Burning Candle Software | Creating fast and reliable software, on time.
Hovverd
Forward Scattering - The Weblog of Nicholas Chapman
Hovverd
24 April 2017
Between the Marlin paper coming out and a contract I’m currently working on, I started thinking about compression again a bit more recently. Friday night, I had an idea (it was quite a bad idea) about using a 16 entry Tunstall dictionary for binary alphabet entropy coding (with variable length output up to 8 bits in length) and using a byte shuffle to lookup the codes from a register in parallel when decoding, throwing in an adaption step.
My initial thought was that maybe it would be a good way to get a cheap compression win for encoding if the next input was a match or a literal in a fast LZ (and using a count trailing bits to count runs of multiple literals). I started to dismiss the idea as particularly useful by Saturday morning, thinking it was the product of too much Friday night cheese and craft beer after thinking about compression all week. However (quite quixotically), I decided to implement it anyway (standalone), because I had so
Burning Candle Software | Creating fast and reliable software, on time.
Hovverd
Forward Scattering - The Weblog of Nicholas Chapman
Hovverd