Well, as I was working on getting the matrix math to work, I have hit my head again and again against a numerical accuracy brick wall. It turns out that, even though 64-bit unsigned integers are good for most precision work within MPGuino, they just can't seem to cut it with regard to performing the needed matrix math without losing accuracy. Adding signed number support would prove to be a coding nightmare, too.
Therefore, I am re-writing SWEET64 to get rid of 64-bit integers, and use 64-bit floating-point numbers instead. Since
most Arduino-specific libraries do not truly support double precision floating point numbers, I am in the process of writing code that will do this economically.
dcb did something similar to this, when he wrote the original 64-bit integer routines for MPGuino - the code was roughly 40% of the size of the native unsigned long code, yet did the job.