[MITgcm-support] RE: a few MITgcm questions
Ichiro Fukumori
fukumori at jpl.nasa.gov
Sat Apr 3 20:46:19 EST 2004
Thanks. FYI, the -D_BYTESWAPIO option does not seem to work for
sequential binary files.
-- Ichiro
At 08:40 PM 4/3/2004 -0500, Chris Hill wrote:
>Ichiro,
>
> -r8 should not be needed, in fact g77 doesn't support it, so the core code
>has to work ok without it. However, it may not have been tested on the Altix
>lately. Because of g77, for the kernel code it never matters, but there
>could be some stuff in KPP where its assumed (Dimitris any thoughts?).
>
> For reading files could you for now do
>
> real*4 KFtheta(1-OLx:sNx+OLx,1-OLy:sNy+OLy,Nr,nSx,nSy)
> _RS KFtmp(1-OLx:sNx+OLx,1-OLy:sNy+OLy,Nr,nSx,nSy)
> character*160 KFthetaFnam
> CALL MDSREADFIELD(KFthetaFnam,precFloat32,'RS',Nr,KFtmp, 1, myThid)
> Kftheta = KFtmp
>
> that should read a 510*6,510 file for your setup that is a straight single
>record direct access form with 32-bit reals.
> Depending on
>
> 1 - compiler options
> 2 - settings of -D_BYTESWAPIO
>
> the input will need to be either little-endian or big-endian
>
> The build options that Dimitri has for the Altix and the Origin both are
>set for big-endian (that's the native format on MIPS but not on Itanium).
>
>It would probably be good to have a 'R4' and 'R8' MDS arg but I have to head
>off to Ames soon so I won't have time - I'll be in touch Monday.
>
>Chris
>
>
>
>
>
>
> > -----Original Message-----
> > From: Ichiro Fukumori [mailto:fukumori at jpl.nasa.gov]
> > Sent: Saturday, April 03, 2004 6:58 PM
> > To: cnh at MIT.EDU
> > Cc: bnc at pacific.jpl.nasa.gov; Dimitris Menemenlis; chenze at nas.nasa.gov
> > Subject: a few MITgcm questions
> >
> > Hi Chris,
> >
> > Is compile option -r8 necessary? It appears to interfere
> > with using sgemv (a single precision BLAS matrix times vector
> > subroutine).
> >
> > I have a 3D array on the cubed sphere grid that I need to read in.
> > Could you tell me the format I should write it to and the
> > subroutine call to read it. Presently I have,
> > real*4 KFtheta(1-OLx:sNx+OLx,1-OLy:sNy+OLy,Nr,nSx,nSy)
> > character*160 KFthetaFnam
> > CALL MDSREADFIELD(KFthetaFnam,precFloat32,'RS',Nr,KFtheta,
> > 1, myThid) Does MDSREADFIELD work with real*4 array (as
> > opposed to _RS)?
> >
> > Thanks,
> > Ichiro
> >
> >
> > ===============================================================
> > Ichiro Fukumori | Jet Propulsion Laboratory
> > e-mail: fukumori at jpl.nasa.gov | Mail Stop 300-323
> > phone: +1 (818) 354-6965 | 4800 Oak Grove Drive
> > fax: +1 (818) 393-6720 | Pasadena, California 91109, U.S.A.
> > ===============================================================
> >
===============================================================
Ichiro Fukumori | Jet Propulsion Laboratory
e-mail: fukumori at jpl.nasa.gov | Mail Stop 300-323
phone: +1 (818) 354-6965 | 4800 Oak Grove Drive
fax: +1 (818) 393-6720 | Pasadena, California 91109, U.S.A.
===============================================================
More information about the MITgcm-support
mailing list