We investigate numerical methods for passive model reduction of linear dynamical systems. This is an important task in circuit simulation when modeling parasitic effects of interconnect. We will show how positive real...
详细信息
We investigate numerical methods for passive model reduction of linear dynamical systems. This is an important task in circuit simulation when modeling parasitic effects of interconnect. We will show how positive real balancing, based on balancing the solutions of two algebraic Riccati equations, can be used for passive model reduction of large-scale systems on parallel computers. Numerical experiments demonstrate the performance of the parallel algorithms using several examples from circuit simulation.
A parallel Hough transform algorithm on linear array with reconfigurable pipeline bus system (LARPBS) is presented. Utilizing the strong communication and computation ability of LARPBS, the algorithm can contract and ...
详细信息
A parallel Hough transform algorithm on linear array with reconfigurable pipeline bus system (LARPBS) is presented. Utilizing the strong communication and computation ability of LARPBS, the algorithm can contract and count the foreground pixels on each line. Suppose the number of θ values considered is m, the algorithm can complete Hough transform for an image with n×n pixels in O(1) time using mn2 processors and get optimal speed and efficiency. We also illustrate how to partition and perform the algorithm on a LARPBS with less than mn2 processors, thus show the algorithm is highly scalable.
To meet the real time requirements an interpolation free, parallel algorithm for the fast Radon transform (FRT) and inverse FRT (IFRT) is proposed. The proposed method solves all the important problems associated with...
详细信息
To meet the real time requirements an interpolation free, parallel algorithm for the fast Radon transform (FRT) and inverse FRT (IFRT) is proposed. The proposed method solves all the important problems associated with the previous interpolation free FRT and IFRT algorithm and reduces the number of computations and algorithmic complexities significantly. The proposed algorithm is highly regular and we also describe a methodology to design a dedicated parallel processing architecture from the view point of its efficient implementation.
We conduct direct numerical simulations with 300 million degrees of freedom of turbulent flows past a stationary and a forced oscillating rigid cylinder at the Reynolds number Re=10,000. This one-order of magnitude in...
详细信息
We conduct direct numerical simulations with 300 million degrees of freedom of turbulent flows past a stationary and a forced oscillating rigid cylinder at the Reynolds number Re=10,000. This one-order of magnitude increase in Reynolds number (compared to previous DNS) is accomplished by employing a multilevel-type parallel algorithm within the spectral element framework. Comparisons with the available experimental data show that the simulation has captured the flow quantities, mean, and rms statistics of the cylinder wake correctly. We also examine the effect of the randomness in the inflow on the vortex formation at a lower Reynolds number. We demonstrate that noisy inflows cause a vortex shedding-mode switching in the cylinder wake.
The goal of this paper is to classify abelian permutation group problems using logspace counting classes. Building on McKenzie and Cook's [MC87] classification of permutation group problems into four NC Turing-equ...
详细信息
The goal of this paper is to classify abelian permutation group problems using logspace counting classes. Building on McKenzie and Cook's [MC87] classification of permutation group problems into four NC Turing-equivalent sets, we show that all these problems are essentially captured by the generalized logspace mod-class ModL, where ModL is the logspace analogue of ModP (defined by Kobler and Toda (KT96)). More precisely, our results are as follows: 1. For abelian permutation groups, the problems of membership testing, isomorphism testing and computing the order of a group are all in ZPL/sup ModL/, and are all hard for ModL under logspace Turing reductions. 2. The problems of computing the intersection of abelian permutation groups, and computing a generator-relator presentation for a given abelian permutation group are in FL/sup ModL//poly. Furthermore, the search version of membership testing is also in FL/sup ModL//poly.
In this paper, a concurrent PSO (CONPSO) algorithm is proposed to alleviate the premature convergence problem of PSO algorithm. It is a type of parallel algorithm in which modified PSO and FDR-PS algorithms are simula...
详细信息
In this paper, a concurrent PSO (CONPSO) algorithm is proposed to alleviate the premature convergence problem of PSO algorithm. It is a type of parallel algorithm in which modified PSO and FDR-PS algorithms are simulated concurrently with frequent message passing between them. This algorithm avoids the possible crosstalk effect of pbest and gbest terms with nbest term in FDR-PSO. Thereby, search efficiency of proposed algorithm is improved. In order to demonstrate the effectiveness of the proposed algorithm, experiments were conducted on six benchmarks continuous optimization problems. Results clearly demonstrate the superior performance of the proposed algorithm in terms of solution quality, average computation time and consistency. This algorithm is very much suitable for the implementation in parallel computer.
Summary form only given. There is today an increasing diversity of parallel execution supports. Solving a target problem by using a single algorithm is not always efficient on any computational support. We present her...
详细信息
Summary form only given. There is today an increasing diversity of parallel execution supports. Solving a target problem by using a single algorithm is not always efficient on any computational support. We present here a polyalgorithmic approach for selecting the most suitable algorithm among various ones for given problem size and available resources. Our principal objective here is to illustrate such an approach on the well-known matrix multiplication problem which is one of the most important basic numerical kernels. More precisely, we propose a polyalgorithm which uses both advantages of standard and fast algorithms which is able to automatically choose the right and suitable algorithm for computing the matrix multiplication of any dimension on a particular parallel system. We target this approach on homogeneous clusters of PCs while providing some experiments.
In this paper we present a general approach for reducing switching activity on the algorithmic level. We concentrate on iterative algorithms that are suitable for an implementation on parallel processor arrays. The re...
详细信息
In this paper we present a general approach for reducing switching activity on the algorithmic level. We concentrate on iterative algorithms that are suitable for an implementation on parallel processor arrays. The reduction is substantially reached by avoiding operations that hardly contribute to the convergence of the implemented algorithm. Our general approach is exemplified on the implementation of a specific algorithm, i.e. the eigenvalue decomposition (EVD) of a real symmetric matrix.
In this paper we report on three parallel algorithms used for 3D reconstruction of asymmetric objects from their 2D projections. We discuss their computational, communication, I/O, and space requirements and present s...
详细信息
Given N matrices A1, A2,..., AN of size NxN, the matrix chain product problem is to compute A1 x A2 xx AN. Given an N x N matrix A, the matrix powers problem is to calculate the first N powers of A, i.e., A, A2, A3,.....
详细信息
暂无评论