The Fraunhofer Competence Center for HPC is engaged in national and international research projects to push performance limits and create innovative solutions in the HPC, Big Data and AI continuum. As Fraunhofer organization we act as research partner for industry to tune their applications and to create new HPC applications. The CC-HPC has invented the GPI (Global Address Space) programming model which drives now the performance in many applications and enables new tools like Tarantella (Distributed Deep Learning Training), GASPILS (A scalable Iterative Solver Library) or GPI-Space (Memory Driven Computing). We provide tools to manage ML developments (CARME) and tune DNN´s with Neural Architecture Search methods (NASE).
The CC-HPC has developed the BeeGFS parallel file system and current research projects contribute to future BeeGFS enhancements and new features.
A major activity at the CC-HPC since 2009 is hardware-software co-design, which led now to the Stencil Accelerator Processor (STX) that is developed within the European Processor Initiative EPI. Extensions of the LLVM compiler and OpenMP will simplify the use of the STX accelerator for a large group of applications.
Stop by our website to get the latest news on our highlights and get in contact with us. We will be happy to meet you online this year. ...just contact us via mail to set up a video chat.