来自"沈志强" <[email protected]>的邮件

Dear all
I have added a pair potential in Lammps(Nov 13) ,and it run well in serial.But when I try to take advantage of the MPI capability of Lammps (Jun 13),the output of the dump file and the fix print file all appear “nan” like below. The related information are attach below.

Could any one help me figure out the problem ?Thank you in advance!

--------------------dumpfile----------------------------\

Step Temp E_pair E_mol TotEng Press
69 1 3.8874321 0.0023160034 5.3896856 nan
nan
nan
70 nan 3.9515662 nan nan nan
71 nan 0 nan nan nan
72 nan 0 nan nan nan
73 nan 0 nan nan nan

[...[

\-----------------------------part of compute-----\
if (rsq < cutsq[itype][jtype])

                { liga = itag;
                     ligd = jtag;
                     ligb = liga-7;
                     ligc = ligd-7;

                  int index[4] =
{atom->map(liga),atom->map(ligb),atom->map(ligc),atom->map(ligd)};//find the
loacal ids of the bonded atom

[...]

                       if(evflag) ev_tally_lire(index[0], index[1],
index[2],index[3], nlocal,newton_pair,evdwl,fabx,faby,fabz, fadx,fady,
fadz,fcdx, fcdy, fcdz, delx, dely, delz, delxii0,delyii0, delzii0, delxjj0,
delyjj0,delzjj0);

1. If you are tallying up a four-body interaction, you should not need
to write a custom tally function, instead you can use the existing
tally function for 4-body: ev_tally4.

2. I am not sure how correct " ligb = liga-7" and " ligc = ligd-7" can be.

3. If you don't have local IDs but only global IDs in your i-j loop,
then most likely your i-j loop is wrong.

Ray