Great.
Some of the data at the end of the .log file is easily recomputed from
the output at every exchange attempt once the .log files are
concatenated, but there will be loss of precision in re-computing the
average exchange acceptance probability, etc.
Mark
On Fri, Apr 5, 2013 at 10:37 AM, Joã
Thank you Mark and Francesco. I will be pleased to contribute with a fix in
case I come up with a fairly general solution for this issue.
Best regards,
João Henriques
On Thu, Apr 4, 2013 at 7:45 PM, Mark Abraham wrote:
> Demux.pl pre-dates .cpt and -append. The only solution is to preserve your
Demux.pl pre-dates .cpt and -append. The only solution is to preserve your
.log files (e.g. with -noappend) and post-process. If you do that by
modifying demux.pl, please consider contributing your fix back.
Mark
On Apr 4, 2013 1:42 PM, "francesco oteri" wrote:
> if your -append option is activa
if your -append option is activated (the default is yes),
maybe Demux.pl reads the exchanging from the .log taking into account the
time in the
log and so you don't need to do anything.
But I don't know how Demux.pl works :(
Francesco
2013/4/4 João Henriques
> That's terrible! I was just about
That's terrible! I was just about to restart 2 hefty REMD simulations...
Maybe I can move the original log files somewhere and combine them with the
restart ones afterwards by using a script. It's just an idea, because I
need to run Demux.pl on the final concatenated log file.
Any other issues I s
This is what I meant,
in particular it is a problem when I want to analyze
the data regarding the exchange probability.
Francesco
2013/4/4 João Henriques
> So let me see if I understood what Francesco said correctly. Restarting a
> REMD job after hitting the cluster wall-time limit resets the i
So let me see if I understood what Francesco said correctly. Restarting a
REMD job after hitting the cluster wall-time limit resets the information
stored in the log files? Can someone shed some light on this subject?
Best regards,
João Henriques
--
gmx-users mailing listgmx-users@gromacs.org
This is, unfortunately, normal. The REMD implementation was never updated
when the checkpointing mechanism was introduced, so this data is not
checkpointed.
Mark
On Sat, Mar 16, 2013 at 12:07 PM, francesco oteri wrote:
> Dear gromacs users,
> I ran a REMD on a HPC cluster equipped with a queue
Dear gromacs users,
I ran a REMD on a HPC cluster equipped with a queue system.
For this reason I carried out the simulation dividing it in chuncks.
Because of this, each run started using the .cpt file from the previous one.
The simulation ran fine, but I noticed that the informations regarding
9 matches
Mail list logo