AFiD-GPU: a versatile Navier-Stokes Solver for Wall-Bounded Turbulent Flows on GPU Clusters

摘要

The AFiD code, an open source solver for the incompressible Navier-Stokesequations ({\color{blue}\burl{http://www.afid.eu}}), has been ported to GPUclusters to tackle large-scale wall-bounded turbulent flow simulations. The GPUporting has been carried out in CUDA Fortran with the extensive use of kernelloop directives (CUF kernels) in order to have a source code as close aspossible to the original CPU version; just a few routines have been manuallyrewritten. A new transpose scheme, which is not limited to the GPU version onlyand can be generally applied to any CFD code that uses pencil distributedparallelization, has been devised to improve the scaling of the Poisson solver,the main bottleneck of incompressible solvers. The GPU version can reduce thewall clock time by an order of magnitude compared to the CPU version for largemeshes. Due to the increased performance and efficient use of memory, the GPUversion of AFiD can perform simulations in parameter ranges that areunprecedented in thermally-driven wall-bounded turbulence. To verify theaccuracy of the code, turbulent Rayleigh-B\'enard convection and plane Couetteflow are simulated and the results are in good agreement with the experimentaland computational data that published in previous literatures.