A parallel mesh chopping algorithm for a class of two-point boundary value problems.pdf

A Parallel Mesh Chopping Algorithm for a Class of Two-Point Boundary Value Problems C. P. KATTI School of Computer and Systems Sciences Jawaharlal Nehru University, New Delhi 110 067, India S. GOEL Computer Centre, Indian Institute of Technology New Delhi 110 016, India (Received January 1996; revised and accepted March 1997) Abstract—-A parallel algorithm for the class of two-point boundary value problems y = f(z, y), y(0) = A, y(1) = B with ~ 0 and ~ continuous on [0,1] x (—oo,oc) is presented. Using an idea similar to that in [1], we divide [0,1] into p different divisions, each division consisting of N or (N+1) (N small) unequal intervals. A high-order finite difference scheme developed for general nonuniform mesh is now applied to the above class of TPBVPs on each of the p divisions and leads to an N x N or (N — 1) x (N — 1) system of linear or nonlinear equations which is solved on p processors (p a power of 2) simultaneously. Keywords-—-Boundary value problems, Parallel chopping algorithm, Hypercube, Discretization, Convergence. I. INTRODUCTION We consider the class of two-point boundary value problems: y = f(x,y), y(0) = A, y(1) = B, (1) where ~ 0 and ~ is continuous on [0,1] x (-cx3,c~). It is well known that (1) has a unique solution [2]. A number of papers have recently appeared in the literature which have considered the solution of (1) on parallel computers; see, e.g., [3-5]. Most of these, however, consider parallelizing the matrix computations which arise when (1) is replaced by its equivalent finite difference scheme. A parallel chopping algorithm is considered in [1] where on a computer with p processors, the BVP is solved numerically at each stage on p meshes using a code based on COLNEW [6,7]. In the following, we present a parallel algorithm to solve (1) on a computer with p processors, p a power of 2. Using an idea similar to that in [1], we consider p different divisions of [0,1], each division consisting of N or (N %1) (N small) unequally sp