Hello,
On Mon, Feb 1, 2010 at 6:43 AM, s. Bill <s_bill36.yahoo.co.uk> wrote:
> Dear AMBER
> I was wonder if there is any one has a parrallel version of MMPBSA.pl.
> I know the new MMPBSA.py is parrallel, but it is just restricted to 3 processors (2 in real worl, where ligand is so small).
As Bill said, the version available on the web will accept 3
processors, or any even number of processors (for instance, 4
processors upon which 2 will be dedicated to the complex and 2 to the
receptor while the ligand is calculated separately later since it's
generally small). However, this only parallelizes the sander energy
calculations done by sander, as the script itself is not parallelized.
Thus, since PB calculations in sander do not run in parallel, so the
only way to effectively use more than 2 processors for PB calculations
is to start several MMPBSA calculations on a subset of your total
number of frames (and only MMPBSA.py will allow you to use 2
processors in a single run). (GB parallelizes well, and using 4
processors in MMPBSA.py shows very close to double the performance of
using 2) Nmode calculations will not parallelize at all in either
version, so if this is done, your only choice is to follow Bill's
advice of splitting your trajectory between MMPBSA calculations.
> Also, how can I restart the job? my jobs have been killed because of the wall clock time, so I would like to continue my jobs rather than to start it from scratch again. How can I do it?
No way of doing that for MMPBSA.py, and I think it is very tricky with
mm_pbsa.pl if it's possible at all. However, the cost of an
MM/PB(GB)SA calculation should be considerably cheaper than the cost
of the initial dynamics used to create the snapshots. Thus, allow
yourself a sufficiently large wallclock time to let the calculation
finish.
Good luck!
Jason
--
---------------------------------------
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 Feb 01 2010 - 05:00:02 PST
