Parallelisation of CABARET
The CABARET (Compact Accurate Boundary Adjusting high REsolution Technique) code may be used to solve the compressible Navier-Stokes equations. The code is based on a low dissipative and low dispersive conservative CABARET method that constitutes a substantial upgrade of the second-order upwind leapfrog scheme. Most notably, the CABARET algorithm has a very local computational stencil that for scalar advection constitutes only one cell in space and time.
The present application of CABARET is for the investigation of aircraft noise, which is currently a very important environmental concern. An important component of aircraft noise is due to airframe/engine installation effects, the reduction of this remains a very challenging problem. In particular, when deployed at a large angle of attack at approach conditions, the wing flaps become a very important noise source. For engine-under-a-wing configurations, flap interaction with the jet can even become a dominant noise component. A crucial element of any noise prediction scheme is the high fidelity Large Eddy simulation (LES) model. For the airframe/engine noise problem, this model needs to accurately capture all important wing-flap, free jet and wing-flap-jet interaction effects.
The overall outcome of this work may be summarised as follows:
- An automated geometrical domain decomposition method for partitioning a Gambit generated unstructured grid has been developed. The resulting partitions are ensured to give good load balancing for the main computation and minimise communication. The partitioner has been demonstrated on a 51.2M cell grid.
- A parallel version of CABARET has been implemented using non-blocking MPI to pass data between internal boundaries (cell faces and sides). This code was validated and tested against the old code using a 3D backward-facing step case and a parallel efficiency of 72% was observed when using 250 cores of the XT4 part of HECToR with the quad core Phase 2a architecture.
- The new code was validated for a larger test case of 51.2M cells on the XT4 part of HECToR, using the 3D backward-facing step case.
- A hybrid (OpenMP/MPI) parallel version of CABARET for Phase 2b and further has been developed. Also a hybrid parallel partitioner has been developed for the domain decomposition of the Gambit generated unstructured grid.
- A 3D backward-facing step test case of 51.2M cells, retains 80% parallel efficiency on 1000 cores of Phase 2b.
- Simulations can now be performed on grids at least 512 times larger than previously possible. Smaller scale simulations can now be performed in a few minutes that would otherwise have taken several days before this work.