🚀 go-pugleaf

I have a machine that assumes that everything done is signed.
For example, addition always produces a fixed point overflow
exception if the sign is wrong. This makes the unsigned data
type hard to support. For math, I'm just extending everything
to long (or more accurately, letting the compiler do the extending
to long), and doing the signed operation in that mode; this will
fail in the (hopefully uncommon) case of two large unsigned numbers
being added together (which would overflow anyways). 

The real problem is with conditional jumps. The only comparison
operation sets less than, equal, or greater, based on a signed
comparison; the subtract instruction will set carry, but it also
causes an overflow exception.

Any ideas on how to support unsigned data on this machine? (Aside
from turning off the overflow trap, which isn't an option).

		Michael