Hi Viktor,
Thanks for the advice! I've tried compiling PLUMED2.5 with MPI support on one of our GPU clusters. The install recipe I have done is
./configure --prefix=/usr --enable-mpi --program-suffix=MPI
make -j 4
make install
I've done what you've suggested, in that I add the following line to my .bashrc file:
export PLUMED_KERNEL=/usr/lib/libplumedMPIKernel.so
However, I still get the same error where it says that plumed has been compiled without MPI support. Do you have any ideas or other flags I can try to rectify this?
Best wishes
Amanda
________________________________
From: viktor drobot <linux776.gmail.com>
Sent: 10 August 2020 17:25
To: AMBER Mailing List <amber.ambermd.org>
Subject: Re: [AMBER] Amber20 with PLUMED not working with multiple CPUs
Hello. Be sure that you have compiler PLUMED with MPI support. Usually all
that you need is to specify --enable-mpi flag during configure. To have
both serial and MPI versions of PLUMED installed on your system you can use
--program-suffix option.
So when you need to run *.MPI versions of sander/pmemd you should export
path of MPI-versioned PLUMED kernel like this:
export PLUMED_KERNEL=/usr/lib/libplumed-mpiKernel.so (your will be
different)
and for serial versions please export path of no-MPI PLUMED kernel:
export PLUMED_KERNEL=/usr/lib/libplumedKernel.so
73, Viktor
ÐÎ, 10 Á×Ç. 2020 Ç., 2:25 Amanda Buyan <Amanda.Buyan.anu.edu.au>:
> Hi everyone,
>
>
> I'm unsure if the last email I got went through, and I'm still having
> trouble with the same problem.
>
>
> I've managed to compile Amber20 on our GPU cluster, and am running
> equilibration simulations. When I minimize using sander, running with
> PLUMED works just fine. However, when I try to run an equilibrium
> simulation using pmemd.MPI, the program asks me to specify the PLUMED
> kernel. When I do, it then says that PLUMED isn't available for MPI. Has
> anyone had this particular error before? How would I go about fixing it?
> I've tried both enabling and disabling the PLUMED library when I compile
> pmemd.MPI, but that doesn't seem to do anything either....
>
>
> Any advice would be very helpful!
>
>
> Best wishes
>
>
> Amanda
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Aug 10 2020 - 21:30:02 PDT
