Same data.file and input script, slightly different results between in Windows and linux, but there is always an error in windows

Dear all:
I have the exactly same data file and input script file, which is a 10000 atoms with random initial configuration, and i apply langevin thermostat to all atoms to control the system temperature. I ran the script both on my desktop(Windows 7) and on a blade server machine(linux).what surprise me is that there is slight difference between results in windows and linux,the thermodynamic data for windows and linux is below.What’s worse, when the simulation ran to a certain time in windows, an error was always encountered. i don’t know why, and now which result should i believe, the windows or linux?or some bug and defect in windows compared with in linux if we run MD simulation…

Step temp temp Poteng Poteng
windows linux windows linux
1000 0.99999 0.99999 -0.00929 -0.00929
2000 1.03309 1.03309 -0.33273 -0.33273
3000 1.02887 1.02865 -0.5399 -0.53891
4000 1.02868 1.02567 -0.71786 -0.70887
5000 1.02945 1.02709 -0.87366 -0.87664
6000 1.01924 1.01242 -1.02001 -1.01556
7000 1.02757 1.02295 -1.13563 -1.14287
8000 1.02361 1.0272 -1.24961 -1.27448
9000 1.01524 1.01038 -1.38932 -1.39967
10000 1.01626 1.01896 -1.51705 -1.53547
11000 1.0196 1.0083 -1.65465 -1.69591
12000 1.01864 1.02179 -1.771 -1.82614
13000 1.02479 1.02234 -1.91406 -1.93274
14000 1.01294 1.02065 -2.0745 -2.08106
15000 1.02142 1.02003 -2.21374 -2.23046
16000 1.02824 1.02266 -2.38336 -2.37435
17000 1.01937 1.01841 -2.55277 -2.52775
18000 1.02453 1.02791 -2.70662 -2.66175
19000 1.01807 1.01493 -2.86115 -2.81889
20000 1.01543 1.024 -2.99956 -2.96433
21000 1.03441 1.02183 -3.13193 -3.10928
22000 1.01062 1.01597 -3.29922 -3.2477
23000 1.02617 1.02178 -3.44287 -3.39148
24000 1.0129 1.00894 -3.62232 -3.54597
25000 1.02434 1.02045 -3.75427 -3.71871
26000 1.01713 1.01535 -3.87747 -3.86579
27000 1.02852 1.01727 -4.03923 -4.01861
28000 1.02305 1.00844 -4.20098 -4.19781
29000 1.02766 1.02655 -4.35916 -4.36012
30000 1.01945 1.02119 -4.50744 -4.50029
31000 1.02106 1.02393 -4.66774 -4.65557
32000 1.00972 1.01787 -4.80983 -4.73875
33000 1.01629 1.02745 -4.94386 -4.89232
34000 1.01976 1.02008 -5.06115 -5.03284
35000 1.0224 1.01516 -5.22286 -5.18223
36000 1.01608 1.03421 -5.35532 -5.32485
37000 1.02182 1.01885 -5.48313 -5.46779
38000 1.01661 1.01749 -5.60281 -5.64273
39000 1.02095 1.01683 -5.73443 -5.79573
40000 1.01407 1.02791 -5.88121 -5.93993
41000 1.00799 1.02199 -5.98232 -6.05922
42000 1.01192 1.01105 -6.11886 -6.16188
43000 1.00438 1.01315 -6.23854 -6.25742
44000 1.01931 1.02233 -6.37419 -6.4011
45000 1.02204 1.02353 -6.48515 -6.53735
46000 1.00656 1.0208 -6.58288 -6.64791
47000 1.01212 1.02948 -6.69286 -6.73881
48000 1.00999 1.00637 -6.85283 -6.87611
49000 1.0135 1.0106 -6.93417 -6.97069
50000 1.01141 1.02738 -7.04317 -7.07784
51000 1.0087 1.00846 -7.14285 -7.18011
52000 1.01351 1.01703 -7.22711 -7.25729
53000 1.01768 1.01825 -7.30364 -7.37323
54000 1.01843 1.0106 -7.42585 -7.46857
55000 1.01045 1.01898 -7.53512 -7.56687
56000 1.02025 1.01726 -7.66796 -7.68184
57000 1.0108 1.00705 -7.75981 -7.79084
58000 1.019 1.01043 -7.82379 -7.88189
59000 1.00689 1.0144 -7.92125 -7.94103
60000 1.01534 1.00644 -8.04316 -8.05097
61000 1.01294 1.00792 -8.08602 -8.1218
62000 1.00112 1.00877 -8.11997 -8.21231
63000 1.01093 1.01758 -8.2102 -8.31214
64000 1.01994 1.01371 -8.27916 -8.41047
65000 1.0082 1.01473 -8.39348 -8.47061
66000 1.00719 1.0076 -8.44043 -8.5413
67000 1.01267 1.01557 -8.53169 -8.61985
68000 1.0012 1.00621 -8.58669 -8.70815
69000 1.01006 1.01011 -8.66703 -8.79496
70000 1.01651 1.00452 -8.71589 -8.85599
71000 1.01351 1.001 -8.77136 -8.91154
72000 1.00862 1.01161 -8.83773 -8.97817
73000 1.00999 1.01927 -8.94437 -9.0326
74000 1.00856 1.00878 -8.99564 -9.1276
75000 1.01414 1.00048 -9.07586 -9.20083
76000 1.00939 1.00042 -9.13904 -9.27644
77000 1.01333 1.00427 -9.19482 -9.34425
78000 1.00908 1.0109 -9.22422 -9.39789
79000 1.00946 1.00095 -9.29409 -9.44213
80000 1.01039 1.00762 -9.30011 -9.4938
81000 1.015 1.0158 -9.36533 -9.57773
82000 1.00982 1.00967 -9.41289 -9.59466
83000 1.00558 1.00077 -9.47086 -9.63924
84000 1.00817 1.01362 -9.52066 -9.69624
85000 1.00441 1.00992 -9.59376 -9.73758
86000 1.00432 0.99704 -9.61869 -9.77794
87000 1.00404 1.00623 -9.64412 -9.84483
88000 1.01235 1.00521 -9.69339 -9.85198
89000 1.00872 1.00286 -9.74298 -9.93249

best
Xuepeng

If you print out thermo output every step for
the first few thousand steps, and plot temp or press,
I imagine you will see they are identical early on
and slowly diverge. That is characteristic of slight
round-offs leading to divergent trajectories. Why
that happens on the same box with Windows vs Linus
could be any number of reasons. E.g. due to math libs
used by the 2 compilers that have slightly different
algorithms or precision for things like sort().

All of that is normal. Why one simulation ends in
error vs not, I have no idea. But it is likely something
about your problem, not something to do with LAMMPS.

Steve

Hi,Steve:
Thank you very much for your explanation! the key is that simulation ends in error" pair distance < table inner cutoff"(my potential is tabulated in lammps) in windows, but it can work in linux! my simulation apply"fix langevin" to control the temperature of system, and the initial configuration of the system is random containing 10000 atoms. why one can work and another can not? i am sure i input the same file and data.file in windows and in linux, and i test it many times. so i wonder if there is bug in windows execute file when we run this kind of simulation?
best
Xuepeng

If you did the test I suggested and verified
that the Windows vs Linux runs drift apart very
slowly, then I am guessing your model does
dynamics that are on the borderline of producing
a pairwise atom separation less than then
the tabulated cutoff you have defined. If that
happens occasionally you will randomly see
it with either executable. One solution would
be to tabulate your potential to a smaller cutoff,
or track which pair of atoms are getting too
close together and figure out why.

Steve

Hi,Steve-
Thank you very much for your comments!Wish you all the best!
Xuepeng