Since this thread is coming to an end, yet may come up as the result of some search, for the record, the polynomial used in the VHDL is:

X¹⁶ + X¹⁵ + X²+ 1

which is used in (among others) DDCMP and MODBUS protocols, but is definitely not:

X¹⁶ + X¹² + X⁵+ 1

which is the so-called CCITT/ISO/ITU/'etc.' that was the primary focus of this thread.

Nevertheless, Jez's referenced code shows one approach (although not without his caveats) to implementing CRC calculations in hardware code (Hehe, "hardware code" I am becoming more accustomed to that term day-by-day). I merely wanted to point out the difference in the polynomials so nobody gets "tripped up" along the way.