TAK codec (Any news, seems silent for some time) Options

[johnsonlam]

Hi Guys,

Thomas seems going silent for a period of time, any news or other developers working on TAK related GUI or something like "TAK drop"?

Also want to ask, any media player can DIRECT decode the HDCD signal inside TAK yet?

Thanks.

[TBeck]

Yes, it became a bit silent. Sometimes i am missing those early days, when i released many test versions and was impatiently waiting for results to see, if the latest optimizations worked well. This was exciting for me and for many testers. Yeah, this was some kind of entertainment!

BTW: I am reading the hydrogen posts on a daily base, but i usually don't have time to answer immideately. Well, there are also a lot of unanswered mails too... Sorry.

As a result of the extensive tuning before the first stable release, there was not much opportunity left to tune the codec further. At some point it's getting more and more difficult to optimize. New processors with new instructions sets can provide opportunities to improve the speed, but the (in some post) mentioned new AMD cpus can't help TAK. I always had possible hardware implementations in mind, and this usually means fast integer implementations, to my knowledge the new AMD cpus only provide improvements for floating point optimizations. But when Intel releaseses their new cpus with AVX-256-Bit integer instructions, i will jump on the train...

So, what was left to do:

V 2.0 (which required a format change) provided relatively small improvements, but the effort to achieve them was quite large. And it would be even more (time wise) expensive to achieve more improvements. It's getting harder and harder. Another point: I am still looking for opportunities to improve the compression ratio, but all i have tried had a very significant effect on the encoding speed (10 to 50 times slower). That's not a good thing, if you want your codec to be associated with speed. Notice: I am still trying a lot, but because nothing worked very well, i don't post about it.

Conclusion: Currently little chance to improve the CD-Audio-compression performance.

Therefore i worked on other things:

V 2.1 initially included a specialized codec for LossyWav-compressed files. But there was very little interest (understandable, if you want to use such files to replace conventionally lossy compressed ones, you usually need hardware playback capabilities, which TAK is lacking).

V2.2 brought the probably most efficient (if you also take speed into account) compression for multi-channel-audio. But that's also only relevant for a small group of users. Possibly more for the video guys, but they would need an open source version of TAK integrated into their video tools.

Currently i have very liitle time i could invest into TAK. To do so, the task has to be attractive. Well, if i get a new exciting idea...

But i still love TAK (developers are allowed to say something like this about their work, i hope) and i definitely want it to have a great future. I am thinking about strategies to spread it. I have some plans, but i don't want to talk about them, this because of bad experiences in the past, when i was to enthusiastic and made promises i couldn't keep. You will hear about it, when i'ts done. And i hope, it will get a bit better than the latest Duke Nukem game...

Thomas

[Porcus]

Another point: I am still looking for opportunities to improve the compression ratio, but all i have tried had a very significant effect on the encoding speed (10 to 50 times slower). That's not a good thing, if you want your codec to be associated with speed. Notice: I am still trying a lot, but because nothing worked very well, i don't post about it.

If you still after these years see the fun in finding room for improvement, then ... did you use much noise music in development? I checked my least-FLACable album and it turned out to be even less TAK-able: http://www.hydrogenaudio.org/forums/index....st&p=800823 . Of course it is only one album, and that's not much of a statistic. Anyway, some figures for the least compressible track (#3, now playing):

1307 for WavPack extra high (not holding my breath)
1395 for FLAC -8
1399 for FLAC -0
1400 for WavPack fast
1402 for WavPack normal (!)
100.04 % for TAK -p4m, reported by Tak.exe (2.2.0) -- indeed, file sizes are 58 178 123 bytes for .wav and 58 198 718 for .tak.
100.96 % for TAK -p0


Cheers from YALAuwivibusoC #666
(Yet Another Lossless Audio user who is very impressed but uses some other Codec)

This post has been edited by Porcus: Sep 23 2012, 15:03