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

970418: fortran interface

Joerg,

>Date: Wed, 16 Apr 1997 21:25:57 -0700 (PDT) 
>From: Joerg Kaduk <address@hidden>
>Organization: Stanford University
>To: address@hidden
>Subject: fortran interface 
>Keywords: 199704170425.WAA11649 

In the above message, you wrote:

> Hi Steve,
> things got much better. After removing a couple of stupidities on
> my side and compiling everything with f77 instead of f90 it
> seems to work! f90 doesn't like the byte statement and gets totally
> mixed up.

We're still working on netCDF-3 and haven't ported it to FORTRAN-90 yet.

> I guess I have to look into the CRAY stuff for that
> since it seems to work for a c90 under f90...
> I think I'm now going to get somewhat used to netCDF. However if
> you have a suggestion to make f90 working on the sun - I'd be very
> happy to hear about, but I guess that's got somewhat lower priority
> in my schedule - so I will not start trying getting it to work right 
> tomorrow... Although as I thnk about that: since I eventually hope
> to run the model on a cray j90 or c90 ...
> is netCDF on the crays writing IEEE as well? 32 or 64 bit? I.e. how
> do I have to specify the types on a cray if I want 64 bit floats?

The external format of a netCDF floating-point value is always IEEE.  It
is 32-bit for NF_FLOATs and 64-bit for NF_DOUBLEs.

> Is there anything else perticular to the crays except the NETCDF_FFIOSPEC
> variable?

The bottom-level I/O layer is different (but transparent to you).
Tuning the NETCDF_FFIOSPEC environment variable can make an order of
magnitude (at least) difference in the I/O speed.

> Thanks again very much!! That was helpful and brought me quite a step
> forward! Now I can finally go on and finish the data output for the
> model run... 
> regards
> Joerg
> --
> address@hidden

--------
Steve Emmerson   <address@hidden>