Skip to content

OpenMM 7.5.0

Choose a tag to compare

@peastman peastman released this 09 Dec 23:39
· 621 commits to master since this release
b49b82e

A major focus of this release is on integrators. There are several new ones:

  • LangevinMiddleIntegrator is mostly a drop-in replacement for LangevinIntegrator, but it uses the LF-Middle discretization of the Langevin equation. This provides substantially lower error, so you can often use a step size roughly twice as large with no loss in accuracy.

  • NoseHooverIntegrator is an alternative that uses a Nose-Hoover thermostat instead of a Langevin thermostat. A particularly useful feature of this integrator is that it lets you couple different parts of the system to different thermostats.

  • DrudeNoseHooverIntegrator is a version of it for use with Drude particles.

  • MTSLangevinIntegrator is a multiple time step integrator with built in thermostat.

There are lots of other new features too. Here are some of the more significant ones:

The most recent (2019) version of the CHARMM Drude force field is now included.

When building membranes, it now offers several new choices for the type of lipid to use.

All integrators have a new setIntegrationForceGroups() option to make them use only a subset of forces when integrating the equations of motion. This is useful when you want to define other forces that can be efficiently queried, but shouldn't directly affect the simulation.

Tabulated functions for use in custom forces can now be periodic.

AmoebaVdwForce has lots of new features, including support for alchemical modifications, the ability to override parameters for particular pairs of particles, and the option to use Lennard-Jones 12-6 instead of buffered 14-7.

There also is a major change that isn't directly visible: we did a refactoring of the OpenCL and CUDA platforms to unify their implementations and eliminate nearly-identical code between them. This eliminated roughly 13,000 lines of code, which will be a big help to ongoing maintenance. It also has an important benefit to anyone writing plugins: in most cases, you can now write a single implementation that works on both CUDA and OpenCL, rather than having to implement each of them separately.

Installation instructions are in the User Guide.