Has anyone had problems with the lj/class2/coul/long/gpu pair style?
When I try to run lammps with more than 6 threads I get the error
Cuda driver error 1 in call at file ‘geryon/nvd_kernel.h’ in line 364.
However changing the pair style to lj/class2/gpu seems to fix the problem?
I’m not sure what info you need, I am using a build from 2015-December-07 with the lmp_intel_openmpi_gpu binary.
Thanks
Christopher Watkins
Applications Support Specialist
IM&T | Scientific Computing
CSIRO
E Chris.Watkins@…4168… T +61 3 9545 2369
Private Bag 10, Clayton South, Vic, 3169
www.csiro.au
PLEASE NOTE
The information contained in this email may be confidential or privileged. Any unauthorised use or disclosure is prohibited. If you have received this email in error, please delete it immediately and notify the sender by return email. Thank you. To the extent permitted by law, CSIRO does not represent, warrant and/or guarantee that the integrity of this communication has been maintained or that the communication is free of errors, virus, interception or interference.
Has anyone had problems with the lj/class2/coul/long/gpu pair style?
When I try to run lammps with more than 6 threads I get the error
what do you mean exactly by "more than 6 threads"?
what kind of GPU? with how much RAM? is this a compute node, or on a
GPU that is also used for display.
how large is the problem? does it happen with smaller or larger problems, too?
Cuda driver error 1 in call at file 'geryon/nvd_kernel.h' in line 364.
However changing the pair style to lj/class2/gpu seems to fix the problem?
that is missing the coulomb computation, so rather a different beast.
I’m not sure what info you need, I am using a build from 2015-December-07
with the lmp_intel_openmpi_gpu binary.
see above. it would also be nice to have a simple test input that
quickly reproduces the issue and instructions how exactly the run was
launched.
axel.
Thanks guys,
what do you mean exactly by "more than 6 threads"?
When I run with mpirun -np 6 (or greater)
what kind of GPU?
Tesla K20m
with how much RAM?
5 GB
is this a compute node, or on a GPU that is also used for display.
Compute node
how large is the problem?
50314 atoms
does it happen with smaller or larger problems, too?
Haven't checked.
it would also be nice to have a simple test input that quickly reproduces the issue and instructions how exactly the run was launched.
Sorry I am not a lammps pro, so I can't be useful here. One suggestion from a colleague is that "there are empty spaces in the input structure, which may result in cells that have no atoms in it, which in turn messes up the kspace solver. Increasing the CPUs makes more (and therefore smaller) cells, and increasing the chance of this happening." Does that sound like a reasonable theory?
Thanks again
Christopher Watkins
Applications Support Specialist
IM&T | Scientific Computing
CSIRO
E [email protected]... T +61 3 9545 2369
Private Bag 10, Clayton South, Vic, 3169
www.csiro.au
PLEASE NOTE
The information contained in this email may be confidential or privileged. Any unauthorised use or disclosure is prohibited. If you have received this email in error, please delete it immediately and notify the sender by return email. Thank you. To the extent permitted by law, CSIRO does not represent, warrant and/or guarantee that the integrity of this communication has been maintained or that the communication is free of errors, virus, interception or interference.