DES
Managed to get my software implementation of DES down to have a cycle count per block of 3.3% the naive implementation, and a key setup time 40% of the original. Optimising code is fun :-) By the same BOE calculation I used before, my implementation would achieve a theoretical 2.8 Mbytes/sec on a 1GHz processor. There are no 64 bit values used at all in this implementation now. I ran it on a real StrongARM and the speedup is very noticable.
More PhDs
Congrats to Ian who has submitted his draft dissertation and is heading of to work for Altera next week.
