Skip to main content

Topic: TAK 1.1.2 - Beta release (Read 26664 times) previous topic - next topic

0 Members and 1 Guest are viewing this topic.
  • TBeck
  • [*][*][*][*][*]
  • Developer
TAK 1.1.2 - Beta release
Beta release of TAK 1.1.2 ((T)om's lossless (A)udio (K)ompressor)

It consists of:

- TAK Applications 1.1.2 Beta 2
- Winamp plugin 1.1.2 Beta 2
- Decoding library 1.1.2 Beta 2

The final release will additionally contain the SDK.

Download:

Download link removed. TAK 1.1.2 Final  has been released.

What's new

Beta 2, 2009-07-18

Improvements:

- Depending on preset and cpu up to 10 percent faster encoding and decoding compared with V1.1.1.

Beta 1, 2009-06-17

New Features:

- Tagging support for the command line encoder.

Improvements:

- Removed some more assembler routines and simplified a lot of code. Most binaries are again smaller now.
- Slightly faster encoding and decoding of LossyWav files.
- Slightly faster encoding and decoding if MMX is disabled and the pure pascal code is beeing used.
- Tiny general speedups for encoding and decoding on my Pentium Dual Core. My Sempron does not benefit from the optimizations.

Fixes:

- The new filter introduced in V1.1.1 revealed a bug in the encoder, which resulted in suboptimal performance especially when compressing LossyWav-files with the presets -p3 or -p4 (BTW: It doesn't make sense to go higher than -p2m when compressing LossyWav-files...).

Known issues:

- If you use pipe decoding and the application reading the pipe is beeing terminated before the whole file has been read, TAKC may get into an endless loop and has to be manually killed with the task manager. I don't think this is a big issue but i will try to fix it in one of the next versions. BTW: Big thanks to shnutils for testing the pipe decoding!
- There seem to be some compatibility issues with pipe decoding to some other applications ("crc1632.exe" has been reported). I will try to fix it in the next release.

Tagging

For your convenience here follows the description of the tag command taken from the applications readme:

Code: [Select]
-tt #      Add textual tag item #, where # is a key/value pair: "key=value",
           for instance "TITLE=A nice song". "key=@file" will read the value
           from the text(!) file "file" in the source directory.

Some remarks:

Tagging is only available when encoding.

TAK does not distinguish between upper and lower case in the keys; therfore "key" and "KEY" are equivalent.

You may assign multiple values to the same key:

Code: [Select]
-tt "Artist=One" -tt "ARTIST=Two"

will create a tag item with the key "Artist" and the value "One Two". The notation of the first appearance of the key is beeing used. Multiple values are beeing separated by zero bytes as defined in the APEv2 specification.

Item values are not allowed to be empty.

Beta testing

The beta version has already gone through extensive testing performed by my automatic scripts. But i haven't performed a noteworthy amount of testing of the new tagging functionality under real world conditions. Please try the beta release and report any bugs in this thread.

I would also be happy about tests of compression efficiency and speed. Because the final release will have identical performance (there may be a speed variation of 1 to 2 percent because of different code alignment of another build), it does make sense to test the beta.

Thanks for testing and have fun

Thomas
  • Last Edit: 27 July, 2009, 05:32:41 PM by TBeck

  • BenniP
  • [*][*]
TAK 1.1.2 - Beta release
Reply #1
Thank you very much for another great (beta) release!

I already love the tagging function
Do the Umlauts work?

  • TBeck
  • [*][*][*][*][*]
  • Developer
TAK 1.1.2 - Beta release
Reply #2
Do the Umlauts work?

Yes!

Anything which can be mapped to window's (locale) ansi char set should work.

  • Alexxander
  • [*][*][*][*]
TAK 1.1.2 - Beta release
Reply #3
I just tried new TAK 1.1.2 Beta with EAC v0.99pb5 and checked with foobar2000 v0.9.6.8 (with foo_input_tak.dll TAK Decoder 0.4.3 and tak_deco_lib.dll v.1.1.2).

