[Mitgcm-support] Re: fftpack.F (PR#17)

[mitgcm-support at dev.mitgcm.org]

chris wrote:
> 1. #ifdef ALLOW_
>   as a complete wrapper for code leaves blank files.

We don't do this anymore. #ifdef ALLOW_ are arranged to
create dummy routines if #undef'd.  

> 
> 2. At some stage we will want to use an FFT on the Sun so
>    then I guess it will get fixed?

fftpack.F needs to be F77ized.

> 3. Is zonalfilt designed to allow any othe FFT library?

Yes - but it needs to know the particular packing of modes
in spectral space which changes from FFT to FFT. Also the normalization
is sometimes different.

> 4. How come zonalfilt isn't disabled by default.

Why should it be disabled when gmredi and kpp are "enabled" by default?

I think a "clean" package should be compilable whether it's
used or not and if not used it should not affect the results
and that allows all code to be enabled by default. I think you
want the reverse to be default.

We haven't thrashed this issue out yet...