[gpumd] Re: [gpumd]

  • From: Alex Gabourie <agabourie47@xxxxxxxxx>
  • To: gpumd@xxxxxxxxxxxxx
  • Date: Wed, 16 Oct 2019 10:24:52 -0700


It's tough to say exactly what might cause that problem, but I think that Zheyong is correct in that the LJ potential definition is incorrect. Since you have two atom types defined, the LJ potential must be of the form 'lj 2'. The fix is simple. Your lj file can look like:

lj 2
2.84e-3 3.405 10.0
2.84e-3 3.405 10.0

Zheyong is also right in that you'll need to increase your neighborlist size but, if that doesn't work, make sure that you are building GPUMD with the -DDEBUG flag removed in the makefile. I've had that cause me memory problems before.


On 10/16/19 6:58 AM, Bruce Fan wrote:

Hi Mostafa,

So the simulation was done using a developing version, right?

First, I add the contents of your LJ potential file here (you sent me this in a private email):

lj 1
2.84e-3 3.405 10.0

Will this contradict the line
potential    potentials/lj/c_clj.txt 0 1
in the run.in <http://run.in> file?

@Alex Gabourie <mailto:gabourie@xxxxxxxxxxxx> How do you think?

Apart from this, I notice a possible source of error in the first line of xyz.in <http://xyz.in>:
 17466 50 11 0 0 1

Here, the parameter M (maximum number of neighbors) is set to 50. This seems to be too small for a cutoff of 10 Angstrom (from the LJ part). You can change it to 500. This might be the reason for the error message you got.


On Wed, Oct 16, 2019 at 4:42 PM Mostafa Valadkhani <mostafa.v68@xxxxxxxxx <mailto:mostafa.v68@xxxxxxxxx>> wrote:

    Sorry if I ask a lot and time to time. Actually I'm calculating
    SHC with lammps using Kimmo Codes and wanna relate it somehow with
    the results from GPUMD.
    I wanna calculate SHC on the bilayer/monolayer graphene as your
    paper with Ali.The attachments are my inputs, when I run the
    script following error comes up:

    CUDA Error:
        File:       neighbor_ON1.cu
        Line:       245
        Error code: 700
        Error text: an illegal memory access was encountered

    can you give me some hints as always?
    And I'm looking forward for your implementation of GKMA on GPUMD too.

Other related posts: