@stephentyrone do you plan to tackle this?

I can make some time for it early next week. Providing default implementations will require the new integer stuff, and I'm not sure we have a good way to support Float80 without the importer being able to map long double, but we can add the API to FloatingPoint and wire up Float and Double quickly enough.

Assigning this to myself so that it has an owner; if someone else wants to jump on it, however, feel free.

What do you mean that the importer can't map long double? Long double on X86 should come in as Float80.

I agree that it should, but it doesn't seem to at present; there are some TODOs scattered in swift around this. I filed a radar yesterday when I became aware of the issue.

Draft implementation (with horrible Float80 workaround) on Github. This is not an optimized implementation by any stretch of the imagination, but should get these working.

#3443

Merged in 9c08fbf.

Optimization opportunity: https://bugs.swift.org/browse/SR-2045