@Charlieface One thing that comes to mind is performance. The double computation may be more costly. I don't know whether there are implementations without floating point hardware out there but on such systems it may be prohibitive.
↧