Hi
Although I used
CUDA_PRECISION = -D_SINHGLE_SINGLE
but according to the cuda profiler, there are some FP64 instructions. Is that normal?
The command I use are
cd lammps-12Dec18/lib/gpu
vim Makefile.linux
CUDA_ARCH = -arch=sm_50
make -f Makefile.linux
cd ../../src
make yes-gpu
make yes-molecule
make yes-kspace
make yes-rigid
make mpi
Regards,
Mahmood