MiMiC: A high-performance framework for multiscale molecular dynamics simulations

Area of Science:

• Computational Chemistry
• Molecular Modeling
• High-Performance Computing

Background:

• Multiscale simulations are crucial for modeling complex systems, but integrating diverse computational approaches remains challenging.
• Existing frameworks often interfere with the native parallelization of underlying programs, limiting flexibility and performance.

Purpose of the Study:

• To introduce MiMiC, a novel, interoperable framework for multiscale simulations.
• To design a system that facilitates efficient data exchange between loosely coupled programs at different resolutions.
• To maintain the original setup and execution of external programs, including hybrid CPU/GPU architectures.

Main Methods:

• MiMiC employs a multiple-program multiple-data (MPMD) paradigm for efficient subsystem interaction.
• It acts as an intermediary, managing data exchange without altering external program parallelization.
• The framework currently implements electrostatic embedding quantum mechanics/molecular mechanics (QM/MM) using CPMD and GROMACS.

Main Results:

• Demonstrated unprecedented parallel scaling for large biomolecular simulations using QM/MM.
• Achieved high efficiency and flexibility by minimizing interference with external programs.
• Validated operability on hybrid CPU/GPU architectures.

Conclusions:

• MiMiC provides a flexible and efficient solution for advanced multiscale simulations.
• The framework is designed for easy extension beyond QM/MM to accommodate general multiscale models.
• MiMiC is well-suited for the demands of exascale high-performance computing.