You are here

Weird results from flexpepdock MPI

4 posts / 0 new
Last post
Weird results from flexpepdock MPI
#1

Hello!

Currentlxz i am running a command like: 

mpiexec -np {0} {1}/source/bin/FlexPepDocking.mpi.linuxgccrelease -database {1}/database -s {2} -ex1 -ex2 -out:file:silent {3}.silent -out:file:silent_struct_type binary -pep_refine -nstruct 100 -unboundrot {4}

 

On 47 cores (so -np 47) with nstruct set to 100. I would expect 100 output energies in the scoing file, however i get 4700. I guess it somehow makes all the 100 simulation on all the nodes. 

 

Is this the expected behaviour? It seems really strange to me. 

Category: 
Post Situation: 
Fri, 2016-06-24 01:18
gerdos

Yup, that's weird!  It should do 100. I'm guessing something is broken with MPI.  What does the MPI executable do if you just try to run it without mpirun?  What MPI are you using?  I assume it compiled just fine.  

Fri, 2016-06-24 07:30
smlewis

 I am using ubuntu 14 lts, and i simply installed openmpi-common with apt-get

Thu, 2016-06-30 00:47
gerdos

What does the MPI executable do if you just try to run it without mpirun?  Does this reproduce with any of the other executables (fixbb with -packing:prevent_repacking is a very fast test).

Thu, 2016-06-30 08:44
smlewis