I am working on file compression. I mentioned in a previous thread
that I did better than others. I was wrong. The experiment was
blotted by a bug. Now I'm getting results about as good as PKWare's
deflate/maximum technique. I still have ideas. Again, it may be due
to an error. Again, as mentioned previously, it is very slow. I have
in mind some techniques to speed up the compression. They wouldn't be
that good:
* Store the first 16 or 32 matches for the last 16 words in the
input that are compressible by lz77 (shortening blocks that have
occured before to the location and length of such comparison). If a
hit occurs, simply compare it to the first matches and scan from the
last match on if full.
* Store the last 16 or 32 non-matches. Scan these for the current
input first, and if found, check it for a match to current and remove
the entry.
Again, I don't think these would help much. In fact, on poorly
compressible files, it may take longer. I am looking for other
techniques to speed up this technique. I assume I must minimize the
hits. Any help would be appreciated.