Running errors of LAMMPS GPU bench test

Dear All,

I'm trying to obtain my first experience of the GPU version of lammps.
My OS is OpenSuSe 12.2, and I installed the driver and cuda toolkit for
the GPU telsa C2075. Then I successfully compiled the lammps source code
with MPICH2 and FFTW3 but without the gpu package, which demonstrated
the current configuration of the Makefile.g++ is ok. After that I
compiled the gpu library /lib/gpu/, and accordingly added gpu_INC,
gpu_SYSLIB and gpu_SYSPATH into the Makefile.g++. Finally I re-compiled
the lammps with make yes-gpu, make g++, and I successfully got the
executable file lmp_g++.

Everything looks fine. When I tried to test the bench, however,
following errors appeared:
[email protected]:~/soft/lammps-24Apr13/bench/GPU> mpirun -np 8 lmp_g++
-sf gpu -c off -v g 2 -v x 32 -v y 32 -v z 64 -v t 100 < in.lj.gpu
[[email protected]] control_cb (./pm/pmiserv/pmiserv_cb.c:202): assert
(!closed) failed
[[email protected]] HYDT_dmxu_poll_wait_for_event
(./tools/demux/demux_poll.c:77): callback returned error status
[[email protected]] HYD_pmci_wait_for_completion
(./pm/pmiserv/pmiserv_pmci.c:197): error waiting for event
[[email protected]] main (./ui/mpich/mpiexec.c:330): process manager
error waiting for completion

Could anybody tell me the origin for this kind error? How should I work
to solve it? Any suggestion is highly apprecaited!

Best regards,
Junjie

Dear All,

I'm trying to obtain my first experience of the GPU version of lammps.
My OS is OpenSuSe 12.2, and I installed the driver and cuda toolkit for
the GPU telsa C2075. Then I successfully compiled the lammps source code
with MPICH2 and FFTW3 but without the gpu package, which demonstrated
the current configuration of the Makefile.g++ is ok. After that I

can you also run successful simulations with it?

compiled the gpu library /lib/gpu/, and accordingly added gpu_INC,
gpu_SYSLIB and gpu_SYSPATH into the Makefile.g++. Finally I re-compiled
the lammps with make yes-gpu, make g++, and I successfully got the
executable file lmp_g++.

Everything looks fine. When I tried to test the bench, however,
following errors appeared:
[email protected]:~/soft/lammps-24Apr13/bench/GPU> mpirun -np 8 lmp_g++
-sf gpu -c off -v g 2 -v x 32 -v y 32 -v z 64 -v t 100 < in.lj.gpu

[[email protected]] control_cb (./pm/pmiserv/pmiserv_cb.c:202): assert
(!closed) failed
[[email protected]] HYDT_dmxu_poll_wait_for_event
(./tools/demux/demux_poll.c:77): callback returned error status
[[email protected]] HYD_pmci_wait_for_completion
(./pm/pmiserv/pmiserv_pmci.c:197): error waiting for event
[[email protected]] main (./ui/mpich/mpiexec.c:330): process manager
error waiting for completion

Could anybody tell me the origin for this kind error? How should I work
to solve it? Any suggestion is highly apprecaited!

those error messages seem to originate from the MPI library. please
test if you can run MPI parallel calculations w/o gpu acceleration
first.
pending success in that, please run with -echo screen and report the
full output leading up to this error messages.

axel.