Parallelisation of a navier-stokes code for two-dimensional13; incompressible flows

An existing Finite Volume code,developed originally. on a sequential computer for calculation of two-dimensional, time-averaged, turbulent, incompressible flows in complex13; configurations is modified suitably to run on FLOSOLVER MK2. The sequential algorithm consists of two major parts13; (i) the generation of coefficients for the diccretised, quasi linear, coupled system, of algebraic equations representing the flux balance, of mass, m o m enta and other scalars over - a control volume (ii) the solution of the system of linear equations for pressure, two velocity components and two turbulent scalars in a decoupled manner. The iterative algorithm and the , corresponding structure of the code is kept unaltered. The computatiorising domain is subdivided into four subdo m sins to be coupled concurrently by four processors and suitable provisions are in ade for interprocessor transfer of data for the variable values and fluxes at the artificial boundaries between two adjacent processors .13; The algorithm employs generalized non-orthogonal coordinate13; system and can handle complex flow boundaries . In the13; present work, however, the performance of the parallel13; version of the code Is dP m c n_-t2rat,=a fvc turbulent flow past a Backward Facing Step using simple Cartesian grids. Comparison of performance between the -sequential and parallel version for 20X20 control! volume computation13; reveals a speed up factor of 3 .15 to achieve the same13; accuracy using four processors