Open Radar

 
Summary:
SignedIntegerType and UnsignedIntegerType have a mechanism to convert between their various types: all of them can be converted to, and constructed from, a 64-bit type, which is typealiased to IntMax or UIntMax. FloatingPointType should have a similar mechanism to convert via Float80.

Steps to Reproduce:
1. Declare a generic function (here, an operator) that converts between any two floating-point types:

    prefix func ^^ <FromFloat: FloatingPointType, ToFloat: FloatingPointType> (x: FromFloat) -> ToFloat {

2. Try to define it:

        return ToFloat.from(x.toFloatMax())

Expected Results:
You can.

Actual Results:
You can’t. This behavior is available on integers but not on floats.

Notes:
I encountered the need for this feature while trying to implement an explicit conversion operator. (The idea is, when you’re writing a math operation, you mark which operands should be converted, but let Swift infer the exact types it should convert them to.)