Dear Users and Developers,
I am trying to compile and run LAMMPS with GPU support. I am definitely running into trouble.
When I run my GPU-enabled script, the pair energy decreases to zero and stays there.
$ ./lmp_openmpi < control
LAMMPS (9 Jan 2011)
Scanning data file …
3 = max bonds/atom
5 = max angles/atom
8 = max dihedrals/atom
2 = max impropers/atom
Reading data file …
orthogonal box = (-1253.83 -1.5 -1253.83) to (1253.83 2507.66 1253.83)
1 by 1 by 1 processor grid
1501 atoms
1500 bonds
1999 angles
1996 dihedrals
498 impropers
Finding 1-2 1-3 1-4 neighbors …
3 = max # of 1-2 neighbors
4 = max # of 1-3 neighbors
6 = max # of 1-4 neighbors
9 = max # of special neighbors
Respa levels:
0 = bond
1 = angle dihedral improper
2 = pair kspace
Dear Users and Developers,
I am trying to compile and run LAMMPS with GPU support. I am definitely
running into trouble.
When I run my GPU-enabled script, the pair energy decreases to zero and
stays there.
$ ./lmp_openmpi < control
LAMMPS (9 Jan 2011)
Scanning data file ...
3 = max bonds/atom
5 = max angles/atom
8 = max dihedrals/atom
2 = max impropers/atom
Reading data file ...
orthogonal box = (-1253.83 -1.5 -1253.83) to (1253.83 2507.66 1253.83)
1 by 1 by 1 processor grid
1501 atoms
1500 bonds
1999 angles
1996 dihedrals
498 impropers
Finding 1-2 1-3 1-4 neighbors ...
3 = max # of 1-2 neighbors
4 = max # of 1-3 neighbors
6 = max # of 1-4 neighbors
9 = max # of special neighbors
Respa levels:
0 = bond
1 = angle dihedral improper
2 = pair kspace
--------------------------------------------------------------------------
- Using GPGPU acceleration for lj/cut:
- with 1 procs per device.
--------------------------------------------------------------------------
GPU 0: GeForce GTX 480, 480 cores, 1.3/1.5 GB, 1.4 GHZ (Double Precision)
--------------------------------------------------------------------------
Initializing GPU and compiling on process 0...Done.
Initializing GPU 0 on core 0...Done.
Setting up run ...
Memory usage per processor = 40.1896 Mbytes
Step Temp E_pair E_mol TotEng Press
...
1160 843.44263 11.444882 16224.237 20006.895 0.13522221
1165 1040.918 4.1453628 15360.852 20019.166 0.081567556
1170 990.88831 2.486279 15581.999 20014.96 -0.0144097
1175 808.83646 2.1481825 16381.667 20000.297 -0.063127345
1180 1007.5124 3.4911719 15480.674 19988.969 -0.0053206663
1185 974.46125 0 15580.861 19937.887 0.089925451
1190 847.08967 0 16141.709 19929.229 0.13508306
1195 965.83775 0 15589.649 19908.117 0.098176491
1200 1002.8156 0 15386.294 19870.098 0.0085066658
1205 840.20433 0 16085.715 19842.449 -0.057383891
1210 948.54388 0 15574.756 19815.9 -0.029410394
1215 1043.9377 0 15114.245 19781.915 0.065152949
...
Starting from alternate configurations and initial conditions
1190 135.53036 -0.91854615 1107.9121 1712.9789 0.0018690215
1195 138.02941 -0.92490086 1095.7895 1712.0238 0.0021082402
1200 137.31521 0 1098.0671 1712.0329 -0.00013053597
1205 135.98878 0 1103.0797 1711.1148 -0.0020139212
...
The compilation seems ok - none of the error messages came up.
but there were several warnings of the type:
warning: unused variable ‘fqi’
I have run the in.lj script given in the "bench" directory with the
following modifications:
newton off
pair_style lj/cut/gpu 2.5
fix gpufix1 all gpu force/neigh 0 0 -1
and the pair energy is
Could you help me?
how does the same input behave with the regular lj/cut pair style?
axel.