on HPC, that will (5) provide some kind of certification, by the Greek HPC ecosystem.
  PubDate: 2023-07-01
SAGE: toward on-the-fly gradient compression ratio scaling
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: Gradient sparsification is widely adopted in distributed training; however, it suffers from a trade-off between computation and communication. The prevalent Top-k sparsifier has a hard constraint on computational overhead while achieving the desired gradient compression ratio. Conversely, the hard-threshold sparsifier eliminates computational constraints but fail to achieve the targeted compression ratio. Motivated by this tradeoff, we designed a novel threshold-based sparsifier called SAGE, which achieves a compression ratio close to that of the Top-k sparsifier with negligible computational overhead. SAGE scales the compression ratio by deriving an adjustable threshold based on each iteration’s heuristics. Experimental results show that SAGE achieves a compression ratio closer to the desired ratio than a hard-threshold sparsifier without exacerbating the accuracy of model training. In terms of computation time for gradient selection, SAGE achieves a speedup of up to \(23.62\times\) over the Top-k sparsifier.
  PubDate: 2023-07-01
Comparing the performance of multi-layer perceptron training on electrical
and optical network-on-chips
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: Multi-layer perceptron (MLP) is a class of Artificial Neural Networks widely used in regression, classification, and prediction. To accelerate the training of MLP, more cores can be used for parallel computing on many-core systems. However, with the increasing number of cores integrated into the chip, the communication bottleneck in the training of MLP on electrical network-on-chip (ENoC) becomes severe, degrading MLP training performance. Replacing ENoC with optical network-on-chip (ONoC) can break the communication bottleneck in MLP training. To facilitate the development of ONoC for MLP training, it is necessary to compare and model the MLP training performance of ONoC and ENoC in advance. This paper first analyzes and compares the differences between ONoC and ENoC. Then, we formulate the performance and energy model of MLP training on ONoC and ENoC by analyzing the communication and computation time, static energy, and dynamic energy consumption, respectively. Furthermore, we conduct extensive simulations to compare their MLP training performance and energy consumption with our simulation infrastructure. The experimental results show the MLP training time of ONoC has been reduced by 65.16% and 52.51% on average in different numbers of cores and batch sizes compared with ENoC. The results also exhibit that ONoC overall has 54.86% and 43.13% on average energy reduction in different numbers of cores and batch sizes compared with ENoC. However, with a small number of cores (e.g., less than 50) in MLP training, ENoC consumes less energy than ONoC. These experiments confirm that generally ONoC is a good replacement for ENoC when using a large number of cores in terms of performance and energy consumption for MLP training.
  PubDate: 2023-07-01
FtCFt: a fault-tolerant coverage preserving strategy for face
topology-based wireless sensor networks
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: Although researchers have investigated multiple facets of fault tolerance, majority of them have overlooked fault tolerance in face structured WSNs. Motivated by this, we propose a Fault-Tolerant Coverage Preserving Strategy for Face Topology-based WSNs (FtCFt). Unlike existing methods of recovering failures by merging the adjacent faces, we propose a coverage-aware node replacement method to replace the failing node with a suitable alternate node. This is significant because a mobile target will go undetected, and no evidence of it can be acquired until it leaves the hole region and is sensed by a node. FtCFt offers fault tolerance by incorporating node self-check and link-check strategies that works in conjunction with one of its mobile target tracking applications. Unlike existing works, the proposed restoration algorithm effectively repairs and restores the face structure to ensure network coverage and connectivity. Simulation results reveal that FtCFt improves coverage, quality of service and WSN liferime.
  PubDate: 2023-07-01
Symmetric key-based authentication and key agreement scheme resistant
against semi-trusted third party for fog and dew computing
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: Fog and dew computing represent relatively new computing paradigms in the literature. The main idea is to offload the computation processes from the device to a more nearby fog or dew server, who further forwards it to the central server. In the case of dew computing, the dew server is considered to lose connection with the central server and should be able to function autonomously most of the time. In the literature, several public-key-based tripartite schemes, offering a full set of security features, have been proposed that can serve the purpose. However, due to the large difference in performance between symmetric and public key-based cryptographic algorithms, this paper proposes a symmetric key-based authentication and key agreement protocol, consisting of a long and a short authentication process, addressing both fog and dew computing scenarios. Moreover, we conduct the informal and formal (ROR logic, GNY logic, and Scyther tool) security analysis to ensure that the scheme satisfies the most important security features described in the literature, in addition to offering protection against a semi-trusted third party. Furthermore, we assess the performance of the long and the short authentication phases in terms of computational, communication, storage costs, and energy consumption, revealing that it is less expensive than its competitors. Additionally, we show that when compared to its competitors, the long and the short authentication phases have less overhead when unknown attacks occur. We also use the NS2 network simulator tool to execute a real-time implementation of the long and the short authentication phase to ensure that it is realistic in practical implementation.
