Navigation

Operating Systems Group
Operating Systems Group

Scalability of Molecular Dynamics Simulations on Heterogeneous Hardware

Molecular dynamics simulations have become a vital research method in biochemistry and materials science. Since the problem size is typically fixed by the physical size of molecules, MD applications target strong scaling. Thus, the computational effort per compute node is very low and MD applications tend to be latency- and synchronization-critical. To exploit the trend towards heterogeneous hardware architectures, MD applications need to adapt to the differences in latency, throughput and memory hierarchy caused by combining multi-core and many-core CPUs with GPUs and accelerators.

Publications

    2020

  1. Laura Morgenstern, Ivo Kabadshow, "The Rocky Road to Tasking: Task Queues Reloaded", 2020
     
  2. Laura Morgenstern, David Haensel, Andreas Beckmann, Ivo Kabadshow, "NUMA-Awareness as a Plug-In for an Eventify-Based Fast Multipole Method", in Computational Science - ICCS 2020, 428-441, 2020 
  3. Laura Morgenstern, Ivo Kabadshow, Matthias Werner, "GPU-Tasking à la Carte? Eventify Meets GPUs", 2020
     
  4. David Haensel, Laura Morgenstern, Andreas Beckmanns, Ivo Kabadshow, Holger Dachsel, "Eventify: Event-Based Task Parallelism for Strong Scaling", in Proceedings of the Platform for Advanced Scientific Computing Conference, 2020 
  5. Laura Morgenstern, "Tasking Meets GPUs: A Closer Look at Locking", 2020
     


  6. 2019

  7. Laura Morgenstern, David Haensel, Andreas Beckmann, Ivo Kabadshow, "NUMA-Awareness as a Plug-In", 2019
     
  8. Laura Morgenstern, Andreas Beckmann Ivo Kabadshow, Matthias Werner, "Tasking Meets GPUs: Fighting Deadlocks and Other Monsters", 2019
     
  9. Laura Morgenstern, Andreas Beckmann, Ivo Kabadshow, "Towards Unified Tasking on CPUs and GPUs", 2019
     
  10. Bartosz Kohnke, Thomas R. Ullmann, Andreas Beckmann, Ivo Kabadshow, David Haensel, Laura Morgenstern, Plamen Dobrev, Gerrit Groenhof, Carsten Kutzner, Berk Hess, Holger Dachsel, Helmut Grubmüller, "A scalable and versatile Fast Multipole Method for biomolecular simulation", in Software for Exascale Computing - SPPEXA 2016-2019, Springer Lecture Notes in Computer Science and Engineering, 2019 


  11. 2018

  12. Laura Morgenstern, "A NUMA-aware Task-based Load-balancing Scheme for the Fast Multipole Method", 2018
     
  13. Laura Morgenstern, "Scalability Enhancements to FMM for MD Simulations", 2018
     
  14. Laura Morgenstern, Andreas Beckmann, Ivo Kabadshow, "Towards Supporting Heterogeneous Hardware in Gromacs", 2018
     


  15. 2017

  16. Laura Morgenstern, "Towards a NUMA-aware Task-based Fast Multipole Method", in Fall School of the HPI Future SOC Lab on Efficient Architectures for Data Science (EADS 2017), 2017 

Contact