From: Peter Freddolino (petefred_at_ks.uiuc.edu)
Date: Thu Nov 15 2007 - 00:17:01 CST
Hi Vlad,
could you try using catdcd on one of the corrupted dcds to see whether
the whole file is corrupted or just the last frame? It should throw an
error if you're reading bad sections of the file, so by specifying -last
to be one before the end you can see where problems occurred. I haven't
been able to reproduce this on my setup, though... has anyone else on
the list seen anything like this?
Peter
Vlad Cojocaru wrote:
> Dear Peter,
>
> I am running on our internal x86_64 Linux cluster (2*Dual Core AMD
> opteron 2218, 4 procs per node, each proc 2600 Mhz), on 4 (1 node) or
> 8 procs (2 nodes).
> The NAMD version is: NAMD_2.6_Linux-amd64-TCP
>
> There is absolutely no error message (maybe next run I could run it
> with strace and see if I can find out more) ... The job looks as if it
> finished normally ... All output is complete (well except for the dcd
> file) ....
>
> The problem appears either with SMD or TclForces enabled and it is
> independent on whether the job runs on 1 node or on 2 different nodes
> .. The problem doesnt appear neither with standard simulations, nor
> with ABF simulations.
>
> It is not a dramatic issue since the .bak file is OK but I was just
> wondering whether anybody else noticed such a problem ...
>
> Cheers
> vlad
>
> Peter Freddolino wrote:
>
>> Hi Vlad,
>> I certainly have not noticed this behavior in any of the SMD
>> trajectories that I've run recently. Could you give a little more
>> information on where you're running? Also, could you check whether your
>> jobs are ending in the middle of a write to the dcd? The dcd.bak file
>> should just contain the same information as the dcd from the previous
>> timestep...
>>
>> Best,
>> Peter
>>
>> Vlad Cojocaru wrote:
>>
>>
>>> Dear NAMD community,
>>>
>>> I noticed that the DCD trajectories produced by SMD simulation in NAMD
>>> 2.6 are corrupted (they cannot be loaded properly in VMD 1.8.6).
>>> Luckily, NAMD creates also a DCD.BAK backup which can be properly
>>> visualized. This problem is independednt of the trajectory size.
>>>
>>> Is there a fix for this problem ?
>>>
>>> Thanks
>>>
>>> Best wishes
>>> vlad
>>>
>>>
>>>
>>>
>>
>>
>>
>
This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:45:32 CST