[Openmm-betatesters] OpenMM 5.0 beta release

[Peter Eastman]

We are very pleased to announce that a beta release of OpenMM 5.0 is now available.  This is a major update with lots of new features and internal changes.  Here are some of the most important ones.

The old CUDA based platform (which was limited in features and hard to maintain) has been eliminated.  Instead, we've written a completely new CUDA based platform modeled after the OpenCL platform.  It supports all the same features as the OpenCL platform, while giving significantly better performance on recent Nvidia processors (Fermi and later).

The CUDA and OpenCL platforms now offer a choice of numerical precision modes.  You can do all calculations in either single or double precision.  There also is a mixed precision mode that gives much better energy conservation than single precision at only a small cost in performance.

There is a new CustomCompoundBondForce class.  It's a very powerful tool for implementing arbitrary bonded interactions.  A "bond" may involve an arbitrary number of atoms, and depend in a user defined way on atomic positions, interatomic distances, angles, etc.

Most Forces now have an updateParametersInContext() method that lets you modify them in an existing Context without having to reinitialize it.  This makes it both easier and more efficient to implement a variety of simulation protocols where, for example, per-particle parameters need to change over the course of the simulation.

There is a new checkpointing feature that lets you save the complete internal state of a simulation to disk, then reload it to continue exactly where it left off.

It can now read Gromacs GRO files.

We also have fixed lots of bugs.  We haven't put together a list of them yet, but we'll do that before it's released.

One other important change: the free energy plugin has been eliminated.  It was never supported by the OpenCL platform, and custom forces are a much more flexible tool for doing the same things.  If you were using the free energy plugin, and you need help converting over to custom forces, please contact us.

Now we need lots of people to start using the release on as many OS/hardware combinations as possible to catch any remaining problems before it goes final.  You can get it from the  OpenMM download page at https://simtk.org/project/xml/downloads.xml?group_id=161.  Scroll down to the bottom of the section for 4.1.1 and click "List all hidden releases".  The link will only show up if you're logged in to your simtk.org account and you are a member of the OpenMM project.  If you aren't, please email me your simtk.org user name and I'll add you.

One other note: we aren't finished updating the documentation yet.  The User Guide is mostly up to date, but there are still some changes left to make, especially in the validation and AMOEBA chapters.  The Developer Guide hasn't been updated at all yet.  We'll post new versions of those once they are finished.

Thank you all for your help in testing OpenMM and making it a great tool!

Peter