Scaling molecular dynamics beyond 100,000 processor cores for large-scale biophysical simulations

Area of Science:

• Computational biology
• Biophysics
• Molecular dynamics simulations

Background:

• Increasing system size in biophysical simulations is crucial for understanding complex biological interactions.
• Existing software requires significant hardware and adaptable solutions for large-scale simulations.

Purpose of the Study:

• To enhance the GENESIS molecular dynamics package for large processor systems.
• To enable efficient simulation of ultra-large biomolecular systems.

Main Methods:

• Parallelization of long-range electrostatic interactions to minimize communication.
• Implementation of a novel algorithm for nonbonded interactions to boost SIMD performance.
• Reduction of memory usage for neighbor searches by approximately 80%.

Main Results:

• Development of strategies for large-scale simulations within the GENESIS package.
• Creation of the first atomistic model of the GATA4 gene locus using experimental data.
• Successful execution of the first billion-atom simulation of an intact biomolecular complex.

Conclusions:

• Optimized GENESIS package enables unprecedented scale in molecular dynamics.
• Achieved 1 ns/day performance scaling to 65,000 processes.
• Paves the way for simulating complex biological systems at the atomistic level.