Towards Domain-Specific Computing for Stencil Codes in HPC

High Performance Computing (HPC) systems are nowadays more and more heterogeneous. Different processor types can be found on a single node including accelerators such as Graphics Processing Units (GPUs). To cope with the challenge of programming such complex systems, this work presents a domain-specific approach to automatically generate code tailored to different processor types. Low-level CUDA and OpenCL is generated from a high-level description of a geometric multigrid algorithm written in a Domain-Specific Language (DSL) instead of writing hand-tuned code for GPU accelerators. By decoupling the algorithm from its schedule, the proposed approach allows to generate efficient stencil codes. Our results show that competitive performance compared to hand-tuned codes can be achieved and that more than 25 frames per second for 16.8 Megapixel images are obtained for full High Dynamic Range (HDR) compression of 2D medical data sets.