[gmx-users] Wrong pressure with "mdrun -rerun" option

[Mark Abraham]

Lanyuan Lu wrote:
> Hi Mark,
> Thanks for your reply. Yes, I did save velocities in the same frequency 
> as positions. And the length of two compared trajectories are exactly 
> the same. I think I've figured it out and I did looked at the average 
> values as the output of the g_energy. It seems the number 545.384 from 
> the original MD edr file is incorrect. I opened the xvg file in xmgrace 
> and found the average pressure should be 1153.22, which is close to the 
> rerun result. I used the command "g_energy -f *.edr -o *.xvg" to dump 
> the pressure from two edr files and found both of the two xvg files have 
> time from 0 ps to 200 ps with 201 frames. I also found for each time 
> step the pressure values are close for the two files. Therefore, I don't 
> understand why I got the incorrect average pressure from the original 
> edr output, while the rerun output was correct. Another test I did was 
> to type "g_energy -f *.edr -o *.xvg -b 100" to get the last 100 ps 
> results. And this time the average number from the original edr file is 
> still incorrect. However, just as in the whole 200 ps case, I can get 
> the right number if I manually calculate the average from the xvg file.
> Lanyuan

Hence...

>> Did your primary mdrun include an
>> equilibration period that the rerun did not?

See appendix C.2. The averages reported by g_energy are not necessarily
the averages of the data in the file. You can see whether this is
happening by comparing a gmxdump of the two .edr files, or using the
gmxcheck facility for comparing them. Assuming that your original mdrun
was pre-equilibrated, I expect you will see the partial sums are
different at each step - and the first one in particular.

Mark