The goal of ORKA-HPC is to develop a software-technical ecosystem for OpenMP implementations so that across all manufacturers FPGAs become accessible for application codes from the HPC and data analytics area. For this purpose, C/C++ code with OpenMP directives is first to be translated into annotated C/C++ code via source-to-source compiler and then translated into a synthesisable high-level representation. Finally, compiler backends generate an executable program for the host CPU and FPGA that can be run in a distributed manner on a host CPU and a connected FPGA.
In addition to the development of the source-to-source compiler, the project aims to develop a productive tool environment that allows a fine-grained resolution of the application code to FPGA concepts such as memory interfaces and internal communication network with optimised routing.