Topic Options
#33801 - 04/15/14 10:50 PM Problem with repdstr
miracle Offline
Forum Member

Registered: 10/04/12
Posts: 5
Hi, all
I have tried to run replica exchange MD with repdstr module in CHARMM c38b1 version.
As explained in REPDSTR.doc, I followed the recommended way to compile repdstr.

./install.com em64t xxlarge M REPDSTR X86_64 IFORT NIH MSCALE ASYNC_PME ALTIX_MPI GENCOMM CMPI.

When I tried to test it with rxsgld.inp in /c36test, the program was terminated with the following message:

* Input
if ?repdstr .ne. 1 then
echo "rxsgld.inp> Test NOT performed."
echo "See repdstr.doc for proper compile procedure."
stop
endif

* output
CHARMM> if ?repdstr .ne. 1 then
RDCMND: can not substitute energy "?REPDSTR"
Comparing "?REPDSTR" and "1".
IF test evaluated as true. Performing command
CHARMM> echo "rxsgld.inp> Test NOT performed."
"rxsgld.inp> Test NOT performed."
CHARMM> echo "See repdstr.doc for proper compile procedure."
"See repdstr.doc for proper compile procedure."
CHARMM> stop

When I just used the rxsgld.inp without the above command, the warning messages (shown below) are appeared after repd command and the program was terminated abnormaly.

***** LEVEL -1 WARNING FROM *****
***** REPlica DiSTRibuted code not compiled.
******************************************

The command I used (/c36test/rxsgld.inp) is the following:

repd nrep @nreps EXCHange FREQuency 50 STEMp 300 MTEMp 400 sump unit 17 -
sgld sgte 300 msgt 600 DSGFt 0.2

Does this mean that repdstr module is not compiled properly or is the problem arising from other reason?
And I cannot figure out what the “?repdstr” exactly means.
I tried to tackle the problem but couldn’t find solution.
Could anybody please give some advice?
Thank you in advance.





Edited by miracle (04/15/14 10:50 PM)

Top
#33802 - 04/16/14 12:11 AM Re: Problem with repdstr [Re: miracle]
rmv Online   content

Forum Member

Registered: 09/17/03
Posts: 8280
Loc: 39 03 48 N, 77 06 54 W
The messages do indicate the REPDSTR code was not compiled correctly. From the c38b1 version of repdstr.doc (available here):

Parallel ditributed replica allows independent replicated systems
over specified number of processors. It mainly works with CMPI
pref.dat keyword (YMMV). REPDSTR is still not the default pref.dat
keyword so the recommended way to compile CHARMM is the following:

install.com gnu xxlarge M mpif90 +REPDSTR +ASYNC_PME +GENCOMM [+MSCALE]

for install.com em64t add +CMPI to the above list.

Also MSCALE is not really needed for pure REPDSTR runs, but it is
needed for triple parallel CHARMM.


The options to install.com are significantly different from what you've posted, which may be from an older version of the documentation.
_________________________
Rick Venable
computational chemist


Top
#33830 - 04/22/14 01:38 AM Re: Problem with repdstr [Re: miracle]
miracle Offline
Forum Member

Registered: 10/04/12
Posts: 5
Thank you! Now it works well smile

Top

Moderator:  BRBrooks, John Legato, wxw