The research project HICFD is a collaborative project within the program "ICT 2020 – Research and Innovation," funded by the Federal Ministry of Education and Research. The aim of the project is to develop new methods and tools for the analysis and optimization of the performance of computational fluid dynamics (CFD) parallel programs on high-performance computers with many-core processors, and to apply these methods as examples to the CFD codes TAU and TRACE of the project partner DLR. The main objectives of the project are presented in this work. The performance of parallel CFD programs is to be improved in this project by optimally utilizing all levels of parallelism. At the highest level (MPI), intelligent grid partitioning is intended to improve load balancing between MPI processes. For block-structured grids, a multi-core compatible partitioning tool is to be developed.

At the multi-core architecture level, highly scalable hybrid OpenMP/MPI methods will be implemented as examples for the two CFD solvers TAU and TRACE. At the processor core level, a preprocessor will be developed to enable the convenient use of parallel SIMD units ("Single Instruction Multiple Data") even for complex applications. To check to what extent the use of the preprocessor influences the performance of parallel programs, the performance analysis software Vampir will be extended towards SIMD.