Hello,
This is probably due to the fact that in the past, all minimizations were
done using sander and the default conjugate gradient minimizer. Nmode
calculations for MMPBSA.py (and mm_pbsa.pl now) are done with a nab program
that uses the xmin minimizer. This minimizer is much better, but also
requires more energy/gradient calculations as I understand it, so it takes
longer than the sander minimization used to.
Also note that only the master thread is reporting the number of threads its
computing, and the master thread is only responsible for 123 frames, so it's
on frame 51 of 123.
One thing that you may be able to do to lessen the amount of time spent
minimizing is to run your trajectory through sander's minimizer (use imin=5
and maxcyc=50000 or something) to use the normal sander minimizer, then use
that trajectory with MMPBSA.py's nmode module.
You should also make sure that the normal mode calculation is actually
taking place (for instance, try with just a single frame), because with 4
threads running you may be getting low on memory, and the calculation quits
after the minimization. This will help save some time if the calculations
have failed after the minimization.
Good luck!
Jason
On Mon, Jan 3, 2011 at 5:21 AM, mish <smncbr.gmail.com> wrote:
> Hi,
>
> I have a little bit confusion about the entropy calculation using perl
> script and MMPBSA.py.MPI
> I did some calculation in the past using perl script in just 1 CPU and
> calculation took about 30 days on 1 CPU where I re-submitted all those
> calculation using MMPBSA.py.MPI in 4CPU (4 threads) and the progress I can
> see is much slower than perl. I am not sure about the iterations used in
> default input file for sander ( all such files have been deleted from perl
> calculation), By default, are we using more cycles in python script than
> perl ? After 11 days of calculation for the same number of snapshots (500)
> I
> can see
>
> Starting nmode calculations...
> master thread is calculating 123 frames
>
> calculating complex for frame number 0
> *
> *
> up to
> calculating complex for frame number 51
>
>
> If it will run for 500 snapshots like this then I suppose it will take more
> time than one CPU ? Can anyone let me know the reason, why is it taking
> more
> time ?
>
> ..Sincerely
> mish
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
--
Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Graduate Student
352-392-4032
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Jan 03 2011 - 15:30:02 PST
