SCM's high-quality modelling software is already used worldwide in academia, government labs, HPC centers and world-leading companies. Nevertheless, for SMEs the barrier for starting new modelling projects is often too high because of the need to hire qualified R&D system maintenance personnel and obtain suitable hardware for realistic, large-size calculations.
The objective of the experiment is to partially rewrite existing density-functional theory (DFT) and especially density-functional based tight-binding (DFTB) as well as (reactive) Molecular Dynamics (MD) codes to optimally enable CPU-GPU hybrid supercomputers architectures. Two-level parallelization will be implemented in existing codes as well as a newly written MD engine, which will make truly massive parallel calculations possible on supercomputers with many thousands of cores.
Once a working implementation of the accelerated code is in place on CPU-GPU and multiprocessor supercomputing architectures, this can then become a one-stop-shop for SMEs that want to start modeling large-scale molecular systems. With SCM's fully integrated GUI, end users can prepare jobs on their own desktop environment, submit the jobs to the cloud, and analyze the results locally as they become available. In collaboration with SURFsara a cloud-based infrastructure will be developed to sustain the delivery to a larger group of SMEs.
The output of the experiment will be optimized DFT(B) and reactive MD (ReaxFF) codes for use on CPU-GPU hybrid supercomputers architecture. Also a business model and licensing scheme will be developed for cloud-based HPC services.
Chemicals & man-made fibres
Modelling & Simulation