Tinker-HP

Tinker-HP accelerates molecular dynamics simulations of large biomolecular systems by providing GPU- and CPU-accelerated, massively parallel implementations of polarizable many-body force fields (AMOEBA) for high-performance computation.


Key Features:

  • GPU Acceleration: Implements single- and multi-GPU execution to accelerate simulations on GPU hardware and complements CPU execution for hybrid workflows.
  • Scalable Strategy: Uses OpenACC and CUDA to achieve scalable performance across diverse GPU platforms, including NVIDIA 2080Ti, 3090, V100, and A100 cards.
  • Multiprecision Capabilities: Supports multiprecision arithmetic with double precision for reference computations and lower-precision modes to improve performance while preserving comparable accuracy.
  • AMOEBA Force Field Implementation: Implements the AMOEBA point dipole polarizable many-body force field for accurate modeling of molecular interactions.
  • Performance Benchmarks: Provides benchmarking results on NVIDIA GPU platforms demonstrating substantial reductions in computation time relative to prior AMOEBA implementations.
  • Strong Scaling and Parallelization: Supports strong-scaling across multiple nodes to enable massive parallelization for large-scale simulations.
  • Adaptive Sampling and Large-Scale Applicability: Integrates unsupervised adaptive sampling techniques to extend applicability to large and complex biophysical systems.

Scientific Applications:

  • Molecular dynamics of large systems: Enables atomistic MD simulations of large biomolecular assemblies using polarizable force fields.
  • Modeling biomolecular interactions: Facilitates accurate modeling of molecular interactions and many-body polarization effects at the atomic level with AMOEBA.
  • High-throughput biophysical studies: Accelerates computational analyses for large-scale research efforts, including investigations related to COVID-19.

Methodology:

Uses OpenACC and CUDA for GPU parallelization; supports single- and multi-GPU execution and multi-node strong-scaling; implements multiprecision arithmetic (double and lower-precision modes); executes the AMOEBA point dipole polarizable force field and incorporates unsupervised adaptive sampling.

Topics

Details

Tool Type:
library
Programming Languages:
C++, Fortran
Added:
12/13/2021
Last Updated:
12/13/2021

Operations

Data Inputs & Outputs

Molecular dynamics

Publications

Adjoua O, Lagardère L, Jolly L, Durocher A, Very T, Dupays I, Wang Z, Inizan TJ, Célerse F, Ren P, Ponder JW, Piquemal J. Tinker-HP: Accelerating Molecular Dynamics Simulations of Large Complex Systems with Advanced Point Dipole Polarizable Force Fields Using GPUs and Multi-GPU Systems. Journal of Chemical Theory and Computation. 2021;17(4):2034-2053. doi:10.1021/acs.jctc.0c01164. PMID:33755446. PMCID:PMC8047816.

PMID: 33755446
PMCID: PMC8047816
Funding: - Agence Nationale de la Recherche: ANR11-IDEX-0004-02 - National Institute of General Medical Sciences: R01GM106137, R01GM114237 - H2020 European Research Council: 810367

Links