Previous Thread
Next Thread
Print Thread
Page 1 of 2 1 2
why I cannot read coordinate and veloci files by traj or merge?
#19627 12/18/08 11:37 PM
Joined: Dec 2008
Posts: 5
D
dswei Offline OP
Forum Member
OP Offline
Forum Member
D
Joined: Dec 2008
Posts: 5
when I use traj or merge commands to operate the trajectory files of .cor and .vel generated in a sequent dynamics simulations, Charmm always displays the following error:

CHARMM> traj coor read
***** ERROR ***** HEADERS DO NOT MATCH �^Qo� CORD VELD
*** LEVEL -1 WARNING *** BOMLEV IS 0
BOMLEV HAS BEEN SATISFIED. TERMINATING.

It seems I cannot read the .cor and .vel files. I run Charmm on the same machine. Does anyone have some ideas? Thanks so much!

Re: why I cannot read coordinate and veloci files by traj or merge?
dswei #19628 12/19/08 12:40 AM
Joined: Sep 2003
Posts: 8,500
rmv Online Content
Forum Member
Online Content
Forum Member
Joined: Sep 2003
Posts: 8,500
As there really isn't a TRAJ COOR READ command, I'm guessing that there may be other errors in your input as well.

Re: why I cannot read coordinate and veloci files by traj or merge?
rmv #19629 12/19/08 04:45 PM
Joined: Dec 2008
Posts: 5
D
dswei Offline OP
Forum Member
OP Offline
Forum Member
D
Joined: Dec 2008
Posts: 5
I found the problem is that the generated .cor or .vel file is not correct. I checked all of my .cor file and found they are the same size. I doubted the .cor file is recorded by REPLACE not by APPEND during dynamics running. When I used the serial Charmm, I got the correct .cor and .vel files. So what's wrong with the parallel charmm? Has anybody met this problem?

Re: why I cannot read coordinate and veloci files by traj or merge?
dswei #19630 12/19/08 06:29 PM
Joined: Sep 2003
Posts: 8,500
rmv Online Content
Forum Member
Online Content
Forum Member
Joined: Sep 2003
Posts: 8,500
There is nothing wrong with parallel CHARMM with regard to writing trajectory files during MD. I suggest some more careful reading of the CHARMM documentation and your output, and then consider a followup question about specific CHARMM commands or options. You question is too vague to answer, and the sentence "I doubted the .cor file is recorded by REPLACE not by APPEND during dynamics running." is not easily understood.

Re: why I cannot read coordinate and veloci files by traj or merge?
rmv #19631 12/19/08 08:16 PM
Joined: Dec 2008
Posts: 5
D
dswei Offline OP
Forum Member
OP Offline
Forum Member
D
Joined: Dec 2008
Posts: 5
I'm sorry I don't make my question well understood. Anyway, thanks so much for rmv's reply. I use c32b2 code to run parallel charmm on IBM/AIX cluster. The following is my dynamics simulation script:

! md run
open unit 30 write file name scr/heat.dcd
open write formatted unit 31 name scr/heat.res
open write unformatted unit 40 name scr/heat.vel

dyna leap verlet start nstep 6200 timestep 0.002 -
firstt 0 finalt @temp -
ihtfrq 10 teminc .5 nsavc 10 -
ieqfrq 0 iasors 1 iasvel 1 nsavv 20 -
isvfrq 500 iprfrq 500 nprint 100 -
iunrea -1 iunwri 31 iuncrd 30 iunvel 40 -
imgfrq 20 inbfrq 10

My question is that when I change the value of nsavc from 10 to 100, the size of the generated .dcd file doesn't change. The .vel file also doesn't. Moreover I cannot use traj or merge commands to operate them. But when I run the serial c32b2 charmm, the generated .dcd and .vel files are normal.

Comparing with the normal files generated by serial running , I found the files generated by parallel running record the coordinates or velocity only at one time not at (nstep/nsavc) times. This makes me stumbled. So I ask for help.

