[ferret_users] Why single precision?

["Ingo Bethke (uib)" <Ingo.Bethke@xxxxxxxxxxxxxx>]

Dear FERRET team,

The FERRET user guide states that all variables are converted to "float" 
(in the guide defined as 32-bit) internally as they are read by Ferret.

Does this mean that also all computations are performed in single 
precision?

I fully agree that single precision should be sufficient for the storage 
of the final result. However, in some critical cases it could be 
beneficial to store the intermediate results with higher precision (at 
least it would save the user to worry about the numerical accuracy...). 
Since cpu-memory has more than doubled during the last years I wonder if 
you had any plans to change the internal format to double precision (64bit)

Best regards,
Ingo Bethke