Probelms using PPPM & comparison of Ewald and PPPM

I’ve managed to get the file to work in parallel when I submit using 4 nodes, but when I try using 8 nodes, it doesn’t work.

Interesting to note that I did a v short test of both the Ewald and the PPPM for my system (17712 LJ/COUL/CUT/LONG atoms, 8978 FENE bonds) and the Ewald one was faster than the PPPM? How do I optimze the PPPM one since it’s meant to be faster than Ewald?

log.lammps:ewald: Loop time of 15.7877 on 4 procs for 100 steps with 17712 atoms
log.lammps.pppm:Loop time of 16.464 on 4 procs for 100 steps with 17712 atoms

Yours

Sam

pppm has better scaling than ewald, but may be slower for small/dilute systems. Axel.

But 17K atoms is more than large enough for PPPM to
be significantly faster. Do the same test with in.rhodo (32K)
atoms and you will see that. Something else must be
odd about your system.

Steve