Hello,
I am trying to print out the per-atom energies to a dump file using CUDA and the following computes:
compute peratom all pe/atom
compute pe all reduce sum c_peratom
thermo 5
thermo_style custom step pe c_pe
thermo_modify format float %15.14g lost warn
dump geometry all custom 100 Ge.dump id type x y z c_peratom
dump_modify geometry format “%d %d %20.15g %20.15g %20.15g %20.15g”
I believe I have found a few bugs:
BUG 1. The per-atom energies, c_peratom (and hence the sum of the per-atom energies, c_pe) drop to zero after 70 timesteps. (See attachments: Ge.dump and log.lammps)
Here is the screen output:
LAMMPS (5 Mar 2012)
Using LAMMPS_CUDA
USER-CUDA mode is enabled (lammps.cpp:396)
Reading data file …
orthogonal box = (-70 -70 -70) to (70 70 70)
1 by 1 by 1 MPI processor grid
1388 atoms
CUDA: Activate GPU
Using device 0: GeForce GTX 560
CUDA: VerletCuda::setup: Allocate memory on device for maximum of 1388 atoms…
CUDA: Using precision: Global: 8 X: 8 V: 8 F: 8 PPPM: 8
Setting up run …
CUDA: VerletCuda::setup: Upload data…
CUDA: Total Device Memory useage post setup: 213.195312 MB
Memory usage per processor = 1.67601 Mbytes
Step PotEng pe
0 -4852.4009374448 -4852.4009374448
5 -4836.3778281971 -4833.4665176848
10 -4791.4542040333 -4791.4542040333
15 -4727.122805875 -4724.2129102419
20 -4656.4539000095 -4653.5480985489
25 -4596.3535288594 -4596.3535288594
30 -4565.1117098309 -4562.21656814
35 -4568.6075739802 -4565.7206205284
40 -4590.5287545806 -4587.655857678
45 -4617.0157756444 -4617.0157756444
50 -4640.5842013672 -4637.7590592372
55 -4654.586859057 -4651.7961199272
60 -4654.8542914239 -4652.103213556
65 -4641.0967615527 -4638.3873700257
70 -4616.6632862688 -4613.9934610603
75 -4588.099067754 0
80 -4564.1069350281 0
85 -4552.6996716958 0
Note: This problem goes away when (1) setting “thermo” to any value less than 5, or (2) choosing a different seed for the “velocity” command.
BUG 2. In some cases the values for “pe” and “c_pe” differ by ~3 eV (See, for example, timestep 5 above.)
Also, LAMMPS slows down drastically when using the above compute.
I am attaching my input script (Ge.in) and geo file (Ge.geo). Please let me know if you need any additional information.
Best regards,
– Willie
Ge.geo (82.9 KB)
in.Ge (619 Bytes)
Ge.dump (1.33 MB)
log.lammps (9.84 KB)