Closed gitkol closed 2 years ago
As I know, PLUMED assumes a constant timestep, so the variable timestep integrators are incompatible.
We need to implement a check in OpenMM-PLUMED, so it fails with a clear error message.
Thanks, @raimis. I was afraid of that. Just wondering, couldn't the plugin provide the then current simulation time to PLUMED at every time step? In a number of biasing methods the time step is irrelevant and the simulation time is only used for plotting the time series of bias values, but I don't think it is a requirement that the time step is constant. Of course, I may be wrong.
Istvan
OpenMM passes to PLUMED the time step during the initialization and at each step updates the number of steps. PLUMED API doesn't support passing the simulation time (https://www.plumed.org/doc-v2.7/developer-doc/html/_how_to_plumed_your_m_d.html).
Got it, thanks!
Hi, I am trying to run OpenMM-PLUMED with the VariableLangevinIntegrator, but even though PLUMED seems to work (it uses CPU cycles), it does not print any bias log data, and it is hard to tell if whatever it calculates makes sense. I guess, the problem is that no time step information is given to PLUMED, see copy of the PLUMED output below.
I wonder, does this mean that I cannot use VariableLangevinIntegrator?
Thank you very much,
Istvan