Parallel Algorithm Based on OpenCL for Simulation of Subsonic Flows with Chemical Reactions

The paper is devoted to developing a parallel computational algorithm based on OpenCL technology for the numerical study of subsonic chemically reacting gas flows with multicomponent diffusion, viscosity, and thermal conductivity. The mathematical model of the process is a modified system of Navier–Stokes equations in a low Mach number approximation. The computational algorithm is based on the splitting by physical processes: the problem is decomposed into several subproblems, each having its own numerical solution method. The parallel algorithm is implemented in C++ using OpenCL technology due to the large number of computational operations with relatively little input data. Serial calculations are performed for the problem of ethane pyrolysis in a one-dimensional formulation. The compact radical scheme, which conveys all the process features, is taken as the kinetic model. Calculations are performed on grids of different dimensions and preserve Courant’s number. We show the coincidence of the results of the sequential and parallel versions. The developed program shows a growth in speedup as the number of calculated cells increases. With 880 cells, the program speedup is more than 11 times that of the sequential version.