  PubDate: 2023-07-01
Shipping code towards data in an inter-region serverless environment to
leverage latency
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: Serverless computing emerges as a new standard to build cloud applications, where developers write compact functions that respond to events in the cloud infrastructure. Several cloud service industries started adopting serverless for deploying their applications. But one key limitation in serverless computing is that it disregards the significance of data. In the age of big data, when applications run around a huge volume, to transfer data from the data side to the computation side to co-allocate the data and code, leads to high latency. All existing serverless architectures are based on the data shipping architecture. In this paper, we present an inter-region code shipping architecture for serverless, that enables the code to flow from computation side to the data side where the size of the code is negligible compared to the data size. We tested our proposed architecture over a real-time cloud platform Amazon Web Services with the integration of the Fission serverless tool. The evaluation of the proposed code shipping architecture shows for a data file size of 64 MB, the latency in the proposed code shipping architecture is 8.36 ms and in existing data shipped architecture is found to be 16.8 ms. Hence, the proposed architecture achieves a speedup of 2x on the round latency for high data sizes in a serverless environment. We define round latency to be the duration to read and write back the data in the storage.
  PubDate: 2023-07-01
Efficient and portable Winograd convolutions for multi-core processors
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: We take a step forward towards developing high-performance codes for the convolution operator, based on the Winograd algorithm, that are easy to customise for general-purpose processor architectures. In our approach, augmenting the portability of the solution is achieved via the introduction of vector instructions from Intel SSE/AVX2/AVX512 and ARM NEON/SVE to exploit the single-instruction multiple-data capabilities of current processors as well as OpenMP pragmas to exploit multi-threaded parallelism. While this comes at the cost of sacrificing a fraction of the computational performance, our experimental results on three distinct processors, with Intel Xeon Skylake, ARM Cortex A57 and Fujitsu A64FX processors, show that the impact is affordable and still renders a Winograd-based solution that is competitive when compared with the lowering gemm-based convolution.
  PubDate: 2023-07-01
Synergistic integration between internet of things and augmented reality
technologies for deaf persons in e-learning platform
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: The development of the Internet of Things (IoT) accentuates the interweaving of the digital with the physical, raising multiple issues, including for learning and more specifically for the learning of the deaf. While emerging technologies may have given rise to various methods and modes of learning, the implications of the IoT for deaf learning are not yet understood. This study aims to propose the possible integration of the IoT and augmented reality (AR) to support deaf learning from three dimensions of analysis: data, interfaces and pervasiveness. The axes of applications identified suggest that the IoT promotes learning characterized by experimentation, adaptation (to the context and to the learner), the manipulation of objects and exploration without the constraints of time or space. In order to achieve this integration, preliminary survey is collected from deaf learners. Based on the results, a deaf intelligent learning model is proposed and discussed.
  PubDate: 2023-07-01
An automatic model management system and its implementation for AIOps on
microservice platforms
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: With the gradual expansion of microservice architecture-based applications, the complexity of system operation and maintenance is also growing significantly. With the advent of AIOps, it is now possible to automatically detect the state of the system, allocate resources, warn, and detect anomalies using machine learning models. Given the dynamic nature of online workloads, the running state of a microservice system in production is constantly in flux. Therefore, it is necessary to continuously train, encapsulate, and deploy models based on the current system status for the AIOps model to dynamically adapt to the system environment. This paper proposes a model update and management pipeline framework for AIOps models in microservices systems in order to accomplish the aforementioned objectives and simplify the process. In addition, a prototype system based on Kubernetes and Gitlab is designed to provide preliminary framework implementation and validation. The system consists of three components: model training, model packaging, and model deploying. Parallelization and parameter search are incorporated into the model training procedure in order to facilitate rapid training of multiple models and automated model hyperparameter tuning. We automate the packaging and deployment process using technology for continuous integration. Experiments are conducted to validate the prototype system, and the results demonstrate the feasibility of the proposed framework. This work serves as a useful resource for constructing an integrated and streamlined AIOps model management system.
