Const Qty & EXT_RESTART

Rolf David rolf.d... at gmail.com
Fri Mar 27 17:36:51 UTC 2015


Ok i'll try this with no change, everything is labelled the same : project 
name/path/relative path, on the wierd system, still huge drift.

I have same problem with another system (good this time) : 

I've done a QM/MM-dynamics starting from a MM-dynamics. 
I've done 1000 steps of 0.5 fs (project 0-1000). I've restarted successfuly 
for another 1000 steps (project 1000-2000 with last restart/wfn from 
0-1000) (Restart default TRUE). And again every 1000 steps, I'm around step 
6000, no problem so far (perfect !).

I also dump a restart/scf history every 50 steps, but when I wanted to 
restart from a file in between (like I restarted from the 750th step, with 
the 1_750.restart and 1_750.wfn, "in the middle of a CP2K run", and not 
from files of the 1000th step, "at the ending of the CP2K run"), I have the 
problem with the Const Qty.

I haven't look at the restarts ("end" one and "middle" one) to see if there 
is a difference here. I'll do that and update here.

So update: I'll find nothing. Everything is pretty much the same except 
coord, velocities, forces, averages, etc.. exactly what I expected to 
change. Now i'll try to start other points, 1_800 for ex, and restart the 
1000.restart one. Because I transfered the file from the HPC to a raid 
storage, but since it's a "ASCI" file it shouldn't have an impact but I'm 
out of idea.

In the meantime, if someone has an idea, I'll take it.

Rolf

On Sunday, March 22, 2015 at 8:15:09 AM UTC+1, Rolf David wrote:
>
> I'll try that just in case
>
> I've tried the same with other system, the "restart transition"  is very 
> smooth so i deducted there was something wrong with my systems and the 
> restart file cannot handle the "exoticity" of the system. But without a 
> 'faulty' system the transition is like it the same calculation all along 
> not a lot divided by restart.
>
> So no problems with CP2K here.
>
>
> On Friday, March 20, 2015 at 8:24:20 AM UTC+1, Samuel Andermatt wrote:
>>
>> Could you try to do it without changing the project name? I am not 
>> familiar with QM/MM but if any other data is needed for the restart except 
>> the run_0-1000.restart file cp2k will probably look for this file under 
>> a path that contains the current project name.
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20150327/d0dd1395/attachment.htm>


More information about the CP2K-user mailing list