[Dock-fans] Errors in running DOCK6 MPI

Scott Brozell sbrozell at rci.rutgers.edu
Tue Jun 14 22:35:33 PDT 2016


Hi,

You are seeing several strange and/or clearly erroneous things.
It is likely that the cause(s) are due to basic mistakes probably
during the preparatory steps.  I recommend that you start over,
review the tutorials, modify those inputs to produce your docking 
inputs, and carefully examine and visualize the results at each step.

The purpose of parallel dock is to facilitate calculations with
large data sets.  You should be confident of the results of several
serial docking calculations before trying parallel dock.

As indicated by Julio and as elaborated in detail here
http://dock.compbio.ucsf.edu/DOCK_6/dock6_manual.htm#ParallelProcessing
parallelization is over ligands and the method is simple.  
It makes no sense that parallel dock is slower than serial dock.
Ignore this for now and concentrate on my first paragraph.

scott

On Tue, Jun 14, 2016 at 09:10:56PM +0000, Julio Dominguez wrote:
> >From the manual:
> If the HA_RMSDs is "-1000.0", then there is an inconsistency in the number of heavy atoms between the reference and the docked conformer.
> But as to the difference for using the two different .in I don't know. However, it might be related to the core use. How many ligands are you docking?As far as I remember, running dock in mpi only improves the docking of several ligands at once, not the performance of one.
> That being said, I have seen very different performances when running dock in AMD vs Intel.
> I hope this helps.
> 
> > From: thanh.q.le at sjsu.edu
> > Date: Tue, 14 Jun 2016 14:00:47 -0700
> > To: dock-fans at docking.org
> > Subject: Re: [Dock-fans] Errors in running  DOCK6 MPI
> > 
> > I forgot to mention one thing, this error can only be seen while using DOCK6 MPI fix.in; regular dock6 flx.in does give the desired numbers.
> > Thanh Le
> > > On Jun 14, 2016, at 1:53 PM, Thanh Le <thanh.q.le at sjsu.edu> wrote:
> > > 
> > > Hi,
> > > The result I received after running dock6 MPI flx.in seems really off:
> > > ########## HA_RMSDs:		-1000.0000
> > > ########## HA_RMSDh:		-1000.0000
> > > ########## HA_RMSDm:		0.0000
> > > ########## Cluster Size:	1
> > > ##########    Grid Score:         -246.470337
> > > ##########      Grid_vdw:         -111.281166
> > > ##########       Grid_es:         -135.189178
> > > ##########    Int_energy:           52.528397
> > > I would like to know why I got -1000.0000 for both HA_RMSDs and HA_RMSDm and 0 for HA_RMSDm.
> > > Also, when using dock6 mpi on a cluster (64 cores), the program runs slower than a single core-must be more than 10 times slower.


More information about the Dock-fans mailing list