ifc, gibbs and ewald

From: David Smith <David.Smith_at_cup.uni-muenchen.de>
Date: Tue 4 Feb 2003 16:49:53 +0000

Hi everyone,

I tried to send this last week but it didn't seem to go through. Apologies
if it comes twice:

I have some questions involving ewald in gibbs and also the intel fortran
compiler.

At the moment I am trying to do a few things with gibbs in a large periodic
system for which I would like to use ewald.  As far as I can tell, the things
I am trying to do require some particular features of gibbs (PMFs,
contributions of constraints etc.) so that I cannot use sander from amber7. I
have been managing OK up unitl now with the amber6 version of gibbs compiled
with either g77 or pgf77.

However, we recently got some new P4 machines here and I had noticed around
the place that lots of people had good experiences with the Intel Fortan
Compiler on these systems. I thought I'd give it a go.

As a first attempt I used the Machine.ifc file from the amber7 distribution
with the code and Makefile from amber6 to see what would happen. Most modules
compiled fine except that gibbs (which is primarily what I need at the
moment) failed.

I get through gib.f and mdread.f but in giba.f I get a bunch of errors like:

Error 173 at (4857:_giba_.f) : Array element subscript is not within array
bounds
      EKSOLT = ENER(3)/fac(2)
                    ^
Error 173 at (4859:_giba_.f) : Array element subscript is not within array
bounds
                           ^
Error 173 at (4859:_giba_.f) : Array element subscript is not within array
bounds
      EKSOLV = ENER(4)/fac(3)

etc.

90 Errors
compilation aborted for _giba_.f (code 1)

I thought I would try the version of gibbs distributed with amber7. This
worked fine but I noticed that -DNOEWALD is specified in the Makefile for
amber7 whereas -DEWALD is the default in the amber6 distribution.

I gave the amber7 distribution a go with -DEWALD and it seems to work (not
extensively tested at this stage but I get an executable).

Which brings me to my questions:

Is it actually OK to use ewald with gibbs in the first place ? In the amber6
manual there is a brief description of how to do it with the old formatted
input like sander_classic (which is what I've been doing). This is gone from
the amber7 manual as is the ability (at least with the Makefile
provided) to use ewald with gibbs. Was this a move in anticipation of moving
all FE calculations to the more efficient sander or does it reflect some
inherent problem with the ewald code in gibbs ?

Assuming it is OK, does anybody know why I get the above errors and if there
is a way to get rid of them (I would have a slight preference for remaining
with amber6 at the moment).

Incidentally, when I was experimenting with resizing the amber6 gibbs code I
found that setting  the parameter (ewmxil) to certain large values (with g77)
gave me similar errors to above. i.e.:

Array element value at (^) out of defined range
_giba_.f:4859: warning:
         EKSOLT = ENER(3)/fac(2)

even though I got an executable in this case. I didn't yet need such a big
nonbond list for ewald but I may. How serious are these "Warnings" ?

Amber7 gibbs with -DEWALD and large values of ewmxil (with ifc) seems to work
OK (again not really tested but I get an executable ad no errors). Anybody
got any idea as to what is happening here ?

Thanks for your time.

David.
 

---------------------------------------
Dr. David Smith
Department of Chemistry
Ludwig Maximilians University
Butenandt-Str. 5-13, D-81377 Munich
Germany
Tel.: +49 (0)89 2180 7740
Fax.: +49 (0)89 2180 7738
e-mail: David.Smith_at_cup.uni-muenchen.de
---------------------------------------
Received on Tue Feb 04 2003 - 08:49:53 PST
Custom Search