I used in EAC external compression pointing to takc.exe and the following command line parameters:
Quote
-e -p4m -tt "artist=%a" -tt "title=%t" -tt "album=%g" -tt "date=%y" -tt "tracknumber=%n" -tt "genre=%m" -tt "comment=EAC Test & Copy Secure no C2 no Cache" %s
All seems to work just fine, I intentionally used latin characters in track title, album title and artist name. In foobar2000 the properties of .tak file showed this:
Quote
Artist Name : Tavarés
Track Title : á ä ñ ç è Heaven Must Be Missing An Angel
Album Title : ñThe Greatest Hitç
Date : 2000
Genre : Rock
Composer :
Performer :
Album Artist :
Track Number : 02
Total Tracks :
Disc Number :
Total Discs :
Comment : EAC Test & Copy Secure no C2 no Cache
The EAC created this .log file (I don't know how to make the box scrollable):
Code: [Select]
Exact Audio Copy V0.99 prebeta 5 from 4. May 2009

EAC extraction logfile from 18. June 2009, 9:55

Tavarés / ñThe Greatest Hitç

Used drive  : PLEXTOR DVD-ROM PX-130A  Adapter: 1  ID: 0

Read mode              : Secure
Utilize accurate stream : Yes
Defeat audio cache      : Yes
Make use of C2 pointers : No

Read offset correction                      : 738
Overread into Lead-In and Lead-Out          : No
Fill up missing offset samples with silence : Yes
Delete leading and trailing silent blocks  : No
Null samples used in CRC calculations      : No
Used interface                              : Native Win32 interface for Win NT & 2000
Gap handling                                : Appended to previous track

Used output format              : User Defined Encoder
Selected bitrate                : 896 kBit/s
Quality                        : High
Add ID3 tag                    : No
Command line compressor        : F:\Progs\TAK_1.1.2_Beta\Applications\Takc.exe
Additional command line options : -e -p4m -tt "artist=%a" -tt "title=%t" -tt "album=%g" -tt "date=%y" -tt "tracknumber=%n" -tt "genre=%m" -tt "comment=EAC Test & Copy Secure no C2 no Cache" %s


TOC of the extracted CD

     Track |  Start  |  Length  | Start sector | End sector
   ---------------------------------------------------------
       1  |  0:00.00 |  4:00.47 |        0    |    18046 
       2  |  4:00.47 |  6:55.25 |    18047    |    49196 
       3  | 10:55.72 |  4:59.43 |    49197    |    71664 
       4  | 15:55.40 |  4:34.70 |    71665    |    92284 
       5  | 20:30.35 |  3:20.32 |    92285    |  107316 
       6  | 23:50.67 |  3:39.70 |    107317    |  123811 
       7  | 27:30.62 |  3:27.25 |    123812    |  139361 
       8  | 30:58.12 |  6:13.15 |    139362    |  167351 
       9  | 37:11.27 |  3:36.23 |    167352    |  183574 
       10  | 40:47.50 |  3:55.27 |    183575    |  201226 
       11  | 44:43.02 |  4:14.48 |    201227    |  220324 
       12  | 48:57.50 |  5:48.45 |    220325    |  246469 
       13  | 54:46.20 |  3:28.10 |    246470    |  262079 
       14  | 58:14.30 |  3:47.40 |    262080    |  279144 
       15  | 62:01.70 |  3:44.60 |    279145    |  296004 
       16  | 65:46.55 |  3:15.40 |    296005    |  310669 


Track  2

     Filename F:\Temp\Tavarés  (2000)  ñThe Greatest Hitç\02 - á ä ñ ç è Heaven Must Be Missing An Angel.wav

     Pre-gap length  0:00:01.35

     Peak level 98.0 %
     Copy CRC EF51C6F0
     Accurately ripped (confidence 8)  [E174C624]
     Copy OK


All tracks accurately ripped

No errors occurred

End of status report
I noticed when that when the DOS Window pops up when takc.exe is launched after ripping, it just shows the temporarily filename (determined by EAC) and some dots to show the progress, i.e. no percentage or any other kind of number that could indicate a correct/succesfull compression was done.

[!--sizeo:1--][span style=\"font-size:8pt;line-height:100%\"][!--/sizeo--]Moderation: Replaced "quote" with "codebox".[/size]
  • Last Edit: 19 June, 2009, 12:59:48 PM by greynol

  • Alexxander
  • [*][*][*][*]
TAK 1.1.2 - Beta release
Reply #4
I tried embedding cuesheet in the big .tak file (=album image created with EAC Copy Image & Create cuesheet) with foobar2000 through Utils >>Edit cuesheet but as it seems it doesn't get saved to the tak file. Embedding cuesheets isn't possible (yet)?

Cheers and thumbs up for Thomas!

