A number of techniques have been proposed over the years to detect clones for improving software maintenance, reusability or security. However, there is still a lack of language agnostic approaches with code granulari...
详细信息
A number of techniques have been proposed over the years to detect clones for improving software maintenance, reusability or security. However, there is still a lack of language agnostic approaches with code granularity flexibility for near-miss clone detection in big code in scale. It is challenging to detect near-miss clones in big code across large scale source repositories with hundreds of millions of lines of code (MLOC) or more. The main reason is that it requires more computing and memory resources as the scale of the source code increases. In particular, near-miss clone detection is more difficult and need more resources. In this paper, we present SNCD, a fast and scalable distributed clone detection approach. It overcomes single node CPU and memory resource limitation with MapReduce and HDFS by scalable distributed parallelization. Furthermore, it is partial index based and optimized with multi-threading strategy which further improve the efficiency. It can not only detect Type-1 and Type-2 clones but can also discover the most computationally expensive Type-3 clones for large repositories. Meanwhile, it works for both function and file granularities, and it supports many different programming languages. Experimental results show that SNCD scales better for big code with the size of code in terms of lines of code increases compared to existing clone detection techniques, with recall and precision comparable to state-of-art approaches. With BigCloneBench and the Mutation Framework, two recent and widely used benchmarks, SNCD achieves both high recall and precision, which is competitive with other existing tools.
Parallel simulation of incompressible fluid flows is considered on networks of homogeneous workstations. Coarse-grain parallelization of a Taylor-Galerkin/pressure-correction finite element algorithm are discussed, ta...
详细信息
Parallel simulation of incompressible fluid flows is considered on networks of homogeneous workstations. Coarse-grain parallelization of a Taylor-Galerkin/pressure-correction finite element algorithm are discussed, taking into account network communication costs. The main issues include the parallelization of system assembly, and iterative and direct solvers, that are of common interest to finite element and general numerical computation. The parallelization strategies are implemented on a Sun workstation cluster using the Parallel Virtual Machine (PVM) message passing library. Test results are obtained with a maximum of nineteen workstations and various PVM configurations are exhibited. Parallel efficiency close to ideal has been achieved for some strategies adopted. It is suggested that load balancing may not always be beneficial on distributed platforms with broadcasting communication connection. (C) 1998 John Wiley & Sons, Ltd.
I-PixelHop is a lightweight learning framework based on successive subspace learning. Compared with the deep learning methods, the rolling bearing fault diagnosis (RBFD) model based on I-PixelHop has lower computation...
详细信息
I-PixelHop is a lightweight learning framework based on successive subspace learning. Compared with the deep learning methods, the rolling bearing fault diagnosis (RBFD) model based on I-PixelHop has lower computational complexity and smaller model size, and it can obtain a good diagnosis accuracy in the complex working conditions. However, the training and diagnosis time of the RBFD model based on I-PixelHop are still long in the face of large-scale RB fault datasets. Therefore, an efficient I-PixelHop framework based on Spark-GPU for intelligent fault diagnosis is proposed. First, a Spark-GPU-based distributed parallel I-PixelHop framework is developed, which can efficiently perform distributed parallel training and diagnosis. Second, an asynchronous parallel execution strategy based on superscalar pipeline (APES-SP) is proposed, which can reduce the waiting time of each functional unit of the distributed parallel I-PixelHop framework. Finally, an ensemble classifier based on Bagging is designed and parallelized, which can improve the diagnosis accuracy of the RBFD model based on distributed parallel I-PixelHop framework. Extensive experiments demonstrate that the proposed framework can not only maintain high diagnosis accuracy but also significantly improve the training performance and diagnosis performance of the RBFD model based on I-PixelHop under industrial big data. [GRAPHICS] .
This paper presents the recent developments in hierarchical genetic algorithms (HGAs) to speed up the optimization of aerodynamic shapes. It first introduces HGAs, a particular instance of parallel GAs based on the no...
详细信息
This paper presents the recent developments in hierarchical genetic algorithms (HGAs) to speed up the optimization of aerodynamic shapes. It first introduces HGAs, a particular instance of parallel GAs based on the notion of interconnected sub-populations evolving independently. Previous studies have shown the advantages of introducing a multi-layered hierarchical topology in parallel GAs. Such a topology allows the use of multiple models for optimization problems, and shows that it is possible to mix fast low-fidelity models for exploration and expensive high-fidelity models for exploitation. Finally, a new class of multi-objective optimizers mixing HGAs and Nash Game Theory is defined. These methods are tested for solving design optimization problems in aerodynamics. A parallel version of this approach running a cluster of PCs demonstrate the convergence speed up on an inverse nozzle problem and a high-lift problem for a multiple element airfoil. (C) 2002 Published by Elsevier Science B.V.
The paper presents parallelization of the boundary element method in distributed memory of a cluster equipped with many-core based compute nodes. A method for efficient distribution of boundary element matrices among ...
详细信息
ISBN:
(纸本)9783319780245;9783319780238
The paper presents parallelization of the boundary element method in distributed memory of a cluster equipped with many-core based compute nodes. A method for efficient distribution of boundary element matrices among MPI processes based on the cyclic graph decompositions is described. In addition, we focus on the intra-node optimization of the code, which is necessary in order to fully utilize the many-core processors with wide SIMD registers. Numerical experiments carried out on a cluster consisting of the Intel Xeon Phi processors of the Knights Landing generation are presented.
Large-scale simulations of blood flow that resolve the 3D deformation of each comprising cell are increasingly popular owing to algorithmic developments in conjunction with advances in compute capability. Among differ...
详细信息
Large-scale simulations of blood flow that resolve the 3D deformation of each comprising cell are increasingly popular owing to algorithmic developments in conjunction with advances in compute capability. Among different approaches for modeling cell-resolved hemodynamics, fluid structure interaction (FSI) algorithms based on the immersed boundary method are frequently employed for coupling separate solvers for the background fluid and the cells within one framework. GPUs can accelerate these simulations;however, both current pre-exascale and future exascale CPU-GPU heterogeneous systems face communication challenges critical to performance and scalability. We describe, to our knowledge, the largest distributed GPU-accelerated FSI simulations of high hematocrit cell-resolved flows with over 17 million red blood cells. We compare scaling on a fat node system with six GPUs per node and on a system with a single GPU per node. Through comparison between the CPU- and GPU-based implementations, we identify the costs of data movement in multiscale multi-grid FSI simulations on heterogeneous systems and show it to be the greatest performance bottleneck on the GPU. (C) 2020 Elsevier B.V. All rights reserved.
暂无评论