TAK 1.1.0 - Beta release Options

[halb27]

Well, TAK's speed compared to FLAC's as taken in for instance SyntheticSoul's comparisons isn't necessarily the same thing on a mobile DAP.
I think so far we all agree. This doesn't say however that TAK can't be a very fast codec on a mobile player.
It's all theory as long as there is no such player. Until then hope is justified by experience with a PC.

@TBeck: Any plans to contact DAP manufacturers? (Time is tight: It will be few years only that we will see giant flash memory so that everybody can encode his collection the lossless [or lossyWAV] way. It would be great to have TAK then as a widespread codec in case the speed hope remains true).

This post has been edited by halb27: Dec 29 2008, 23:15

[TBeck]

No doubt: looking at compression ratio and important decompression speed TAK is the best player in the lossless field. Congratulations!

Thank you!

a) TAK yields the best combination together with lossyWAV right now. Some months ago you announced specific development for use with lossyWAV to achieve further improvement. What's the current state here?

I will make those specific improvements part of the new codec which will be introduced with TAK 2.0. Currently i am expecting these advantages for LossyWav files:

- At least 10 kbps better compression.
- The compression of audio parts where LossyWav hasn't removed any bits will be about as good as for the standard codec. Currently the compression here is suffering from the small frame sizes used for LossyWav.

b) The impressive decompression speed makes TAK highly desirable for use with mobile DAPs (like FLAC, but with a better compression ratio). Has there been any contacts with DAP producers allowing them to implement TAK and giving them the necessary information? This would be great, especially as the time for lossless codecs on mobile DAPs is yet to come, but we may be close to it right now. It would be great to have TAK be a major player in this field.

@TBeck: Any plans to contact DAP manufacturers? (Time is tight: It will be few years only that we will see giant flash memory so that everybody can encode his collection the lossless [or lossyWAV] way. It would be great to have TAK then as a widespread codec in case the speed hope remains true).

There is contact with one Rockbox member. But i don't think i will release any code before the beginning of the TAK 2.0 line.

Could I suggest that you also change the behaviour of TAKC so that if doesn't report an "invalid file extension" whenever you try to give an output filename an extension anything other than tak? FLAC will let you call the output file anything you want, with no extension added if not present in the output filename.

I will think about it. For this release it's too late, because i would have to perform a significant amount of testing.

Thomas

[TBeck]

Beta 4 has been released

The download link and the history are in the first post.

Probably i could have made this a final release, but i am a bit cautious these days...

Well, if there are no problems with this beta, i will release it as final version this weekend.

There may be some very tiny speed improvements because i have optimized the code alignment of some functions. This shouldn't really matter, but i want to make sure, that accidental misaligment isn't the cause for the slightly worse decoding speed of TAK 1.1.0 in Synthetic Soul's comparison.

I still think that subtle io issues are responsible for the speed drop, but i have deceided to investigate this for the next version, which hopefully will be released in about 3 months.

Thomas

[adamjk]

You are the best!

[Synthetic Soul]

CODE

     |  1.1.0b3       |  1.1.0b4
     |  Enc     Dec   |  Enc     Dec
=====================================
p0   |  130x    141x  |  132x    142x
p0e  |  108x          |  110x
p0m  |   61x          |   61x
p1   |  109x    140x  |  109x    141x
p1e  |   90x          |   91x
p1m  |   51x          |   50x
p2   |   66x    125x  |   67x    125x
p2e  |   53x          |   53x
p2m  |   33x          |   33x
p3   |   38x    111x  |   38x    112x
p3e  |   31x          |   31x
p3m  |   20x          |   20x
p4   |   24x    104x  |   24x    104x
p4e  |   15x          |   15x
p4m  |   14x          |   14x

[TBeck]

You are the best!

No, mommy is the best!

CODE

     |  1.1.0b3       |  1.1.0b4
     |  Enc     Dec   |  Enc     Dec
=====================================
p0   |  130x    141x  |  132x    142x
...

Thank you, that was fast!

Thomas

This post has been edited by TBeck: Jan 3 2009, 14:33

[TBeck]

Prepearing the final release

I want to release the final version within the next two days. Well, i am already working on V1.1.1...

Please report any bugs or misbehaviour you may have encountered with the latest beta 4.

Thomas

[Synthetic Soul]

Thank you, that was fast!

No worries, the holiday season has given me a little more time to let tests run.

I know that you had said previously that -p0 was only required, but I didn't know whether that was still the case, and I was keen to perform yet another test to correlate my figures further. I was pleased to see very similar values. Where values are different I believe it is only with the faster speeds - where hundredths of a second may make a difference - and only a variance of one unit. Except -p0 and -p0m's encoding speed, which were out by two. Is this relevant to your changes do you think? Were you hoping to see a larger variation?

[TBeck]

Thank you, that was fast!

No worries, the holiday season has given me a little more time to let tests run.

Good for me!

I know that you had said previously that -p0 was only required, but I didn't know whether that was still the case, and I was keen to perform yet another test to correlate my figures further. I was pleased to see very similar values. Where values are different I believe it is only with the faster speeds - where hundredths of a second may make a difference - and only a variance of one unit. Except -p0 and -p0m's encoding speed, which were out by two. Is this relevant to your changes do you think? Were you hoping to see a larger variation?

No. That's quite exactly what i expected. Thank you again!

Thomas


TAK 1.1.0 final has been released

See this thread.