Dear LAMMPS user,
I am using the latest lammps version 5Jun2019.
Here I try to implement SW potential.
I have performed the MD simulations of system consist of poly methoxyethane and calculate the radius of gyration of the molecule. I am using the same input file for GPU and CPU, however I get different result in GPU and CPU. I have the same issue as this one
https://sourceforge.net/p/lammps/mailman/message/36661154/
When I run using CPU, I get the correct radius of gyration of the molecule. However, when I use GPU, the potential energy is much higher and cannot reproduce the paper result. I also try to increase the “skin” but it does not help.
Could someone help me to solve this issue?
This is my input file
dme.inp (1.82 KB)
dme.data (489 KB)
dme.sw (5.63 KB)
Dear LAMMPS user,
I am using the latest lammps version 5Jun2019.
Here I try to implement SW potential.
I have performed the MD simulations of system consist of poly methoxyethane and calculate the radius of gyration of the molecule. I am using the same input file for GPU and CPU, however I get different result in GPU and CPU. I have the same issue as this one
https://sourceforge.net/p/lammps/mailman/message/36661154/
When I run using CPU, I get the correct radius of gyration of the molecule. However, when I use GPU, the potential energy is much higher and cannot reproduce the paper result. I also try to increase the “skin” but it does not help.
Could someone help me to solve this issue?
your input is bogus. you are using bonded interactions with a manybody potential, which assumes for the computation of manybody terms, that there are no excluded pairs in the neighbor list. but there will be excluded pairs because of the bonds and no adjustment to the special_bonds command.
axel.
Dear Axel, thanks for your reply.
I following the suggestion in the lammps manual in special_bond command, “The recommended course of action is to remove such bonds, or - if that is not possible - use a special bonds setting of 1.0 1.0 1.0.”
So I add one more line in my input file
special_bond lj/coul 1.0 1.0 1.0
When I am using the CPU, the potential energy become absolutely different.
However, there is no change in potential energy when I run the simulation using GPU whether I add that line or not (special_bond lj/coul 0.0 0.0 0.0).
Could you help me with this, please?
Best regards,
Ravi Mahesta
Before even thinking about using the GPU version, you should verify that the cpu version is producing correct results for this very unusual model. Axel.