std.complex

[Ola Fosheim Grøstad" <ola.fosheim.grostad+dlang at gmail.com>]

On Thursday, 2 January 2014 at 19:58:27 UTC, Joseph Rushton 
Wakeling wrote:
> I agree that should be supported. I just think that template 
> constraints should only be relaxed in the context of concrete 
> test cases, otherwise you're making promises you don't know you 
> can keep.

Yes, that is very true. Fixed-point is very sensitive to 
precision-loss so it requires some careful thinking, so it is 
definitively not a drop-in replacement type.

(It is actually useful on x86 too when you want reproducible 
results across binaries and hardware, such as in peer-to-peer 
syncing of computed state.)