I’ve installed TESLA card in my computer and downloaded LAMMPS of windows and cuda version.
I’ve runned the test job but it seems that no big increase of computational efficiency.
Do I have to manipulate somethings so that the use of TELSA card of NVIDIA may be valid?
I’d like to restain force and movement of positions for atoms in a certain
selected region. The paper of “Anomalous Strength Characteristics of Tilt Grain Boundaries in Graphene, Rassin Grantab, et al. Science 330, 946 (2010)” deals the system in this way using LAMMPS program.
That is, they restrained the movement and made force zero in a certain direction (e.g. y-axis) for a 2.5 A wide region at two ends,
permitting force application and movement of atoms in this region freely in case they are perpendicular to this direction.
Would you let me know what I do for it in detail (what commands etc.), please?
1. I've installed TESLA card in my computer and downloaded LAMMPS of windows
and cuda version.
I've runned the test job but it seems that no big increase of computational
efficiency.
which test job? are you saying that you are trying to running a gpu
accelerated lammps version on windows???
i would strongly recommend using linux for this. that is well tested.
for some comparison, you can run the tests and compare to the numbers
posted on this page (n.b. the USER-CUDA version has some recent
improvements that are not yet covered by those benchmarks). http://code.google.com/p/gpulammps/wiki/SingleGPUBenchmarkResults
Do I have to manipulate somethings so that the use of TELSA card of NVIDIA
may be valid?
whether and how much of a speedup you get from a gpu depends
a lot on the system size and other parameters. you have to provide
more specific details for anybody to comment and give advice.
2. I'd like to restain force and movement of positions for atoms in a
certain
selected region. The paper of "Anomalous Strength Characteristics of Tilt
Grain Boundaries in Graphene, Rassin Grantab, et al. Science 330, 946
(2010)" deals the system in this way using LAMMPS program.
That is, they restrained the movement and made force zero in a certain
direction (e.g. y-axis) for a 2.5 A wide region at two ends,
permitting force application and movement of atoms in this region freely in
case they are perpendicular to this direction.
Would you let me know what I do for it in detail (what commands etc.),
please?
you have the lammps documentation and can ask the authors what
they did. it is not our job to do _your_ job.
Many thanks for your answer, Steve.
I’d like to ask following questions again.
At what step, force is calculated in LAMMPS?
If “fix npt temp” command is used, I think it may include force calculation and integration steps together.
Then, if I’d like to use “setforce” command, how can I apply it to this case?
We should apply setforce between general force calculation and integration, I think.
I applied setforce right befor fix npt temp but it seems that it wasn’t applied correctly.
Forces are calculated every step. Integration is
indeed done as part of fix npt. Fix setforce simply
changes the forces that have been calculated each
step, and does it properly within the integration
framework. You don't need to worry about when it
is done.