TaLPas: Task-Based Load Balancing and Auto-Tuning in Particle Simulations
TimeTuesday, June 23rd4:35pm - 4:37pm
DescriptionTaLPas provides a solution to fast and robust simulation of many, inter-dependent particle systems in peta- and exascale supercomputing environments. This will be beneficial for a wide range of applications. TaLPas focuses on sampling in molecular dynamics (rare event sampling, sampling of equations of state, etc.).
1. the development of an auto-tuning based particle simulation library AutoPas to leverage optimal node-level performance,
2. the development of a scalable workflow manager to optimally distribute inter-dependent particle simulation tasks on HPC compute resources,
3. the investigation of performance prediction methods for particle simulations to support auto-tuning and to feed the workflow manager with accurate runtime predictions,
4. the integration of 1-3, augmented by visualization of the sampling (parameter space exploration) and an approach to resilience. The latter will guarantee robustness at peta- and exascale.
On this poster, we discuss latest work in the project that has just entered its final phase. We discuss the integration of the auto-tuning library AutoPas in the MD code ls1 mardyn and its use in massively parallel, load-balanced simulations. Furthermore, we outline the application of the TaLPas workflow manager for MD sampling in adsorption processes, including latest developments on performance prediction. Besides, visualization aspects are addressed.