[SCore-users-jp] Re: [SCore-users] charmm under SCore

[Bogdan Costescu]

Sorry for the late reply, I had a very busy end of week...

On Thu, 24 Oct 2002, Nick Birkett wrote:

> Hi I am having a problem compiling charmm (both  c25b2 and  c28b2 versions).

Hmm, c25b2 is so old, but I'm quite surprised about c28b2. I think that I 
haven't got any similar error when compiling it - I'm using RedHat Linux 
7.2 with all the updates.

> If I use the gnu compiler with -O2 (mpif77) I get a compiler error (g77):

I assume that you are using a 2.9x gcc/g77.

> nmdimb.f:1215: Internal compiler error in make_edges, at flow.c:1071

That's the reason why the "FC0" variable is set in the Makefile. It's the 
compile line used to disable optimizations for files where optimizations 
are known to screw up the compilation; gcc 3.x (I don't know exactly 
from what "x", but 3.2 seems to work) fixes this. In the attached 
Makefile, FC0 is not set properly. And keep in mind: Makefiles can be 
different in different CHARMM versions, so don't just copy the Makefile 
from c25b2 to c28b2 (I would actually copy it in the other direction !).

> parset.f: In function `nnods':
> parset.f:29:
>          NNODS = NUMNOD
>                  ^
> Invalid declaration of or reference to symbol `numnod' at (^) [initially seen 
> at (^)]
> parset.f: In function `mnods':
> parset.f:49:
>          MNODS=MYNOD

I've never seen these errors in recent versions of CHARMM, so I assume 
that this is still c25b2. I can't give you any advice here, as I don't 
think that I ever compiled pre-c26 versions. However, from what I 
remember, c25 was the first version to work in parallel on Linux and you 
still needed to do some tweeking in c26, so maybe this was just fixed in 
some later version.

> This is my gnu pref.dat :

You should keep in mind that not all modules of CHARMM work in parallel. I 
usually compile here only the "lite" version in parallel.

OK, so this had nothing to do with SCore :-( You should have probably 
written to the CHARMM users' list or to CCL.

> which i assume is a preprocessing error.

But this reminded me of something that _is_ related to SCore: while a 
normal invocation of the compiler works with something like:

gcc -DOS="LINUX" -o x x.c

where x.c is something like:

#include <stdio.h>
int main(void) {
printf("OS is: %s\n", OS);
}

the SCore MPI wrappers don't, as the command line is parsed by the scripts 
and not sent to gcc in its original form. Can something be done about this ?

-- 
Bogdan Costescu

IWR - Interdisziplinaeres Zentrum fuer Wissenschaftliches Rechnen
Universitaet Heidelberg, INF 368, D-69120 Heidelberg, GERMANY
Telephone: +49 6221 54 8869, Telefax: +49 6221 54 8868
E-mail: Bogdan.Costescu ＠ IWR.Uni-Heidelberg.De


_______________________________________________
SCore-users mailing list
SCore-users ＠ pccluster.org
http://www.pccluster.org/mailman/listinfo/score-users