Hi all,
I'm not sure if this is a real bug, or a user error ;-) But here we go... I'm running a simulation using a hybrid granular/molecular pair style with some simple aggregates of non-overlapping spheres. The aggregates are allowed to settle in a frictional box, and this all works fine. I then start to "wiggle" the floor, and it works fine for a while, and then crashes with the following segfault:
825000 26595 -0 0.54522916 139251.09 0.32
827500 26595 -0 0.53139484 139251.09 0.32
[comp-l05:30593] *** Process received signal ***
[comp-l05:30593] Signal: Segmentation fault (11)
[comp-l05:30593] Signal code: Address not mapped (1)
[comp-l05:30593] Failing at address: 0x50d2000
[comp-l05:30593] [ 0] /lib/libpthread.so.0(+0xf8f0) [0x7f8ae95208f0]
[comp-l05:30593] [ 1] /home/mab/bin/lmp_openmpi(_ZN9LAMMPS_NS8Neighbor22granular_bin_no_newtonEPNS_9NeighListE+0x603) [0x6dc5a3]
[comp-l05:30593] [ 2] /home/mab/bin/lmp_openmpi(_ZN9LAMMPS_NS8Neighbor5buildEv+0xeb) [0x6d3ddb]
[comp-l05:30593] [ 3] /home/mab/bin/lmp_openmpi(_ZN9LAMMPS_NS6Verlet3runEi+0x141) [0x7ca0b1]
[comp-l05:30593] [ 4] /home/mab/bin/lmp_openmpi(_ZN9LAMMPS_NS3Run7commandEiPPc+0x26a) [0x79e52a]
[comp-l05:30593] [ 5] /home/mab/bin/lmp_openmpi(_ZN9LAMMPS_NS5Input15execute_commandEv+0x9ae) [0x6a2dee]
[comp-l05:30593] [ 6] /home/mab/bin/lmp_openmpi(_ZN9LAMMPS_NS5Input4fileEv+0x3c8) [0x6a3648]
[comp-l05:30593] [ 7] /home/mab/bin/lmp_openmpi(main+0x49) [0x6ac869]
[comp-l05:30593] [ 8] /lib/libc.so.6(__libc_start_main+0xfd) [0x7f8ae91acc4d]
[comp-l05:30593] [ 9] /home/mab/bin/lmp_openmpi() [0x479ad9]
[comp-l05:30593] *** End of error message ***
Any suggestions as to if this is a user error or a bug? I experienced something similar in a smaller test sim, and found that by turning down the vibration amplitude the error went away, but this is a larger sim/deeper bed and I have already turned the amp down. It could still just be a mis-match of timesteps/spring constants etc. but any insights would be great! I can provide my input script if needed.
Regards,
Mark
ckloss_ | Fri, 08/12/2011 - 22:46
Hi Mark, my guess would be
Hi Mark,
my guess would be partly/partly. First of all, there should not be a segfault popping up. On the otherhand, this specific segfault is typical if the systems "blows" due to some bad dynamics / bad boundary conditions, so this would be my guess.
Chris
PS: I know you were interested in the bond model - I am out of office now, but if you want I can send it to you as I return early sept - just drop me an email
msbentley | Fri, 08/19/2011 - 15:58
Yes, a lower amplitude helped!
Thanks Chris, I tried again with a lower vibration amplitude and the problem went away. I will have to go back and do some "back of the envelope" calculations later to confirm the sanity of my input values!