compute pe/atom error with gpu package

Dear Axel,

Thank you very much for your reply.

please *always* reply to the mailing list. if other people run into
the same issue, they should see from the mailing list archives how
this is resolved. thanks

I didn't attach the data.pos file, for it's too large(70MB). And I tried to
use "tar" to compress the data file, but it is still very large (35MB) for

nobody is going to run a test on a huge system anyway. if you want
somebody to follow up on your problem, try to construct to the
smallest possible input or post an input script that adds only the
problematic command(s) to one of the input sets distributed with the
LAMMPS sources.

attachment. The script running just on CPU did not output any error info.
I find I say something wrong in last email, "I think GPU package does
support this "compute". What I really wanted to say is that I don't think
GPU package support this "compute". I guess in the GPU package the gpu
calculates the force and total energy but could not return the energy per
atom back to CPU. Any way to solve this problem?

without being able to easily reproduce it, nobody will look into it.
and without looking into it, it is impossible to give any

well, outside of not using a GPU since it seems to be working on the CPU.