Hi all,
It seems that OpenMM has a problem with data overflow, (or addressing). The running time files produced by gromacs, with a system of 3361 atoms, showed that, for double version, the running could last for like 20 steps and for single version, it would run several steps, maybe even less. In latter times, both the programs continued running, but the data were kept the same, and even the time steps. The first lines looks like following, and the numbers has no difference in the rest of the result:
Generated by trjconv : 3361 system t= 0.00000
3361
1nit N1 1-214748.3594-214748.3594-214748.3594
1nit N2 2-214748.3594-214748.3594-214748.3594
2nit N1 3-214748.3594-214748.3594-214748.3594
2nit N2 4-214748.3594-214748.3594-214748.3594
3nit N1 5-214748.3594-214748.3594-214748.3594
3nit N2 6-214748.3594-214748.3594-214748.3594
4nit N1 7-214748.3594-214748.3594-214748.3594
4nit N2 8-214748.3594-214748.3594-214748.3594
We appreciate any suggestions on solving this problem. Thanks!