[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Netcdf on Nec SX-5

Marc,

[In future correspondence, would you please configure your mailer to
avoid unnecessary blank lines and "="s at the end of lines.  Thanks.]

>Date: Mon, 06 Dec 1999 19:31:54 +0100
>From: Marc Guyon <address@hidden>
>Organization: .
>To: address@hidden
>Subject: Netcdf on Nec SX-5
>Keywords: 199912021756.KAA15510

In the above message, you wrote:

> many thanks for your reply, I followed your instructions, excepted that I=
>  =
> prefered to avoid optimisation options and to use debug options for the C=
>  =
> (-h0,ansi -g) and the f90 (-Cdebug) compiler : the test results are the s=
> ame =
> with the same message of compilation without the vectorization warning...=
>  I =
> didn't build the c++ interface because it is safe...
> 
> so the the different variables are :
> 
> uqbar-ssos555 : echo $CC
> cc
> uqbar-ssos555 :  echo $FC
> f90
> uqbar-ssos555 : echo $CFLAGS
> - -h0,ansi -g
> uqbar-ssos555 : echo $FFLAGS
> - -dw -dW -Cdebug
> 
> the A, B, C, D, E, F and G item have been placed in the compressed (gzip)=
> tarfile attached above :
> 
> mira : ls
> VERSION        configure.log  path.log       uname.log
> config.log     make.log       test.log
> 
> mira : pwd
> /tmp/Netcdf.Log
> 
> mira : =
> 
> 
> Here, there is the true result of the ncgen test, I suppose you are right=
>  and =
> 
> the differences are due to numbers representation and machine precision..=
> =2E
> 
> *** ncgen -b test successful ***
> =2E/ncgen -c -o ctest0.nc c0.cdl > ctest.c && \
> cc -o ctest -h2,float0,ansi  -I../libsrc -I.  -DNDEBUG ctest.c ../libsrc/=
> libnet
> cdf.a  && \
> =2E/ctest && \
> =2E./ncdump/ncdump -n c1 ctest0.nc > ctest1.cdl
> 21c21
> <               i:d =3D -1.e+308, 0., 1.e+308 ;
> - ---
> >               i:d =3D -9.99999999999999e+307, 0., 9.99999999999999e+307=
>  ;
> 101c101
> <               :Gd =3D -1.e+308, 0., 1.e+308 ;
> - ---
> >               :Gd =3D -9.99999999999999e+307, 0., 9.99999999999999e+307=
>  ;
> 128c128
> <  dr =3D -1e+308, 1e+308 ;
> - ---
> >  dr =3D -9.99999999999999e+307, 9.99999999999999e+307 ;
> 140c140
> <  d1 =3D -1e+308 ;
> - ---
> >  d1 =3D -9.99999999999999e+307 ;
> 152c152
> <  d2 =3D -1e+308, 1e+308 ;
> - ---
> >  d2 =3D -9.99999999999999e+307, 9.99999999999999e+307 ;
> 164c164
> <  d3 =3D -1e+308, 0, 1e+308 ;
> - ---
> >  d3 =3D -9.99999999999999e+307, 0, 9.99999999999999e+307 ;
> *** ncgen -c test failed ***
> gmake: *** [c-test] Error 1

Yes.  I wouldn't worry about the above "errors".  They have more to do
with printing capabilities than an actual problem.

In your file "test.log", the following occur:

    *** Testing nf_get_var1_int1 ... 
     Range error: No error
     Range error: No error
     Range error: No error
     Range error: No error

    *** Testing nf_get_var_int1 ... 
     Range error: No error
     Range error: No error
     Range error: No error
     Range error: No error

    *** Testing nf_get_vara_int1 ... 
     Range error: No error
     Range error: No error
     Range error: No error
     Range error: No error

    *** Testing nf_get_vars_int1 ... 
     Range error: No error
     Range error: No error
     Range error: No error
     Range error: No error

    *** Testing nf_get_varm_int1 ... 
     Range error: No error
     Range error: No error
     Range error: No error
     Range error: No error

    *** Testing nf_get_att_int1 ... 
     Range error: No error
     Range error: No error

Based on the commentary in the INSTALL file regarding the SX-4 computer,
I belive that the above error may be safely ignored.  The "errors"
come from reading in an NF_BYTE value that shouldn't be possible, but on
your system, NF_BYTE-s are 16-bit integers and have a greater range than
the test assumes.

To sumarize, I belive that your build of this version of the netCDF 
package was successful.

--------
Steve Emmerson   <http://www.unidata.ucar.edu>