[lammps-users] lammps failing to run on gpu

LAMMPS LAMMPS Mailing List Mirror
1

Dear Users,

I’m trying to run LAMMPS (1 Apr 2010) on a GPU (NVIDIA Tesla C2050) based RHEL-5.3 LInux System. I’ve followed the steps from:

http://lammps.sandia.gov/workshops/Feb10/Mike_Brown/gpu_tut.pdf

It got installed successfully.

But the sample input file(as mentioned in pdf) supporting gpu is failing with following error:

cat gpu_test_lj_out

lmp_openmpi: pair_gpu_cell.cu:486: void build_cell_list(double*, int*, cell_list&, int, int, int, int, int, int, int, int): Assertion `err == cudaSuccess’ failed.
LAMMPS (7 Apr 2010)
Lattice spacing in x,y,z = 1.6796 1.6796 1.6796
Created orthogonal box = (0 0 0) to (33.5919 33.5919 33.5919)
1 by 1 by 1 processor grid
Created 32000 atoms

2

Sangamesh,

We’ve made some recent improvements to the GPU package for LAMMPS that have not yet been migrated into the main repository. Would you mind trying this more recent developmental version? It can be found here:

http://code.google.com/p/gpulammps/

Best wishes,

Paul

3

Dear Paul,

I’ve tried the latest version of LAMMPS through SVN:
But still the problem persist.

…/…/src/lmp_openmpi < in.lj

LAMMPS (5 Jun 2010)
Lattice spacing in x,y,z = 1.6796 1.6796 1.6796
Created orthogonal box = (0 0 0) to (33.5919 33.5919 33.5919)
1 by 1 by 1 processor grid
Created 32000 atoms
ERROR: Illegal pair_style command

It failed in a very initial stage.

Do you have any other version(old), which work successfully with GPU?

Thanks

4

Hi Sangamesh. Thanks for trying the version GPU modules that is on the google codes SVN repository. From the error message you’re reporting it looks like there’s an inconsistency between the input script you are using and the source code. Could you try a different input script? Please see the “tests” folder on the google codes SVN repository. Specifically, could you try the “lj_cut.in” example. That example input script should work with your build.

Paul

5

If you are trying to use code from the google repository, please open doc/Section_start.html in your browser and read about running on GPUs. This has changed significantly. - Mike