Description:
The aim of this workshop is to deliver a "traning on the job" school based on a class of selected numerical methods for parallel Computational Fluid Dynamics (CFD). The workshop aims to share the methodologies, numerical methods and their implementation used by the state-of-the-art numerical codes used on High Performance Computing (HPC) clusters. The lectures will present the challenges of numerically solving Partial Differential Equations (PDE) in problems related to fluid-dynamics, using massively parallel clusters. The lectures will give a step-by-step walk through the numerical methods and their parallel aspects, starting from a serial code up to scalablity on clusters, including strategies for parallelization (MPI, OpenMPI, use of Accelerators, plug-in of numerical libraries,....) with hands-on during the lectures. Profiling and optimization techiques on standar and heterogeneous clustes will be shown during the school. Further information will be available later for participants upon confirmations of the speakers.
Skills:
At the end of the course, the student will possess and know how to use the following skills:
- Numerical analysis
- Algorithms for PDE Solution
- Parallel computing (MPI, OpenMP, Accelerators)
- HPC architecture
- Strategies for massively parallelization of numerical methods
- Numerical Libraries for HPC
Target audience:
MSc/PhD students, Post-Docs, Academic and industrial researchers, software developers which use / are planning to use / develop a code for CFD
Pre-requisites:
Previuos course(s) on parallel computing, numerical analysis and algorithms for p.d.e. solution.