PME tuning sometimes not performed

User discussions
1

GROMACS version: 2023.3
GROMACS modification: No

Running the same simulation several times, the PME tuning stage is sometimes not performed. In some output files I can see the tuning output at the beginning of the simulation:

10000 steps,     20.0 ps.
step 0
step  480: timed with pme grid 320 320 320, coulomb cutoff 1.200: 10531.1 M-cycles
step  640: timed with pme grid 288 288 288, coulomb cutoff 1.302: 11526.9 M-cycles
step  800: timed with pme grid 256 256 256, coulomb cutoff 1.465: 14821.5 M-cycles
step  960: timed with pme grid 280 280 280, coulomb cutoff 1.339: 12217.3 M-cycles
step 1120: timed with pme grid 288 288 288, coulomb cutoff 1.302: 11478.3 M-cycles
step 1280: timed with pme grid 300 300 300, coulomb cutoff 1.250: 10975.4 M-cycles
step 1440: timed with pme grid 320 320 320, coulomb cutoff 1.200: 10631.6 M-cycles
              optimal pme grid 320 320 320, coulomb cutoff 1.200
imb F  2% pme/F 1.11 step 2000, will finish Tue Feb 20 14:39:19 2024

But sometimes this is missing completely:

10000 steps,     20.0 ps.
step 0
step 500, will finish Wed Feb 21 05:37:27 2024
step 1000, will finish Wed Feb 21 05:37:21 2024
step 1500, will finish Wed Feb 21 05:37:19 2024
imb F  2% pme/F 0.70 step 2000, will finish Wed Feb 21 05:37:18 2024

The command I’m using is the following:

gmx_mpi mdrun -v -stepout 500 -s benchPEP-h.tpr -nsteps 10000 -resetstep 5000 -noconfout -ntomp 1 -dlb auto -tunepme -pin on -nstlist 80

Why is PME tuning run only sometimes, despite -tunepme being enabled and no other settings changing between the runs?

Thank you.

2

If the PME/force ratio is below 1 then there will be no benefit of PME tuning and it will not be performed. You see that in your second example. In the first example the scan is performed, but the finally chosen setup is the same as the start setup.