764N/AThis is my implementation of RIPEMD-160. The pentium assember is a little
919N/Aoff the pace since I only get 1050 cycles, while the best is 1013.
919N/AI have a few ideas for how to get another 20 or so cycles, but at
919N/Athis point I will not bother right now. I belive the trick will be
919N/Ato remove my 'copy X array onto stack' until inside the RIP1() finctions the
919N/Afirst time round. To do this I need another register and will only have one
764N/Atemporary one. A bit tricky.... I can also cleanup the saving of the 5 words
919N/Aafter the first half of the calculation. I should read the origional
919N/Avalue, add then write. Currently I just save the new and read the origioal.
919N/AI then read both at the end. Bad.