Re: [Libjpeg-turbo-users] Position statement regarding mozjpeg
SIMD-accelerated libjpeg-compatible JPEG codec library
Brought to you by:
dcommander
From: Roger P. <rog...@gm...> - 2014-04-28 21:40:34
|
On 3/9/14, DRC <dco...@us...> wrote: > For those who have asked about the mozjpeg project and why that code > wasn't integrated "upstream" (libjpeg-turbo is "upstream" in this case), > I prepared a position statement with further details: > > http://www.libjpeg-turbo.org/About/Mozjpeg > > Executive summary: > > -- The jpgcrush code breaks ABI compatibility, so it would either have > to be reworked using new get/set methods (a lot of trouble), or it would > have to be #ifdef'ed, which means that Mozilla would have to maintain > their own build anyhow. Because of the ABI compatibility issue, the > jpgcrush code, as currently written, could not be enabled by default in > the official builds of libjpeg-turbo, nor could it be enabled in the > builds that are deployed by various O/S distributors. > > -- My performance results indicate that most of the 10% improvement in > compression ratio that Mozilla is claiming comes from simply using > progressive encoding. The addition of jpgcrush relative to "plain" > progressive encoding is only about a 2% difference, on average (and no > more than a 4% difference was observed in the best case.) Maybe if they have some optimizing commits (where possible) you can just cherry-pick them? There may be some interesting commits in there :) https://github.com/mozilla/mozjpeg/commits/master |