CHARMM Development Project Forums User Discussion & Questions Performance and Benchmarking charmm eef1 implicit membrane parallel processing

I use Free Version charmm 43b1, ./install.com gnu M .
Then run parallel: mpirun -np 4 charmm -i run.inp -o run.out
and serial : charmm -i run.inp -o run.out


main part of run.inp:
---------------------------------
eef1 setup temp 298.15 unit 93 name toppar/solvpar.inp -
membrane slvt water slv2 chex nsmth 10 width 16 aemp 0.85 -
nbond ctonnb 7.0 ctofnb 9.0 cutnb 11.0 group rdie

set fbeta = 5.0
scalar fbeta set @fbeta
scalar fbeta set 0.0 select type H* end

shake bonh param fast
open write unit 12 card name run.rst
open write unit 13 file name run.dcd

DYNAMICS LANGEVIN start nstep 1000000 timestp 0.002 -
nprint 100 iprfrq 5000 isvfrq 5000 ntrfrq 500 -
inbfrq -1 imgfrq -1 ihbfrq 0 ilbfrq 0 -
firstt 300 finalt 300 iseed 47858436 -
tbath 300 rbuf 0.0 -
iunread -1 iunwrite 12 iuncrd 13 iunvelo -1 -
nsavcrd 500 nsavvelo 0 -
iasvel 1
---------------------------------


I do implicit membrane MD of a transmembrane helix. Parallel processing get a unfolding structure and will get out of the membrane in about 5ns, serial processing keep the transmembrane helix in about 150ns.
The default parallel model of charmm is atom decomposition, the EEF1 solvation model can be parallel as described in charmm/doc/parallel.doc .

I have run testcases and find there are only testcases about implicit water, lack of implicit membrane.
if only {eef1 setup temp 298.15 unit 93 name toppar/solvpar.inp} in run.inp, it means implicit water around the protein.
if {eef1 setup temp 298.15 unit 93 name toppar/solvpar.inp membrane slvt water slv2 chex nsmth 10 width 16 aemp 0.85 } in run.inp, it means implicit membrane .

May be there are many researches use eef1 implicit water, but little about eef1 implicit membrane, I wander if eef1 implicit membrane doesn't work well when parallel.

Or can anyone help me do test if you have enhanced version of CHARMM.
My email: millymei@mail.ustc.edu.cn
Thanks very much !

Yes, in the past IMM1 was not parallelized. But in August 2016 I submitted a fix that allowed it to run in parallel. The fix was incorporated into version c42a1. In principle, it should also have made it into version c43b1, I don't know why it didn't.

Actually it did. But we broke the parallel IMM1 code when we submitted the extension to curved membranes last year. We'll submit another fix. Sorry about that.

Thank you for the information !