The first Tinker developer’s conference was held in St. Louis in March 2017. This interdisciplinary workshop gathered theoretical chemists, biophysicists, physicists, biomedical engineers and applied mathematicians from various institutions including Washington University, University of Texas at Austin, Sorbonne university, Stanford Univ., University of Iowa, NIH, Florida State Univ, UC Berkeley etc. The discussions planned the developments of the current 3 Tinker codes co-existing within the Tinker 8 distribution:
- Tinker (canonical code version 8, OpenMP) https://dasher.wustl.edu/tinker/
- Tinker-HP (high performance implementation of Tinker, MPI/OpenMP) http://www.ip2ct.upmc.fr/tinkerHP/
- Tinker-GPU (GPU version of Tinker) http://biomol.bme.utexas.edu/tinkergpu
The interdisciplinary aspect of the workshop allowed sharing advances between communities. Discussions about the implementation within the Tinker codes of various novel and mathematically sound algorithms took place. Such new methods ranged from polarization solvers (iterative, fixed cost/analytic forces with error control, non-iterative/extended Lagrangian) to new efficient implicit polarizable solvation models and advanced sampling techniques. The new context of high performance computing were also highlighted including GPUs, new plateforms such as Xeon-Phi or the perspective of more refined mathematical parallelization strategies, i.e parallelism in time. Finally, the evolution of polarizable force fields models within the Tinker software were discussed.