Help：LAMMPS GPU

LAMMPS LAMMPS Mailing List Mirror
1

hello, I know why something went wrong now.

maybe because lib/gpu, when I "git clean -Xdf ",
0. mkdir build

  1. cmake -D PKG_MANYBODY=yes -D BUILD_OMP=yes -D PKG_GPU=on -D GPU_API=cuda -D GPU_ARCH=sm_61 …/cmake/
  2. make -j
  3. make install
    all is OK

Thanks, “if you use CMake, you don’t need to do anything in lib/gpu (which is also explained in the manual).” is right

Axel Kohlmeyer <[email protected]> 于2020年6月9日周二 上午5:32写道：

image.png

image.png

image.png

image.png

image.png

image.png

image.png

image.png

2

thanks a lot for this feedback. this is extremely useful. there are indeed generated files in the lib/gpu/ folder that are picked up by the cmake based compilation. if those have the wrong arch and compilation settings then those will cause the unexpected error, since they will be picked up by accident by the cmake compile. there is a similar problem when enabling packages for the traditional build system in the src folder, which we already test for.
we will thus add a test to the cmake based procedure to detect those files and print a suitable error message with a suggestion to resolve this issue as well.

thanks again,
axel.