[Dock-fans] dock6.mpi is running in serial on 48 CPUs?

Jonathan Gough jonathan.d.gough at gmail.com
Thu Mar 24 07:17:17 PDT 2016


There must have been something wrong with the mpich install. I sort of
cheated by re-compiling mpich using the amber mpich install helper script.
(you essentially unpack the tarball into the AmberTools/src directory and
then execute their install/config script) which installed mpich into the
amber14/bin path. Then compiled dock6.mpi against that. When I then
launched dock6.mpi It worked efficiently. Another sign that it worked was
that it spawned the extra screen.out.# out files.



> Does your dock output file contain, eg:
> Parallel dock running 48 MPI processes
> ?
>

This was a very helpful clue, thank you Scott!

Oddly, the machine was indeed launching 48 dock6.mpi processes, but the log
file and the calculations did not reflect that.



>
> How many ligands are in the file specified by ligand_atom_file ?
>

I was using 100


> What happens when you run the mpi test:
> cd dock6/install/test; make parallel
> ?
>
>
I got:
-bash-4.2$ make parallel
MPICH_HOME is defined to /home/jgough/dock6/mpich-install
cd mpi && make test
make[1]: Entering directory `/home/jgough/dock6/install/test/mpi'
make[2]: Entering directory `/home/jgough/dock6/install/test/mpi'
cd ../grid_generation && make test
make[3]: Entering directory
`/home/jgough/dock6/install/test/grid_generation'
# Construct box to enclose spheres
../../../bin/showbox < box.in > /dev/null
../dockdif box.pdb.save box.pdb
diffing box.pdb.save with box.pdb
PASSED
==============================================================
# Compute scoring grids
../../../bin/grid -i grid.in -o grid.out
../dockdif grid.out.save grid.out
diffing grid.out.save with grid.out
PASSED
==============================================================
make[3]: Leaving directory `/home/jgough/dock6/install/test/grid_generation'
make[2]: Leaving directory `/home/jgough/dock6/install/test/mpi'
make[2]: Entering directory `/home/jgough/dock6/install/test/mpi'

Processing parallel dock6 test mpi
/home/jgough/dock6/mpich-install/bin/mpirun -np 2 ../../../bin/dock6.mpi -i
mpi.dockin -o mpi.dockmpiout
../dockdif -t 8 mpi.dockmpiout.save mpi.dockmpiout
diffing mpi.dockmpiout.save with mpi.dockmpiout
possible FAILURE:  check mpi.dockmpiout.dif
==============================================================
diffing mpi_ranked.mol2.save with mpi_ranked.mol2
PASSED
==============================================================
make[2]: Leaving directory `/home/jgough/dock6/install/test/mpi'
make[1]: Leaving directory `/home/jgough/dock6/install/test/mpi'

Contents of mpi.dockmpiout.dif:

6,7d5
< Parallel dock running 2 MPI processes
< --------------------------------------
102a101
>
------------------------------------------------------------------------------------------
111a111,115
> Molecule: ZINC00158442
>  ERROR:  Could not complete growth.
>          Confirm that the grid box is large enough to contain the ligand,
>          and try increasing max_orientations.
> -----------------------------------
200d203
< Finalizing MPI Routines...




> If you need more help then please send verbose and verbatim details.
>
> scott
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.docking.org/pipermail/dock-fans/attachments/20160324/d2649ed6/attachment-0001.html 


More information about the Dock-fans mailing list