On Sat, Aug 19, 2017, JiYuan Liu wrote:
>
> I have compiled amber16 with ambertools17 in serial and parallel
> successfully that used mpich2 v3.1.4, but when I performed
> MMPBSA.py.MPI, it displayed the error "could not import mpi4py
> package ! use serial version or instal mpi4py", then I typed
> the command "from mpi4py import MPI", it also displayed a error
> "/amber16/lib/python2.7/site-packages/mpi4py/MPI.so:undefined
> symbol:MPI_File_iread_at_all.
Please see note 2 on p. 25 of the Amber 2017 Reference Manual. What operating
system are you using, and how did you install python? (Did you accept
the offer to install miniconda?) [Developers: our configure_python
script doesn't attempt to get mpi4py. Is there a reason for this? Should
we give better instructions in the manual about why this is, and what to do?]
>
> Btw, I configured the mpich2 with the commmand "./configure CC=icc
> CXX=icpc FC=ifort F77=ifort CFLAGS="-fPIC" CXXFLAGS="-fPIC"
> FFLAGS="-fPIC" --enable-shared --prefix=/usr/local/mpich2".
The python problem has nothing to do with how you installed mpich2.
>
> Another question is how to install amber12 with ambertools13 in the cuda
> version on RHEL7.3, I have tried, but failed since the cuda verion 5.5
> could not be installed on RHEL7.3.
Oooh....why would you even think about installing pmemd.cuda from Amber12,
when you have Amber16 available? It *is* the case that we can't help very
much if you are unable to install the nvidia cuda tools on your machine.
(You don't give any information about why not; but this is not the right
mailing list for that sort of question anyway.)
....dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Aug 18 2017 - 14:00:04 PDT
