Fwd: Wrong exit from RAMD runs

From: Francesco Pietra (chiendarret_at_gmail.com)
Date: Tue Apr 05 2011 - 15:34:19 CDT

This is to add that now, for the first time out of so many runs in the
past and present, I have seen a formally correct exit from a ramd run.
Same system as the wrong exit described below. The only difference was
the acceleration.
ENERGY:  196500       935.6674      1456.8738       592.0131
99.5454        -124940.0012     10232.3179         0.0000
0.0000     21324.5279         -90299.0557       309.2950
-111623.5835    -90268.0435       309.4156           -144.5299
147.9136    331258.5733       141.8714       139.1275

WRITING EXTENDED SYSTEM TO RESTART FILE AT STEP 196500
WRITING COORDINATES TO DCD FILE AT STEP 196500
WRITING COORDINATES TO RESTART FILE AT STEP 196500
FINISHED WRITING RESTART COORDINATES
The last position output (seq=196500) takes 0.021 seconds, 16315.520
MB of memory in use
WRITING VELOCITIES TO RESTART FILE AT STEP 196500
FINISHED WRITING RESTART VELOCITIES
The last velocity output (seq=196500) takes 0.004 seconds, 16315.262
MB of memory in use
TCL: EXIT: 196500  > MAX DISTANCE LIGAND COM - PROTEIN COM REACHED
TCL: EXIT: 196500  > LIGAND EXIT EVENT DETECTED: STOP SIMULATION
TCL: EXIT: 196500  > EXIT NAMD

However, what I have also always noticed in the past with a different
protein-ligand,  and in these days, playing in VMD shows that the
ligand left the protein, entering surrounding water, much before. NAMD
did not stop at ligand entering the surrounding water and forces on
ligand continued, pushing it to the other side of the surrounding
water, without touching the protein.

francesco

---------- Forwarded message ----------
From: Francesco Pietra <chiendarret_at_gmail.com>
Date: Tue, Apr 5, 2011 at 10:11 AM
Subject: namd-l: Wrong exit from RAMD runs
To: NAMD <namd-l_at_ks.uiuc.edu>

Hi:
like in the past - when ramd tcl scripts were external to namd - i am
observing a wrong exit from namd runs.

With acceleration 0.1:

WRITING EXTENDED SYSTEM TO RESTART FILE AT STEP 34700
WRITING COORDINATES TO DCD FILE AT STEP 34700
WRITING COORDINATES TO RESTART FILE AT STEP 34700
FINISHED WRITING RESTART COORDINATES
The last position output (seq=34700) takes 0.019 seconds, 2907.915 MB
of memory in use
WRITING VELOCITIES TO RESTART FILE AT STEP 34700
FINISHED WRITING RESTART VELOCITIES
The last velocity output (seq=34700) takes 0.003 seconds, 2907.782 MB
of memory in use
TCL: EXIT: 34700  > MAX DISTANCE LIGAND COM - PROTEIN COM REACHED
TCL: EXIT: 34700  > LIGAND EXIT EVENT DETECTED: STOP SIMULATION
TCL: EXIT: 34700  > EXIT NAMD
Charmrun: error on request socket--
Socket closed before recv.

Actually, this has no consequences. Except for the wrong exit,
everything was in order.
*******************************
Now, however, with the same system, with acceleration 0.05:
WRITING EXTENDED SYSTEM TO RESTART FILE AT STEP 493670
WRITING COORDINATES TO RESTART FILE AT STEP 493670
FINISHED WRITING RESTART COORDINATES
The last position output (seq=493670) takes 0.014 seconds, 40916.619
MB of memory in use
WRITING VELOCITIES TO RESTART FILE AT STEP 493670
FINISHED WRITING RESTART VELOCITIES
The last velocity output (seq=493670) takes 0.004 seconds, 40916.769
MB of memory in use
Charmrun: error on request socket--
Socket closed before recv.

That is, the simulation crashed at step 493670 without the ligand
having left the protein.

In both cases, the 30 March namd night-build GNU/Linux 64 (no cuda)
Debian amd64 was used on a shared memory multi opteron machine with
huge ram. chamm force field 27, npt 1 bar, 310K. The system had been
accurately equilibrated in his solvated box, and simulation was with
ts 1 fs, rigid bonds for water only.
****************************
ATTACHED please find more detailed end of log, which I was unable to
to represent correctly in gmail plain text. There it is seen that
energies were in order. Additionally, please notice that with namd
alone (without ramd) I never had crashes with this system (same ff,
same ts = 1fs, rigid bonds water, 310K) even for longer runs.

cheers
francesco pietra


This archive was generated by hypermail 2.1.6 : Mon Dec 31 2012 - 23:20:05 CST