Software floating point library?

[Ivan Voras]

Does anyone know of a simple, BSD (or compatible) licensed software
math library that emulates floating point numbers? Ideally, it should
be a single .c file library that implements basic operations (+, -, *,
/, negation, comparisons, convert to/from machine integers and
strings), that uses a standard C type (int or long long) to store the
value, and is usable in the kernel.

Alternatively, how about a fixed-point library with the same
properties? (I know it's easy to do, but don't want to reinvent the
wheel again if I don't have to).

Please disregard my earlier message on this subject, if you received it.