On Nov 6, the first ENCCS developer progress meeting was held. We took an in-depth look at the work Mark Abraham will do in Life Science, particularly with GROMACS. GROMACS is a mature widely-used code with a clearly articulated exascale roadmap (see https://zenodo.org/record/3563747). Mark’s work will focus on areas of that roadmap most relevant to Swedish researchers, particularly for using new HPC infrastructure such as the LUMI supercomputer (https://www.lumi-supercomputer.eu/).
- Collaborate with AMD to find a suitable way to support the GPUs on LUMI, either via AMD’s vendor-owned HIP language or the upcoming standard SYCL language via hipSYCL (https://github.com/illuhad/hipSYCL),
- Contribute design and refactoring work to enable researchers to easily implement robust and performant API-driven workflows in Python (http://manual.gromacs.org/current/gmxapi/) or C++ (upcoming in GROMACS 2021)
- Develop a new modern C++ FFT library to provide an independent showcase of ENCCS development capabilities, while delivering performance enhancements to GROMACS, which uses small 3D FFTs not typically optimized well by existing libraries