Dear All
I am using Jan 9th 2009 version of LAMMPS. I am doing a minimization problem using CG
with line quadratic min_modify option. In the script, I am minimizing the system for
different displacement increments. It runs fine for several displacement increments but
gives me a segmentation fault after about 100 displacement increments and backtracks to linemin_quadratic. Any suggestions?.
Thanks
Best Regards
Sreekanth
P.S: I am attaching the input script, log file and the error file
in.rough.0.6.blk.min.discrete (1.88 KB)
log.lammps (152 KB)
lmp_xeon_mar_2009.80s-17111,hhpc123.btr (819 Bytes)
I would see if you can save the snapshot before the last bad
minimization and see if you can reproduce the problem with a single
minimization, starting from that state.
Maybe Aidan has another idea.
Steve
Hi Nitin,
If I understand your question appropriately, you can possibly use the fix viscous command to achieve this. To help you in implementing this, I recommend you read the following:
Phys Rev B, vol 73, #224111 (2006),
particularly section III. I also recommend reading the “fix viscous” and “fix langevin” documentation (and related information).
Best of luck,
Ed
I have restarted the simulation from the last minimized state. I can’t reproduce the segmentation fault but it does not output anything. I mean, it took about 10 hours and did not write anything in the log.lammps nor in the out.screen file. Any suggestions?
Thanks
Best Regards
Sreekanth
If it ran for so long, and you can't reproduce it from a restart, the first
thing I would suspect is the computer. Try running exactly the same calc and
see what happens.
Aidan
Nope - someone would have to add that as a fix.
Steve
I doubt “absorbing boundaries” are compatible to lammps since they are not convenient for parallelization.
Z. Insepov
Steve Plimpton wrote:
hi Nitin
we used the standard lammps commands (fix viscous) to create a non-reflecting boundary condition. If that is what you are asking about I can give you the gist and the full treatment should appear in PhysRevB soon.
Reese
Steve Plimpton wrote:
Nitin,
Nice to hear you from the email list. How are you recently.
Can you let me know what is the meaning of “Absorbing boundary conditions” ? I have not heard it before. Sounds like a new methodology. Just want to keep up with new techniques and terminologies.
AC
Dear All
I am using Jan 9th 2009 version of LAMMPS. I am doing a minimization problem using CG
with line quadratic min_modify option. In the script, I am minimizing the system for
different displacement increments. It runs fine for several displacement increments but
gives me a segmentation fault after about 100 displacement increments and backtracks to linemin_quadratic.
I have tried to restart the simulation and reproduce the error to make sure if its not something non-related to the code. But, when i restart the simulation, it runs for about 10 hours and does not write anything in log file or out file. Its not even able to open and read the restart file. I thought of the possibility that the particular restart file to be corrupted and tried with another restart file, but it behaves the same way.
I was running many such cases with same number of atoms and processors. All of them except this one, ran successfully without any error. By this, i could not even assert that its the problem of the cluster.
Moreover, i have checked the last snapshot before the breakdown for any anomalies. I could not find any. I am going to try restarting with different number of processors if that could change anything. If anybody ran into such problem, please let me know. Please find the attachments as input, log and error file.
Thanks
Best Regards
Sreekanth
in.rough.0.6.blk.min.discrete (1.88 KB)
lmp_xeon_mar_2009.80s-17111,hhpc123.btr (819 Bytes)
log.lammps (152 KB)