First I must say, I know nothing about space vector modulators etc.
As said, I doubt you will find a straighforward path from matlab to a '51. Even having generated some C code, modifying it for the peculiarities of the 51 may turn out to be as laborious as rewriting it totally, based on the ideas from the matlab program.
But first, a couple of questions on the matlab program (the source would be better, if you can post it):
- how many variables are you using and what is the required precision?
- do you use complex numbers?
- do you use matrix arithmetics?
- do you perform calls to library or built-in functions?
- how fast should the calculation be, or, how much time is between consecutive pulses?

Answers to these questions may give you (or us) an idea on complexity of your problem.

By the way, I don't know how comes an EPROM to matlab calculations; I suppose you mentioned is as you have already seen a construction similar to yours, can you give more details?

Jan Waclawek