Parallel Agglomeration Strategies for Industrial Unstructured Solvers

The use of fluid dynamics codes with unstructured grids continues to grow in popularity. This is particularly true for fluid applications that do not involve strong separation or turbulence or where the viscous/inviscid interaction is weak. In these cases the Navier-Stokes equations can be reduced to the Euler equations. For many practical applications, the flow of air over an aircraft can be modeled using the Euler equations because the Reynolds number is high and the viscous effects are primarily restricted to the boundary layer at the aircraft's surface. The computational grid required to model a complete aircraft can be very complex. In this chapter, an industrial aerospace code is parallelized for execution on shared memory and distributed memory architectures. The flow solver uses an explicit time integration approach and utilizes an agglomeration multigrid strategy to accelerate the convergence. A novel method has been used that efficiently parallelizes the multigrid solver and minimizes any intrusion into the flow solver. The parallel code uses standard message passing routines and is demonstrated to be portable.