switch long range force calculation from ewald to pppm method on the half way

Dear LAMMPS Users,

Currently,I am applying the ewald of K_SPACE style to calculate the long range.

There should be 12 cases for my simulation. And compare the results between them.

Now I have finished only 2 cases.I want to switch the K_space style to PPPM. Is it possible by doing this?

Another, I would like to know whether the relative error will occur by running the same jobs on CPU and GPU?

Thank you very much

Liyi

Dear LAMMPS Users,

Currently,I am applying the ewald of K_SPACE style to calculate the long range.

There should be 12 cases for my simulation. And compare the results between them.

Now I have finished only 2 cases.I want to switch the K_space style to PPPM. Is it possible by doing this?

just issue a new kspace_style command.

Another, I would like to know whether the relative error will occur by running the same jobs on CPU and GPU?

when you compile the GPU support with full double precision, the results should be the same (albeit not identical, since floating point math is not associative and operations are processed in a different order on the GPU than on the CPU. i.e. the difference should be comparable to running on a different number of CPUs or changing from newton on to newton off.

axel.