From: brian.radak_at_gmail.com
Date: Sat Mar 12 2022 - 13:05:44 CST
Hi Atanu,
This does not sound like intended behavior. Can you say a little bit
more about what kind of settings you are using (i.e. the specific number
of MD/MC cycles and the value of cphRestartFreq)? One possibility is
that the Tcl I/O buffer is not getting flushed even though the file is
being closed -- that could behavior could be system dependent.
In general I would highly recommend calibrating your settings so that
NAMD completes normally in the allowed walltime. Like most MD codes,
this has not been aggressively tested for other scenarios. A good way to
get a conservative estimate is to estimate the total wall time from the
benchmark ns/day, adding ~30% to the cost for the longest possible
switch run at each cycle.
wall time in hrs = [<numsteps> + 1.3*max(<numsteps per switch>)] *
(<time step in ps> / 1000 ps/ns) * ( 24 hrs/day / (ns/day) )
Many pH values will probably complete faster when fewer residues
titrate, but your overall throughput will be much faster due to
stability (and maybe even accrued priority, depending on the queue).
HTH,
Brian
On 3/11/22 5:52 PM, Acharya, Atanu wrote:
> Hello everyone,
> I am running constant pH simulations using NAMD 2.13. Because of a
> relatively small walltime enforced by our supercomputer, I needed to
> restart the constant pH simulation from incomplete runs. However, NAMD
> is not writing the cphrst files needed to restart a run. I have also
> tried to change the **cphRestartFreq* *without any luck. **
> This question was posted a few years ago. Is there a solution to it?
>
> Thank you so much for looking into it.
>
> Best,
> Atanu
This archive was generated by hypermail 2.1.6 : Tue Dec 13 2022 - 14:32:44 CST