[gmx-users] Testing the performance of

Richa Singh richa.s.rathorerr at gmail.com
Wed Jul 10 16:29:37 CEST 2013


Thanks! Mark.

I got solution to my problem. For my system, setting pinoffset = 0, 16
and pinstride = 1 for two simultaneous simulations gave much
satisfactory performance both for Verlet and Group cutoff-scheme. I
tried it for 4 simultaneous simulations also with pinoffset = 0, 8,
16, 24 and pinstride = 1 , and it works!!

Command:     mdrun -nt 16 -v -pin on -pinoffset 0 -pinstride 1 -deffnm file &
                     mdrun -nt 16 -v -pin on -pinoffset 16 -pinstride
1 -deffnm file &


Group: 21.559+21.879 = 43.348 ns/day
Verlet: 20+20.293 = 40.293 ns/day




Richa.



More information about the gromacs.org_gmx-users mailing list