  PubDate: 2023-07-01
Trust factor-based analysis of user behavior using sequential pattern
mining for detecting intrusive transactions in databases
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: Organizations today are employing databases on a large scale to store data essential for their functioning. Malicious access and modifications of the databases may lead to adverse financial and legal implications. In recent years, security researchers have focused on detecting abuse of access privileges by employees of an organization. Identifying threats from insiders is hard because they are aware of the organization of the database in addition to having authorised access privileges. To detect insider attacks effectively and efficiently, we present a novel approach to dynamically determine the malicious transactions using historical data. We propose Trust factor-based user behavior analysis using sequential pattern mining for database intrusion detection systems (TFUBID). Since, groups of users access the organizational database for similar purposes, we cluster user behavior vectors using fuzzy clustering and define a class of Integral Data Attributes using sequential pattern mining to model trust factor-based behavioral patterns of employees accessing the database assigning higher weight to critical elements and Directly Correlated Attributes. A comprehensive experimental evaluation on our synthetic dataset adhering to TPC-C standard benchmark revealed that TFUBID achieved an accuracy of 94% for detecting malicious transactions and outperforms competing state-of-the-art techniques on several performance measures.
  PubDate: 2023-07-01
Embedding hierarchical folded cubes into linear arrays and complete binary
trees with minimum wirelength
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: Graph embedding maps a guest graph into a host graph, thus enabling structural simulation, processor allocation, and algorithm porting. It is used to design the physical layout of Network-on-Chip (NoC) and to study the simulation capabilities of a parallel architecture. Wirelength is one of the indicators to measure the quality of graph embedding. Minimum wirelength in NoC design means a smaller wiring area and less wiring cost. In parallel computing, it means shorter communication time and delay. In this paper, the guest graph is the hierarchical folded cube with good communication and fault tolerance capabilities. The host graphs are the linear array and the complete binary tree, both of which are widely used in graph embeddings. We solve the embedding problems in linear time for hierarchical folded cubes into linear arrays and complete binary trees with minimum wirelength, respectively.
  PubDate: 2023-07-01
Estimation and prediction of the multiply exponentially decaying daily
case fatality rate of COVID-19
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: The spread of the COVID-19 disease has had significant social and economic impacts all over the world. Numerous measures such as school closures, social distancing, and travel restrictions were implemented during the COVID-19 pandemic outbreak. Currently, as we move into the post-COVID-19 world, we must be prepared for another pandemic outbreak in the future. Having experienced the COVID-19 pandemic, it is imperative to ascertain the conclusion of the pandemic to return to normalcy and plan for the future. One of the beneficial features for deciding the termination of the pandemic disease is the small value of the case fatality rate (CFR) of coronavirus disease 2019 (COVID-19). There is a tendency of gradually decreasing CFR after several increases in CFR during the COVID-19 pandemic outbreak. However, it is difficult to capture the time-dependent CFR of a pandemic outbreak using a single exponential coefficient because it contains multiple exponential decays, i.e., fast and slow decays. Therefore, in this study, we develop a mathematical model for estimating and predicting the multiply exponentially decaying CFRs of the COVID-19 pandemic in different nations: the Republic of Korea, the USA, Japan, and the UK. We perform numerical experiments to validate the proposed method with COVID-19 data from the above-mentioned four nations.
  PubDate: 2023-07-01
An improved symbiotic organisms search algorithm with good point set and
memory mechanism
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: Symbiotic organisms search (SOS) algorithm is a current popular stochastic optimization algorithm. It has been widely used to handle all kinds of optimization problems, whereas SOS has some disadvantages, such as over-exploration phenomenon and unbalance between exploration and exploitation. To improve the search capability of SOS, in this study, a novel improved SOS (GMSOS) with good point set and memory mechanism is presented. For enhancing the population diversity and the optimization ability of SOS algorithm, good point set instead of uniform distribution is utilized to produce the initial population, and memory mechanism is employed in three stages of SOS algorithm. In the mutualism stage and commensalism stage, history best organism in memory takes the place of the current best organism. In the parasitism stage, the new parasite vector based on history best organism is produced. These strategies help to effectively provide a better trade-off between exploration and exploitation in the search scope, and avoiding falling into local optima synchronously. The performance of the presented SOS is evaluated on 35 typical benchmark functions and 3 engineering design problems. The experimental results attest that the proposed algorithm is competitive as compared to other algorithms considered.
