[MITgcm-support] reading error

[chris hill]

Hi Paola et. al,

 I think Patrick's fix should work. 
 The issue Ed raises is only needed for experiments with the "forcing is
periodic in time flag" set true - which it aint for this experiment.

Chris
On Mon, 2004-11-08 at 18:23, Ed Hill wrote:
> On Mon, 2004-11-08 at 17:42, Paola Cessi wrote:
> > Hi - I am trying to run the MITgcm_checkpoint54 version on a 
> > 256-processors ALTIX (SGI) at ORNL, but at runtime I get an error trying 
> > to read the input binary file 'Qnet.circle', generated by a local fortran 
> > file in real*8 words (I get the same problem if the output is real*4).
> > 
> > mitgcmuv >output.txt
> > forrtl: severe (36): attempt to access non-existent record, unit 9, file 
> > /autofs/dfs/users/u001/cessi/MITgcm/verification/exp5/build/Qnet.circle
> > Image          PC                  Routine       Line     Source
> > mitgcmuv       0x40000000002f4630  Unknown       Unknown  Unknown
> > mitgcmuv       0x400000000025d640  Unknown       Unknown  Unknown
> > mitgcmuv       0x400000000025deb0  Unknown       Unknown  Unknown
> > mitgcmuv       0x4000000000294260  Unknown       Unknown  Unknown
> > mitgcmuv       0x4000000000291f30  Unknown       Unknown  Unknown
> > mitgcmuv       0x400000000003cb20  Unknown       Unknown  Unknown
> > mitgcmuv       0x40000000000c64a0  Unknown       Unknown  Unknown
> > mitgcmuv       0x40000000001f6f20  Unknown       Unknown  Unknown
> > mitgcmuv       0x4000000000232120  Unknown       Unknown  Unknown
> > mitgcmuv       0x400000000021e510  Unknown       Unknown  Unknown
> > mitgcmuv       0x400000000024bc00  Unknown       Unknown  Unknown
> > mitgcmuv       0x400000000024c1f0  Unknown       Unknown  Unknown
> > mitgcmuv       0x4000000000160870  Unknown       Unknown  Unknown
> > mitgcmuv       0x4000000000003b50  Unknown       Unknown  Unknown
> > libc.so.6.1    0x20000000001fe990  Unknown       Unknown  Unknown
> > 
> > ------------------------------------------------------------
> > 
> > I am using the following optfile to compile (non MPI, for now)
> > 
> > CPP='/usr/bin/cpp  -traditional -P'
> > FFLAGS='-mp -132 -r8 -i4 -w95 -W0 -WB '
> > FOPTIM='-O3 -align'
> > LIBS='-L/opt/intel/efc8046/lib'
> > FC=ifort
> > DEFINES=' -DWORDLENGTH=4'
> > INCLUDES='-I/opt/intel/efc8046/include'
> > INCLUDEDIRS='/opt/intel/efc8046/include'
> > CC='icc -static'
> > 
> > Any suggestion about what I am doing wrong?
> 
> 
> Hi Paola,
> 
> Are you certain that the "Qnet.circle" input file you have generated is
> the right size for your domain?
> 
> Looking in "MITgcm/model/src/external_fields_load.F" it seems that the
> file needs to be at least twice the size of the surface of the domain
> since it is interpolated in time.  So for one of our verification
> experiments ("exp5") that uses this input file, we would expect a file
> size of:
> 
>   sNx * sNy * nSx * nSy * nPx * nPy * WORDLENGTH * 2
>     = 32*32*2*2*1*1*4*2
>     = 32768
> 
> which agrees with the size of the input file (in "data", we see that
> "surfQfile='Qnet.circle'") which is provided.
> 
> And if I've missed something here, I'm hoping someone else will correct
> me...
> 
> Ed