I'm currently doing a project that involves 2 separate pic processors on 2 separate boards doing completely different things. One of the boards is more avionics based and one of its tasks is to process GPS data to establish a required heading and distance to a destination point. I'm using the Cordic tools posted here by Walter a few years ago to help with the trig processing.
Also, I like the idea of transitioning from previous 28 pin PICs to the 18F26K83 due to numerous virtues it has; however one thing I discovered attempting to implement it for this particular application is that the 18F26K83 and the Cordic routines don't see eye-to-eye and produce erroneous results. Well, for some reason the results (particularly for atan2) always appear to exceed the maximum.
If I change to an 18F26K22 with exactly the same code (except the osc settings which are different from the 28K83), the routines work perfectly. This is just straight math processing, no external inputs to factor for this debugging.

Troy