Best!

dswei

Re: why I cannot read coordinate and veloci files by traj or merge?
dswei #19632 12/19/08 09:06 PM
Joined: Sep 2003
Posts: 8,500
rmv Online Content
Forum Member
Online Content
Forum Member
Joined: Sep 2003
Posts: 8,500
The symptoms suggest there may be some issues in how the parallel version was compiled for that system type. There may a problem with the parallel executable you are trying to use.

A couple additional notes--

* heating from zero usually isn't needed or recommended
* it is usually counterproductive to specify ISVFRQ
* unless CUTIM is a couple A larger than CUTNB, IMGFRQ may be too large
* it's usually best to start with INBFRQ -1 IMGFRQ -1 to start with

Re: why I cannot read coordinate and veloci files by traj or merge?
rmv #19633 12/22/08 11:45 PM
Joined: Dec 2008
Posts: 5
D
dswei Offline OP
Forum Member
OP Offline
Forum Member
D
Joined: Dec 2008
Posts: 5
hi rmv,

I still not solve this problem. Today I did an another test. When setting the IUNXYZ, NSAVX and MXYZ parameters to print out the xyz file at different running times by the parallel Charmm, I can get the correct .xyz file. Because of the .xyz file is formatted, I thought whether or not the cluster IBM/AIX I submit job has a problem with the unformatted file output? Do you have any advice? Thanks!

Regards!

Dongshan

Re: why I cannot read coordinate and veloci files by traj or merge?
dswei #31315 01/23/13 07:06 PM
Joined: Nov 2008
Posts: 11
C
Forum Member
Offline
Forum Member
C
Joined: Nov 2008
Posts: 11
Dear All,

I have a similar issue with writting the velocities file card in the gnu open and intel MPI version of c36b1 (tested under E5620 w/gcc and E5-2660 w/icc, both with leap and vver integrators).

The velocity card is written WITHOUT updating according to nsavv frames in the parallel version while in the serial version it updates each nsavv frames accordingly.
To be clear: a velocity file with identical content for each nsavv frame is written in the parallel version. The serial version behaves as it should.

Making more tests, I also noticed I could not make nsavx to write in serial nor parallel, but this may be a different issue. (unit stays at 0KB, and is strange that nsavx is not shown in the output even prnlev 10; below the prnlev 10 trukated CHARMM output). Also using last Dongs' trick of changing the nsavx and nsavc frequencies does not make a difference.


Has this problem been pinpointed yet? I will keep trying other versions. Your help will be kindly acknowledged, many thanks!






--------------------------------------

CHARMM> open write unit 20 file name dcd/rotor.@te1.@ii.dcd
OPNLGU> Unit 20 opened for WRITE access to dcd/rotor.300.2.dcd

CHARMM> open write unit 21 file name dcd/rotor.@te1.@ii.vel
OPNLGU> Unit 21 opened for WRITE access to dcd/rotor.300.2.vel

CHARMM> open write unit 22 form name dcd/rotor.@te1.@ii.xyz
OPNLGU> Unit 22 opened for WRITE access to dcd/rotor.300.2.xyz

CHARMM> open write unit 23 form name dcd/rotor.@te1.@ii.tem
OPNLGU> Unit 23 opened for WRITE access to dcd/rotor.300.2.tem

CHARMM> ! NOSE
CHARMM> dynamics vver nose restart qref @qfriction tref @te1 ncyc 4 nsnos @sfrq -
CHARMM> ! LANGEVIN
CHARMM> !dynamics leap langevin restart -
CHARMM> timestp 0.002 nstep @te1stp ilbfrq @te1stp rbuf 0.0 tbath @te1 -
CHARMM> iprfrq @sfrq ihtfrq 0 ieqfrq 0 ntrfrq 0 -
CHARMM> iunrea 16 iunwri 15 iuncrd 20 iunvel 21 iunxyz 22 kunit 23 mxyz 3 -
CHARMM> isvfrq @te1stp nsavc @sfrq nsavv @sfrq nsavx @sfrq nprint @sfrq inbfrq 0 ihbfrq 0
CHARMM> firstt @te1 finalt @te1 tstruc @te1 -
CHARMM> iasors 0 iasvel 1 ichecw 0 twindl 0.0 twindh 0.0

