Dawning Nebulae is a heterogeneous system composed of 9280 multi-core x86 CPUs and 4640 NVIDIA Fermi GPUs. With a Linpack performance of 1.271 petaFLOPS, it was ranked the second in the TOP500 List released in June 20...
详细信息
Dawning Nebulae is a heterogeneous system composed of 9280 multi-core x86 CPUs and 4640 NVIDIA Fermi GPUs. With a Linpack performance of 1.271 petaFLOPS, it was ranked the second in the TOP500 List released in June 2010. In this paper, key issues in the system design of Dawning Nebulae are introduced. system tuning methodologies aiming at petaFLOPS Linpack result are presented, including algorithmic optimization and communication improvement. The design of its file I/O subsystem, including HVFS and the underlying DCFS3, is also described. Performance evaluations show that the Linpack efficiency of each node reaches 69.89%, and 1024-node aggregate read and write bandwidths exceed 100 GB/s and 70 GB/s respectively. The success of Dawning Nebulae has demonstrated the viability of CPU/GPU heterogeneous structure for future designs of supercomputers.
Synthetic aperture radar (SAR) image segmentation is the basis of the image understanding. Combined with the Nyström sampling technique and the graph spectral theory, a new improved algorithm with fast and effect...
详细信息
Synthetic aperture radar (SAR) image segmentation is the basis of the image understanding. Combined with the Nyström sampling technique and the graph spectral theory, a new improved algorithm with fast and effective spectral clustering is proposed for the SAR image. Based on the matrix perturbation analysis theory, the automatic determining class number criterion suitable for the SAR image segmentation is constructed. Based on analysis of influence of proportion parameters on the spectral clustering algorithm, according to the global construction characteristics of the SAR image, auto-adaptive neighborhood estimate method of scale parameter is proposed. According to the gray value and the spatial location of each pixel in the SAR image, the affinity function better describing the essence structure of the SAR image is constructed, and then the improved spectral clustering algorithm is researched. The proposed algorithm is applied to the simulation experiment and in the actual SAR image segmentation, and also it is compared with the traditional spectral clustering method.
A novel layered method was proposed to solve the problem of Web services *** this method,services composition problem was formally transformed into the optimal matching problem of every layer,then optimal matching pro...
详细信息
A novel layered method was proposed to solve the problem of Web services *** this method,services composition problem was formally transformed into the optimal matching problem of every layer,then optimal matching problem was modeled based on the hypergraph theory,and solved by computing the minimal transversals of the ***,two optimization algorithms were designed to discard some useless states at the intermediary steps of the composition *** effectiveness of the composition method was tested by a set of experiments,in addition,an example regarding the travel services composition was also *** experimental results show that this method not only can automatically generate composition tree whose leaf nodes correspond to services composition solutions,but also has better performance on execution time and solution quality by adopting two proposed optimization algorithms.
Introduce the basic concept of best rational approximation and constructing procedure of the best rational approximation function in Laplace domain. By an illustrative example the design method of FOC based on best ra...
详细信息
Introduce the basic concept of best rational approximation and constructing procedure of the best rational approximation function in Laplace domain. By an illustrative example the design method of FOC based on best rational approximation is discussed. The performance comparison of transfer functions between the FOC obtained by best rational approximation method and conventional PID shows that it is effective to apply the method in design of FOC, which can guarantee maximum absolute error of amplitude frequency characteristic of approximation function in a given error tolerance.
Swarm intelligence is an umbrella for amount optimization algorithms. This discipline deals with natural and artificial systems composed of many individuals that coordinate their activities using decentralized control...
详细信息
Traditional temporal logics such as LTL (Linear Temporal Logic) and CTL (Computation Tree Logic) have shown tremendous success in specifying and verifying hardware and software systems. However, this kind of logic can...
详细信息
In this paper we investigate the stochastic multisymplectic methods to solve the stochastic partial differential equation. The stochastic KdV equations are considered. Besides conserving the multi-symplectic structure...
In this paper we investigate the stochastic multisymplectic methods to solve the stochastic partial differential equation. The stochastic KdV equations are considered. Besides conserving the multi-symplectic structure of original equation, the stochastic multi-symplectic methods are also investigated for the conservation of various conservation laws. We deduce the transit laws of the specific formal conservation laws. Numerical experiments are illustrated to verify the good behaviors of stochastic multisymplectic methods.
This work presents an iterative liveness-enforcing method for a class of generalized Petri nets, which can model flexible manufacturing systems. The proposed method checks the liveness of net models using mixed intege...
详细信息
This work presents an iterative liveness-enforcing method for a class of generalized Petri nets, which can model flexible manufacturing systems. The proposed method checks the liveness of net models using mixed integer programming and controls the token allocations of resource places instead of siphons using a liveness and resource usage ratio-enforcing supervisor. The enumeration of a kind of special structures, which is required in the previous work, is avoided and the number of iterations is bounded by the number of shared resource places in the net model. All strict minimal siphons in the controlled systems are minimally controlled. Several explanatory examples are used to illustrate this method.
A novel metric for full-reference image quality assessment (IQA) is proposed in this paper. Based on the sparse representation in independent component analysis (ICA) domain, the image basis is generated from natural ...
详细信息
A novel metric for full-reference image quality assessment (IQA) is proposed in this paper. Based on the sparse representation in independent component analysis (ICA) domain, the image basis is generated from natural images adaptively, which coincides with the characteristics of human vision system (HVS). In order to extract the feature vector, a hybrid norm optimization strategy is introduced for achieving more stable computational performances. The proposed IQA metric is calculated as a correlation coefficient between the two feature vectors from reference and distorted images, respectively. Experimental results on the LIVE Database Release 2 demonstrate that the proposed metric can achieve competitive performances as compared to the well-known structural similarity (SSIM) metric.
In high frequency radar, we should avoid noise disturbances in the radar's working-frequency segment. Moreover, the sidelobes of strong targets interfere with the detection of weak targets. A new method based on a...
详细信息
In high frequency radar, we should avoid noise disturbances in the radar's working-frequency segment. Moreover, the sidelobes of strong targets interfere with the detection of weak targets. A new method based on an adaptive selecting working-frequency is proposed. Frequency spectrum monitor is designed for selecting quiet frequency segment for the radar. Frequency spectrum monitor and the receiver of the radar are arranged to work according to special time periods respectively. So the radar can work in the frequency segments with lower noise disturbances. Moreover, there is no correlation between the noise and the useful echo signal, though the correlation between noises over very short time periods is strong, the noise data produced by frequency spectrum monitor can be exploited effectively Adjusting system parameters in real-time by adaptive methods can be utilized to reduce noise disturbances. Algorithm based on the properties of crosscorrelation between noise and target is exploited for suppressing sidelobe disturbances of strong targets. Lastly, the feasibility of the methods is verified by processing actual radar data.
暂无评论