[cam-users] virtual memory

[Alexander Sen Gupta]

Hi Peter,
The the IT guys that run the server, found a simpler solution. It was just
a matter of going to a different compiler version. The magic command
was:
setenv DECF90 /usr/lib/cmplrs/fort90/hot-f90-3112-48D3A
if that means anything to anyone.

So now the whole code compiles. Of course on executing it crashes after a few seconds
but thats the challenge for next week.

Thanks again for your help.
Alex


On Wed, 19 Nov 2003, Peter Paul Smolka wrote:

> On Wed, 19 Nov 2003, Alexander Sen Gupta wrote:
> 
> > Date: Wed, 19 Nov 2003 16:28:17 +1100 (EST)
> > From: Alexander Sen Gupta <alexg@maths.unsw.edu.au>
> > To: cam-users@ucar.edu
> > Subject: [cam-users] virtual memory
> >
> > Hi,
> > Im a new CCM user in need of some help. Im trying to compile the coupled
> > model on an alpha server. After some minor tweaking compilation proceeds
> > through the various components (in a few 10s of seconds), but on reaching
> > the ocean component, the virtual memory requirement starts creeping up.
> > After about an hour the compliation crashes with the message:
> > Fatal:  Virtual address space exceeded.
> > gmake: *** [initial.o] Error 1
> 
> Dear Alexander,
> 
> With ccm3.6 a similar phenomenon could be observed when compiling
> for improved optimization all files into one file so cross procedure
> optimization can take place.
> 
> In this case the solution was:
> 
> Separating parts of code out e.g.:
> 
> Not the one extreme, compiling each program separately and then
> linking but also not the other compiling all into one large object
> file.
> 
> Potentially you generated this effect indirectly.
> 
> Checking your explicite and implicite compiler options as well
> as scaling back optimization one or two steps for that part might be
> useful. One of the Alpha compilers I know has these effects.
> 
> By the way:
> 
> A question:
> 
> Do you or anybody else have ESMC in Fortran? E.g. not
> only the ESMF part linked to *.C programs but the *.C programs
> themselves?
> 
> If yes I ll save the recoding.
> 
> The same applies in principle to netCDF. I am quite sure it does
> not exist in Fortran but sometimes luck exists.
> >
> > The virtual memory requirement is in excess of 2GB!
> > Is this memory requirement normal? If not, any ideas what might be causing
> > the build up?
> >
> > Thanks for any help.
> > Alex
> >
> 
> Best regards, Peter
> 
> 
> 
> **********************************************************************
> Dr. Peter P. Smolka
> University Muenster
> Geological Institute
> Corrensstr. 24
> D-48149 Muenster
> 
> Tel.: +49/251/833-3989   +49/2533/4401
> Fax:  +49/251/833-3989   +49/2533/4401
> E-Mail: smolka@uni-muenster.de
> E-Mail: PSmolka@T-Online.de
> **********************************************************************
> 
> _______________________________________________
> cam-users mailing list
> cam-users@ucar.edu
> http://mailman.ucar.edu/mailman/listinfo/cam-users
>