IUNREA = 16 IUNWRI = 15 IUNOS = -1
IUNCRD = 20 IUNVEL = 21 KUNIT = 23
TITLE> * STREAM TO REAL FILE
TITLE> * DATE: 1/23/13 18:33: 3 CREATED BY USER: palma
TITLE> *
READYN> dynamics restart file was read. Current step= 15000
NSTEP = 15000 JHSTRT = 15000
DCNTRL> Nose-Hoover dynamcs requested.
Reference temperature = 300.0000000 K.
Temperature coupling constant = 1000.0000000Kcal sec**2
Cycles of convergence = 4

NSTEP = 100000 NSAVC = 100 NSAVV = 100
ISCALE = 0 ISCVEL = 0 IASORS = 0
IASVEL = 1 ICHECW = 0 NTRFRQ = 0
IHTFRQ = 0 IEQFRQ = 0 NPRINT = 100
INBFRQ = 0 IHBFRQ = 0 IPRFRQ = 100000
ILBFRQ = 100000 IMGFRQ = 0
ISVFRQ = 100000 NCYCLE = 100000 NSNOS = 100
FIRSTT = 300.000 TEMINC = 10.000 TSTRUC = 300.000
FINALT = 300.000 TWINDH = 0.000 TWINDL = 0.000


Re: why I cannot read coordinate and veloci files by traj or merge?
dswei #31317 01/23/13 10:45 PM
Joined: Sep 2003
Posts: 8,500
rmv Online Content
Forum Member
Online Content
Forum Member
Joined: Sep 2003
Posts: 8,500
This probably should have been a new topic, as it isn't at all clear that this is the same problem, esp. since since you are using a much more recent CHARMM release, from after the Fortran95 conversion. Also, this post probably belongs in the Molecular Dynamics forum, since it concerns issues related to running MD. Guidelines for posting can be found in the READ BEFORE POSTING topic, at the top of each forum.


We don't know if the problems @dswei was having were resolved or not; there was no further response.

I'm not aware of any other reports of problems writing velocity files in parallel, and it worked as expected the last time I tried it (using LEAP CPT).

It should be noted that VVER NOSE is older code that isn't as well supported as some of the newer code.

I don't know how well supported the XYZ output code is, either; I've never made use of it myself. Apparently, no one has ever bothered to add NSAVX to the values printed at DYNAmics startup.


Rick Venable
computational chemist

Re: why I cannot read coordinate and veloci files by traj or merge?
dswei #31318 01/24/13 12:12 PM
Joined: Nov 2008
Posts: 11
C
Forum Member
Offline
Forum Member
C
Joined: Nov 2008
Posts: 11
Ok sry sure, thanks again for the reply.

In the meantime the issue is SOLVED. Indeed as you pointed out this is a bug of the VVER NOSE integrator, much to my 300K CPU hours demise. LEAP LANG works fine.

Bests!

Last edited by cpalma; 01/24/13 12:14 PM.
Page 1 of 2 1 2

Moderated by  lennart, rmv 

Link Copied to Clipboard
Powered by UBB.threads™ PHP Forum Software 7.7.4
(Release build 20200307)
Responsive Width:

PHP: 5.6.33-0+deb8u1 Page Time: 0.016s Queries: 34 (0.007s) Memory: 0.9903 MB (Peak: 1.1224 MB) Data Comp: Off Server Time: 2020-10-01 19:19:41 UTC
Valid HTML 5 and Valid CSS