Tridiagonal matrix algorithm

Abstract: This paper proposes an accurate integral-based scheme for solving the advection–diffusion equation. In the proposed scheme the advection–diffusion equation is integrated over a computational element using the quadratic polynomial interpolation function. Then elements are connected by the continuity of first derivative at boundary points of adjacent elements. The proposed scheme is unconditionally stable and results in a tridiagonal system of equations which can be solved efficiently by the Thomas algorithm. Using the method of fractional steps, the proposed scheme can be extended straightforwardly from one-dimensional to multi-dimensional problems without much difficulty and complication. To investigate the computational performances of the proposed scheme five numerical examples are considered: (i) dispersion of Gaussian concentration distribution in one-dimensional uniform flow; (ii) one-dimensional viscous Burgers equation; (iii) pure advection of Gaussian concentration distribution in two-dimensional uniform flow; (iv) pure advection of Gaussian concentration distribution in two-dimensional rigid-body rotating flow; and (v) three-dimensional diffusion in a shear flow. In comparison not only with the QUICKEST scheme, the fully time-centred implicit QUICK scheme and the fully time-centred implicit TCSD scheme for one-dimensional problem but also with the ADI-QUICK scheme, the ADI-TCSD scheme and the MOSQUITO scheme for two-dimensional problems, the proposed scheme shows convincing computational performances. Copyright © 2001 John Wiley & Sons, Ltd.

Abstract: Cyclic reduction, originally proposed by Hockney and Golub, is the most popular algorithm for solving tridiagonal linear systems on SIMD-type computers like CRAY-1 or CDC CYBER 205. That algorithm seems to be the adequate one for the IBM 3090 VF (uni-processor), too, although the overall expected speedup over Gaussian elimination, specialized for tridiagonal systems, is not as high as for the CRAY-1 or the CYBER 205. That is because the excellent scalar speed of the IBM 3090 makes its vector-to-scalar speed ratio relatively moderate. The idea of the cyclic reduction algorithm can be generalized and modified in various directions. A polyalgorithm can be derived which takes into account much better the given architecture of the IBM 3090 VF than the ‘pure’ cyclic reduction algorithm as described for instance by Kershaw. This is mainly achieved by introducing more locality into the formulae. For large systems of equations the well-known cache problems are prevented.