simtk.org

(This is a response to the questions about custom forces in https://simtk.org/forums/viewtopic.php?f=161&t=4497. I'm responding here to keep the two subjects separate.)

I don't know of any reason for the behavior to have changed. CustomNonbondedForce gained several new features, but all of them are disabled by default.

Can you provide a simple test case that shows the difference? Something with just a few atoms would be ideal, but if it needs to be larger that's ok. A serialized System and State would be great, since then I can be sure I'm exactly reproducing what you're doing.

Peter

Hi Peter,

Thanks for responding. I am not sure I can produce it with a simple test since my OpenMM/python skills are basically non-existent. All of this is occurring through the CHARMM/OpenMM interface. I have forgotten, do you guys have CHARMM? I remember there was a request for a license sometime back. Anyway, I'll need to dig down and see if I can simplify the system enough to make the error occur in a more clear manner. What I know is that the same CHARMM unit test I wrote for this worked/works with the same code w/ OpenMM 5.1 and CUDA 5.0. Will get back when I have more.

Hi there,

We did get a CHARMM license just a few months back.

- Lee-Ping

Unfortunately, I don't know how to use CHARMM! Even if I did, that's a much more difficult environment for me to try to debug the problem in. Can you let CHARMM build the system, then just use XMLSerializer to write it to a file?

Peter

Hi Peter,

We're trying to get the serializer into CHARMM - we need to write a f90 wrapper. Hopefully we can get this together and find the issue.