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

960426: 2.3.2 vs. 2.4 netCDF question

Hi John,

>Date: Fri, 26 Apr 1996 15:02:19 -0400 (EDT) 
>From: address@hidden (John Sheldon)
>Organization: NOAA/GFDL
>To: address@hidden (Russ Rew)
>Subject: Re: 2.3.2 vs. 2.4 netCDF question 
>Keywords: 199604222201.AA01657

In the above message you wrote:

> Hi again-
> 
> Yesterday, I CC'd you on some mail I sent to Jeff Kuehn about a timing
> code which demonstrated the CP-time problem with pre-filling on our
> Unicos 8 C90 under 2.4.1.
> 
> Here's a modified version of that code that pretty simply demonstrates
> that netCDF is writing one rendition of "1.E35" to the "missing_value"
> attribute and another rendition to the data portion of the file.
> 
> I hope we can get to the bottom of these issues pretty soon.  I'm
> riding what appears to be the crest of a tidal wave of new netCDF users
> here at GFDL, and the fewer reefs I encounter, the better!

We've discovered the problem and have generated a workaround (which
will also be in the next release).  The problem is that the writing of
a floating-point attribute used the CRI-supplied xdr_float() function;
whereas, the writing of a data value used the (also CRI-supplied) IEG
parallel routines.  Unfortunately, CRI didn't make these routines behave
the same: for certain values, they create different XDR bit-patterns.

We've modified the code to always use the IEG routines when appropriate.
You'll find a patch for the file libsrc/array.c enclosed.

Please let us know if this helps.

Regards,
Steve Emmerson   <address@hidden>

--------Begin patch
Index: array.c
===================================================================
RCS file: /upc/share/CVS/netcdf/libsrc/array.c,v
retrieving revision 1.54
diff -c -r1.54 array.c
*** 1.54        1995/10/13 19:28:51
--- array.c     1996/04/26 20:50:11
***************
*** 561,571 ****
                xdr_NC_fnct = xdr_nclong ;
                goto loop ;
        case NC_FLOAT :
!               xdr_NC_fnct = xdr_float ;
!               goto loop ;
        case NC_DOUBLE : 
!               xdr_NC_fnct = xdr_double ;
!               goto loop ;
  /* private types */
        case NC_STRING :
                xdr_NC_fnct = xdr_NC_string ;
--- 561,571 ----
                xdr_NC_fnct = xdr_nclong ;
                goto loop ;
        case NC_FLOAT :
!               xdr_NC_fnct = xdr_floats ;
!               goto func;
        case NC_DOUBLE : 
!               xdr_NC_fnct = xdr_doubles ;
!               goto func;
  /* private types */
        case NC_STRING :
                xdr_NC_fnct = xdr_NC_string ;