Hi all,
Today I’ve received a new computer to run my LAMMPS simulations on. This beauty has a Intel Xeon E5-2609 v2 quadcore CPU, but the cores are split over two independent chips (2 cores each). When I try to configure LAMMPS to utilise all 4 cores (command “processors 2 2 1”) I get an error: “ERROR: Specified processors != physical processors”. I assume this error is raised because a single chip only has 2 cores, but so far I couldn’t figure out how to include the other 2 cores as well. Can anybody tell me what I’m supposed to do to get this working?
(I’m using the 2 Apr 2013 version of LAMMPS)
Regards,
Martijn van den Ende
Hi all,
Today I've received a new computer to run my LAMMPS simulations on. This
beauty has a Intel Xeon E5-2609 v2 quadcore CPU, but the cores are split
over two independent chips (2 cores each). When I try to configure LAMMPS to
utilise all 4 cores (command "processors 2 2 1") I get an error: "ERROR:
Specified processors != physical processors". I assume this error is raised
Not likely. Have you checked your input script if there is a
"processor" command defined?
Ray
Hi all,
Today I've received a new computer to run my LAMMPS simulations on. This
beauty has a Intel Xeon E5-2609 v2 quadcore CPU, but the cores are split
over two independent chips (2 cores each). When I try to configure LAMMPS to
utilise all 4 cores (command "processors 2 2 1") I get an error: "ERROR:
no. this is not what the processors command does. please read the
documentation more carefully. the processors command only determines
how to *map* the processors to the x-, y-, and z- dimension. the total
number of processors is specified via the mpirun or mpiexec command.
of course, for that you first have to have LAMMPS compiled with MPI
support (or install a suitable pre-compiled version).
Specified processors != physical processors". I assume this error is raised
because a single chip only has 2 cores, but so far I couldn't figure out how
no.
to include the other 2 cores as well. Can anybody tell me what I'm supposed
to do to get this working?
have a look at the documentation and (perhaps) at an MPI tutorial.
axel.
Thanks for your reply, I found the problem. When I still had 2 cores available and I wanted to run LAMMPS, I invoked the command "mpirun -np 2 lmp_openmpi < scriptfile.in". But since I didn't want to type that entire line each time, I made an alias for that. On my new computer I adjusted the alias in my .bashrc file to use 4 cores, but I forgot to source it before running the code again... (awkward)
It works now on all cores, thanks for your help.
Martijn
Thanks for your reply, I found the problem. When I still had 2 cores
available and I wanted to run LAMMPS, I invoked the command "mpirun -np 2
lmp_openmpi < scriptfile.in". But since I didn't want to type that entire
line each time, I made an alias for that. On my new computer I adjusted the
alias in my .bashrc file to use 4 cores, but I forgot to source it before
running the code again... (awkward)
not awkward at all.
however, it shows the problem with using aliases (you forget the real command).
that is why i prefer to put a shell script called ./run.sh into my
working directories when i run interactive production calculations.
this way i also save effort (even more than with an alias) *and* have
a record of how i ran the code.
axel.