[gmx-users] Question about use of tpbconv

[Mark Abraham]

Joanne Hanna wrote:
> Hi
> 
> I have a quick question. When you use tpbconv to continue runs what exactly is taken from the *.edr and *.trr files? For instance if I have to run my simulation in short batches of 1ns is it sufficient to just use the run3 files to continue the run (2ns-3ns) or should thr run1, run2 and run3 files by concatenated and then a run4 file generated using the complete edr and trr files.

Exactly what gets taken will vary a bit with the kind of simulation, but 
only information from the last step is pertinent, and only that gets 
used. In a very real sense, with positions and momenta and Newton's Laws 
known, the system doesn't need "memory" in order to determine the next 
time step. You can verify that by starting a simulation with "run3" and 
another with a concatenation of runs 1-3. Unless the concatenation has 
involved some lossy conversion of numbers from one format to another, 
the final step will be identical and the subsequent simulation identical.

Multiple time-step integration would be a slightly different story, of 
course.

Mark