For this project, two very important CFD applications have been upgraded: one which is used to study mesoscale eddy processes in ocean modelling - PEQUOD (the Parallel Quasi-Geostrophic Model) and the other, Cfoam-CABARET for quasi-DNS simulations for aeroacoustics. Both use the CABARET finite difference scheme to resolve the advection equations. In PEQUOD, the global collectives for the parallel 1D tri-diagonal solver were updated by implementing the hardware optimised collective MPI_ALLTOALL. For a representative grid size of 1025, this now enables a 2 times speedup and good weak scaling on a fixed number of grid points per MPI task basis. For Cfoam-CABARET, MPI-IO was introduced for the restart and Tecplot360 visualisation data files. This eliminates the requirement for many thousands of files to be present at run-time and enables more practical use of the application on HECToR and future HPC architectures.
Future results obtained using the PEQUOD code will be beneficial for several communities. Climate modellers will be interested for understanding how internal nonlinear dynamics of the ocean can contribute to the climate variability. Ocean observers will be interested in knowing what are the key features and properties of the large-scale low-frequency variability and the eddy field that needs to be accurately estimated. Ocean modellers who work with comprehensive general circulation models might be able to enhance their non-eddy-resolving or partially eddy-resolving codes; results from PEQUOD will provide some guidance. Theoreticians will also benefit because the issues of the intrinsic large-scale low-frequency variability and the associated eddy effects are very fundamental ones.
Many computational research papers deal with various aspects of the quasigeostrophic turbulence, and there is growing demand for more turbulent and dynamically realistic solutions. There is a set of very interesting and important results that are waiting to be revisited and upgraded, with a completely new level of dynamical realism that can only be achieved with PEQUOD and HECToR, This is particularly relevant to the NERC project (NE/H020837/1) - "A new approach to parameterising ocean eddies: energetics, conservation and flow stability", October 2010-September 2013.
For Cfoam-CABARET and jet-flap-noise, this dCSE project will enhance the machinery required to answer a few important scientific questions. One of them will be "What is the effect of fine-scale-flow structures on far-field noise in the audible range of frequencies?" With the enhanced numerical model based on the unstructured CABARET and HECToR, this will cope with a significant increase in the grid resolution, so that this question may be answered. The EPSRC projects (EP/I017747/1) "Flap Noise", April 2011-September 2013 and (EP/I017771/1) "Aerodynamics and aeroacoustics of complex geometry hot jets", November 2011- September 2015.
On HECToR Phase 2b, 20,000kAUs were used with CABARET and so far more than 17,000kAUs have been used on Phase 3. Under a recent Flap Noise RAP Top Up award a further 19,600 kAUs were received, which together with the remaining 3,000k AUs awarded under the original EPSRC Flap Noise project will be used by September 2013.
Phil Ridley 2012-10-01