A MESSAGE-PASSING, DISTRIBUTED MEMORY PARALLEL ALGORITHN FOR A THIN LAYER NAVIER-STOKES SOLVER

In this paper the results of parallelizing a 3-D TLNS (Thin Layer Navier-Stokes) solver for supersonic turbulent flows are presented. As a serial code this solver is very time consuming and takes a large part of memory due to the complicated and lengthy computations. Also for complicated geometries an exceeding number of grid points are required which necessitate larger serial computation times. Therefore, parallelizing this code has brought a lot of savings in computer time and memory for each computational node. For executing this code on parallel processors, the domain is partitioned in a proper form and different methods for communicating the data such as blocking send and non-blocking send are investigated. The performance of each method is evaluated and the best method for the problem at hand is indicated. The results are reported in terms of speedup and efficiency obtained from running the parallel code on 1,2,4,8 and 16 processors. Also the contours of Mach number and pressure for a supersonic turbulent flow around a tangentogive at angle of attack for serial and parallel executions are presented.