Re: CRC slicing-by-8

[Antonio Diaz Diaz]

p.z.l wrote:
> Have you got any idea why this:
>      c = crc32[(c ^ buffer[i]) & 0xFF] ^ ( c >> 8 ); /// lzip
> is faster than this:
>      c = (c >> 8) ^ crc32[(c & 0xFF) ^ buffer[i]]; /// sbrumme crc

May be because of the different ordering of the operations. You may try to 
transform one of the lines above into the other step by step, and test the 
speed of the intermediate stages.