  PubDate: 2023-07-01
A transformer with layer-cross decoding for remaining useful life
prediction
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: Remaining useful life (RUL) prediction is critical for industrial equipment status detection, and the accurate prediction results provide decision-makers with actionable information. Preventive maintenance can be carried out to prevent the sudden failure of the equipment effectively based on the predicted results. However, as Industry 4.0 technologies develop, the amount of data collected by sensors is also rapidly increasing. The existing RUL prediction methods are gradually unable to cope with complex industrial equipment data, and deep learning methods gradually come to the fore. In this background, this paper proposes a transformer-based model with a multi-layer encoder–decoder structure to extract domain-invariant features. The decoder in the traditional transformer structure only obtains a single piece of information from the last layer of the encoder, and this paper uses an integrated layer-cross decoding strategy to compensate. Based on the encoder–decoder cross-connection, each decoder layer is provided with global view information from the final encoder layer simultaneously, improving the model’s performance. The validity and superiorities of the proposed method are evaluated through several experiments on the publicly available C-MAPSS dataset provided by NASA. As can be seen from the results, the proposed method gets higher prediction accuracy than other network architectures and state-of-the-art approaches.
  PubDate: 2023-07-01
An improved discrete flower pollination algorithm for fuzzy QoS-aware IoT
services composition based on skyline operator
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: The Quality of Service (QoS)-aware Service Composition (QSC) for Internet of Things (IoT) consists of connecting different available atomic IoT-services to produce a Composite of IoT-Services (CS) which satisfies the requirements of end users. With the growing number of atomic IoT-services that have similar functionalities with different values in their QoS parameters, it has been a challenging issue to select the suitable ones in order to generate an optimal CS with high quality in terms of QoS values which should fulfill the end users’ constraints. This problem, which is an NP-hard constrained optimization one, has been generally solved under the assumption of precise and deterministic QoS values, which is not fully advisable. Since the QoS values of an IoT-service are doomed to be altered at any point, due to changes in topological structure of IoT networks, mobility of IoT devices, IoT systems congestion, and economic policies. Hence, the ambiguity of the QoS parameters is represented using the generalized trapezoidal fuzzy number (GTrFN). Moreover, a novel efficient approach combining two modules (1) a fuzzy skyline-based module and (2) an improved discrete flower pollination algorithm is proposed to solve the QSC in Fuzzy IoT environments (QSCFIoT). The performance and the efficiency of the proposal are validated on different scales of QSCFIoT using fuzzy versions of the real QWS and a large-sized synthetic datasets; while the experimental results demonstrate that the proposed approach is superior to some recently proposed QSC optimization algorithms such as EFPA, PSO and ITL-QCA in terms of composition’s quality, time, and stability
  PubDate: 2023-07-01
Parallel implementations of randomized vector algorithm for solving large
systems of linear equations
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: The results of a parallel implementation of a randomized vector algorithm for solving systems of linear equations are presented in the paper. The solution is represented in the form of a Neumann series. The stochastic method computes this series by sampling only random columns, avoiding multiplication of matrix by matrix and matrix by vector. We consider the case when the matrix is too large to fit in random-access memory (RAM). We use two approaches to solve this problem. In the first approach, the matrix is divided into parts that are distributed among MPI processes and stored in the available RAM of the cluster nodes. In the second approach, the entire matrix is stored on each node’s hard drive, loaded into RAM, and processed in parts. Independent Monte Carlo experiments for random column indices are distributed among MPI processes or OpenMP threads for both approaches to matrix storage. The efficiency of parallel implementations is analyzed. Results are given for a system governed by dense matrices of size \(10^4\) and \(10^5\) .
  PubDate: 2023-07-01
swParaFEM: a highly efficient parallel finite element solver on Sunway
many-core architecture
- Free pre-print version: Loading...
  Rate this result: What is this?
  
  Please help us test our new pre-print finding feature by giving the pre-print link a rating.
  A 5 star rating indicates the linked pre-print has the exact same content as the published article.
  
  Abstract: The simulation of three-dimensional stress and strain is a research hot spot of computational structural mechanics. As the complexity of the project increasing, the size of the matrix generated increases during the simulation. Therefore, a fast and efficient solver is needed. In this paper, we present swParaFEM, a highly efficient parallel finite element solver on Sunway many-core architecture. It is based on preconditioned conjugate gradient iteration algorithm. We launch a master–slave acceleration model to exploit the computational power of Sunway supercomputer. The kernel aggregation optimization scheme is proposed to deal with the problem that threads’ frequent creation and destruction waste computing resources. Moreover, we improve the data transfer speed from the slave core to the master core through memory access optimization. Using several optimizations, we achieve a speedup of 10.5 \(\times\) compared to the naive implementation on one compute group of an SW26010-Pro processor and a strong scaling efficiency of 62.8% on 512 compute groups.
  PubDate: 2023-07-01