A good starting point with this processor is the example program ecdh2m.c

This runs "out-of-the-box", with no modifications required, on the 
VisualDSP++ simulator. Takes about 17 Million clock cycles (release mode) 
to complete both sides of a Diffie-Hellman key exchange using C-only code.

The Blackfin is a 16/32 bit processor. For GF(2^m) applications it is best 
considered 32-bit, but over GF(p) it is probably best to consider it as a 
16-bit processor. 

There is a macro file blackfin.mcs which can be used to build the application 
ecdhp16.c using VisualDSP++. Takes about 9 Million clock cycles to complete
both sides of a Diffie-Hellman key exchange.