Re: Cache Digest Diffs, first results

On Wed, 11 Aug 1999, Henrik Nordstrom wrote:

> I have now completed a initial round of simulations of "Huffman encoded
> bit change distance".
>
> Initial results are:
> %bits changed %improvement from cd_diff
> 0.50% 39%
> 0.60% 41%
> 1.05% 46%
> 1.70% 50%
> 3.02% 54%
> 5.00% 57%
> 7.02% 58%
> 10.00% 58%
>
> Known limitations of this initial version:
> * Can't handle "no changes". It will give false results if changes are
> further apart than 2^16 bits.
> * It is a simulator, and I have not yet verified that encoding+decoding
> gives correct output. It should, but I cannot make any guarantees until
> I actually have written and tested an complete encoder+decoder.

Henrik,

        Your preliminary results look great! I can only hope that the
implementation will be as effective. Please post the final version when it
is ready.

Thanks,

Alex.
Received on Tue Jul 29 2003 - 13:15:59 MDT