Dear LAMMPS Developers and User
Does anybody know how to accelerate Kspace calculation?
or a tricky way of neglecting its effect?
Many Thanks
kspace_modify command (several options), increase the real space
cutoff, lower precision FFTs, use run_style verlet/split,
Wolf summation …
Steve
Thanks, Steve
Does run_style verlet/split also speedup Kspace calculations on GPU?
I’m using GeForce GPU paralleled with a Core i7 CPU. What’s the best practical way to speedup Kspace calculations in my case?
The GPU package already overlaps Kspace time with pair time. I’d suggest balancing the two costs, and then Kspace should be negligible. Or use wolf/dsf pair styles that run really well on GPUs. See http://lammps.sandia.gov/doc/pair_coul.html.
Stan
