Dear Prof. Axel,
Thank you so much for your kindly instructions.
Please aware that I didn’t mean the problem come from Lammps GPU source code, since I also have an installation of Lammps gpu package on Linux cluster, and everything looks to going well.
I just try to figure out what is problem on my windows gpu package.
I know that for the windows pre-built installer, we can not get double precision. So whether we can set the precision from input file? (I mean set it to just single or mixed precision)
I just do some MD calculations using EAM potential.
Thank you so much
Dear Prof. Axel,
Thank you so much for your kindly instructions.
Please aware that I didn't mean the problem come from Lammps GPU source
code, since I also have an installation of Lammps gpu package on Linux
cluster, and everything looks to going well.
i didn't understand it as criticism, but just wanted to support my claim,
that the issue is likely outside of the LAMMPS source code.
I just try to figure out what is problem on my windows gpu package.
I know that for the windows pre-built installer, we can not get double
precision. So whether we can set the precision from input file? (I mean set
it to just single or mixed precision)
I just do some MD calculations using EAM potential.
the GPU precision setting is a compile time option. if you want different
options, you have to compile LAMMPS for windows by yourself. mind you,
that you will still be locked into using OpenCL instead of CUDA, since my
recollection is, that CUDA for windows only supports the visual c++
compilers, while LAMMPS will likely need some porting effort to correctly
work with those. the windows binaries are build with a linux-to-windows
cross compiler.
axel.