On Sat, Jul 9, 2011 at 2:39 PM, Huan Truong <[email protected]> wrote:
> On Sat, 09 Jul 2011 20:16 +0100, "Alex Horn" <[email protected]> > wrote: > > > It's much more important to be near the left of the chart (i.e. fast) > than > > > it is to be near the top of the chart (i.e. space efficient) > > http://phantasie.tonempire.net/t96-compression-benchmark#149 > > The charts over here is much easier to see and to figure out what's the > best choice. > Maybe, though I think the table data is more useful. For example, what is more important encoding speed or decoding speed? If your interest is in transferring a disk image over the network, then the answer is the "slowest" of the two, since the encoder will run at one end at the same time that the decoder runs at the other and the transfer will bottleneck at whichever end is slower. If you reorder the table by minimum of encoding and decoding throughput and take the max (why does this sound like game theory class?) then the best algorithm for piping an image over the network would be: LZP2 followed pretty closely by QuickLZ-L1 and LZBW1 and a bit farther by tornado-1 and LZ4. Gzip is way at the bottom, which agrees with anecdotal experience. I wish the had include lzop in the table. I'd have like to see how it fared in this kind of analysis. Don
