A new parallel overlapped domain decomposition method for nonlinear dynamic finite element analysis

In this paper a new parallel algorithm for nonlinear transient dynamic analysis of large structures has been presented. An unconditionally stable Newmark-beta method (constant average acceleration technique) has been employed for time integration. The proposed parallel algorithm has been devised within the broad framework of domain decomposition techniques. However, unlike most of the existing parallel algorithms (devised for structural dynamic applications) which are basically derived using nonoverlapped domains, the proposed algorithm uses overlapped domains. The parallel overlapped domain decomposition algorithm proposed in this paper has been formulated by splitting the mass, damping and stiffness matrices arises out of finite element discretisation of a given structure. A predictor-corrector scheme has been formulated for iteratively improving the solution in each step. A computer program based on the proposed algorithm has been developed and implemented with message passing interface as software development environment. PARAM-10000 MIMD parallel computer has been used to evaluate the performances. Numerical experiments have been conducted to validate as well as to evaluate the performance of the proposed parallel algorithm. Comparisons have been made with the conventional nonoverlapped domain decomposition algorithms. Numerical studies indicate that the proposed algorithm is superior in performance to the conventional domain decomposition algorithms. (C) 2003 Elsevier Ltd. All rights reserved.

Topics in randomized numerical linear algebra

This thesis studies three classes of randomized numerical linear algebra algorithms, namely: (i) randomized matrix sparsification algorithms, (ii) low-rank approximation algorithms that use randomized unitary transformations, and (iii) low-rank approximation algorithms for positive-semidefinite (PSD) matrices.

Randomized matrix sparsification algorithms set randomly chosen entries of the input matrix to zero. When the approximant is substituted for the original matrix in computations, its sparsity allows one to employ faster sparsity-exploiting algorithms. This thesis contributes bounds on the approximation error of nonuniform randomized sparsification schemes, measured in the spectral norm and two NP-hard norms that are of interest in computational graph theory and subset selection applications.

Low-rank approximations based on randomized unitary transformations have several desirable properties: they have low communication costs, are amenable to parallel implementation, and exploit the existence of fast transform algorithms. This thesis investigates the tradeoff between the accuracy and cost of generating such approximations. State-of-the-art spectral and Frobenius-norm error bounds are provided.

The last class of algorithms considered are SPSD "sketching" algorithms. Such sketches can be computed faster than approximations based on projecting onto mixtures of the columns of the matrix. The performance of several such sketching schemes is empirically evaluated using a suite of canonical matrices drawn from machine learning and data analysis applications, and a framework is developed for establishing theoretical error bounds.

In addition to studying these algorithms, this thesis extends the Matrix Laplace Transform framework to derive Chernoff and Bernstein inequalities that apply to all the eigenvalues of certain classes of random matrices. These inequalities are used to investigate the behavior of the singular values of a matrix under random sampling, and to derive convergence rates for each individual eigenvalue of a sample covariance matrix.

On parallel computing for stochastic optimization models and algorithms

167 p.

Parallel Simulation of Subsonic Fluid Dynamics on a Cluster of Workstations

An effective approach of simulating fluid dynamics on a cluster of non- dedicated workstations is presented. The approach uses local interaction algorithms, small communication capacity, and automatic migration of parallel processes from busy hosts to free hosts. The approach is well- suited for simulating subsonic flow problems which involve both hydrodynamics and acoustic waves; for example, the flow of air inside wind musical instruments. Typical simulations achieve $80\\%$ parallel efficiency (speedup/processors) using 20 HP-Apollo workstations. Detailed measurements of the parallel efficiency of 2D and 3D simulations are presented, and a theoretical model of efficiency is developed which fits closely the measurements. Two numerical methods of fluid dynamics are tested: explicit finite differences, and the lattice Boltzmann method.

Automatic parallel code generation for message passing on distributed memory systems

The availability of a very accurate dependence graph for a scalar code is the basis for the automatic generation of an efficient parallel implementation. The strategy for this task which is encapsulated in a comprehensive data partitioning code generation algorithm is described. This algorithm involves the data partition, calculation of assignment ranges for partitioned arrays, addition of a comprehensive set of execution control masks, altering loop limits, addition and optimisation of communications for all data. In this context, the development and implementation of strategies to merge communications wherever possible has proved an important feature in producing efficient parallel implementations for numerical mesh based codes. The code generation strategies described here are embedded within the Computer Aided Parallelisation tools (CAPTools) software as a key part of a toolkit for automating as much as possible of the parallelisation process for mesh based numerical codes. The algorithms used enables parallelisation of real computational mechanics codes with only minor user interaction and without any prior manual customisation of the serial code to suit the parallelisation tool.

An application of quasi-Newton methods for the numerical solution of interface problems

Quasi-Newton methods are applied to solve interface problems which arise from domain decomposition methods. These interface problems are usually sparse systems of linear or nonlinear equations. We are interested in applying these methods to systems of linear equations where we are not able or willing to calculate the Jacobian matrices as well as to systems of nonlinear equations resulting from nonlinear elliptic problems in the context of domain decomposition. Suitability for parallel implementation of these algorithms on coarse-grained parallel computers is discussed.

Numerical simulation of incompressible flow problems using an unstructured staggered mesh method

A number of two dimensional staggered unstructured discretisation schemes for the solution of fluid flow and heat transfer problems have been developed. All schemes store and solve velocity vector components at cell faces with scalar variables solved at cell centres. The velocity is resolved into face-normal and face-parallel components and the various schemes investigated differ in the treatment of the parallel component. Steady-state and time-dependent fluid flow and thermal energy equations are solved with the well known pressure correction scheme, SIMPLE, employed to couple continuity and momentum. The numerical methods developed are tested on well known benchmark cases: the Lid-Driven Cavity, Natural Convection in a Cavity and Melting of Gallium in a rectangular domain. The results obtained are shown to be comparable to benchmark, but with accuracy dependent on scheme selection.

Design of parallel algorithms for fractal video compression

The intrinsic independent features of the optimal codebook cubes searching process in fractal video compression systems are examined and exploited. The design of a suitable parallel algorithm reflecting the concept is presented. The Message Passing Interface (MPI) is chosen to be the communication tool for the implementation of the parallel algorithm on distributed memory parallel computers. Experimental results show that the parallel algorithm is able to reduce the compression time and achieve a high speed-up without changing the compression ratio and the quality of the decompressed image. A scalability test was also performed, and the results show that this parallel algorithm is scalable.