  • buktore
  • [*][*][*][*][*]
TAK 1.1.2 - Beta release
Reply #5
Slightly faster decoding on my Athlon 83.9x vs 85.4x

Thanks, Thomas 

  • foosion
  • [*][*][*][*][*]
  • FB2K Moderator (Donating)
TAK 1.1.2 - Beta release
Reply #6
I tried embedding cuesheet in the big .tak file (=album image created with EAC Copy Image & Create cuesheet) with foobar2000 through Utils >>Edit cuesheet but as it seems it doesn't get saved to the tak file. Embedding cuesheets isn't possible (yet)?
This is a feature of foobar2000 respectively the foo_input_tak plugin. The TAK decoder library is not involved in this at all. Check the foobar2000 console, the cue sheet probably contains syntax errors.
http://foosion.foobar2000.org/ - my components for foobar2000

  • Alexxander
  • [*][*][*][*]
TAK 1.1.2 - Beta release
Reply #7
I tried embedding cuesheet in the big .tak file (=album image created with EAC Copy Image & Create cuesheet) with foobar2000 through Utils >>Edit cuesheet but as it seems it doesn't get saved to the tak file. Embedding cuesheets isn't possible (yet)?
This is a feature of foobar2000 respectively the foo_input_tak plugin. The TAK decoder library is not involved in this at all. Check the foobar2000 console, the cue sheet probably contains syntax errors.
I have played around a bit more with album image in TAK format. I add cuesheet loading it via Utils>>Edit cuesheet but as I said yesterday it doesn't get saved. Today I first tried manual editing of tags and this works (through Propertied dialogue), the embedded cuesheet is updated as well when foobar2000 calculates ReplayGain values or when I use tags retrieved from freedb.

So after this on purpose I manually changed some track titles, artist name and album title (through Properties dialogue). I checked via Edit Cuesheet that these changes were applied. Then I loaded again the original cuesheet and this resulted in a weird behaviour: track titles were updated in embedded cuesheet but not the album title nor artist name! I opened with Notepad the original cuesheet to be sure that this file wasn't modified and indeed it wasn't (album title and artist name were still correct). Manually editing through Properties dialogue was necessary to get have again correct album name and artist name.

All the time I couldn't see anything strange in the console, just these kind of information:
Quote
Opening track for playback: "F:\Temp\Tavares - The Greatest Hits.tak" / index: 1
Properties dialog refresh: 0:00.000164
Properties dialog display refresh: 0:00.002740

The cuesheet I use is generated by EAC Copy Image & Create Cuesheet. It is supposed to have correct syntax, right? This is its content:
Code: [Select]
REM GENRE Rock
REM DATE 2000
REM DISCID FB102E10
REM COMMENT "ExactAudioCopy v0.99pb5"
PERFORMER "Tavares"
TITLE "The Greatest Hits"
FILE "Tavares - The Greatest Hits.tak" WAVE
  TRACK 01 AUDIO
   TITLE "It Only Takes A Minute"
   PERFORMER "Tavares"
   INDEX 01 00:00:00
  TRACK 02 AUDIO
   TITLE "Heaven Must Be Missing An Angel"
   PERFORMER "Tavares"
   INDEX 00 03:59:12
   INDEX 01 04:00:47
  TRACK 03 AUDIO
   TITLE "Slow Train To Paradise"
   PERFORMER "Tavares"
   INDEX 00 10:54:65
   INDEX 01 10:55:72
  TRACK 04 AUDIO
   TITLE "Never Had A Love Like This Before"
   PERFORMER "Tavares"
   INDEX 00 15:54:15
   INDEX 01 15:55:40
  TRACK 05 AUDIO
   TITLE "Check It Out"
   PERFORMER "Tavares"
   INDEX 00 20:29:10
   INDEX 01 20:30:35
  TRACK 06 AUDIO
   TITLE "She's Gone"
   PERFORMER "Tavares"
   INDEX 00 23:50:07
   INDEX 01 23:50:67
  TRACK 07 AUDIO
   TITLE "My Ship"
   PERFORMER "Tavares"
   INDEX 00 27:29:72
   INDEX 01 27:30:62
  TRACK 08 AUDIO
   TITLE "Don't Take Away The Music"
   PERFORMER "Tavares"
   INDEX 00 30:57:10
   INDEX 01 30:58:12
  TRACK 09 AUDIO
   TITLE "Whodunit"
   PERFORMER "Tavares"
   INDEX 00 37:09:57
   INDEX 01 37:11:27
  TRACK 10 AUDIO
   TITLE "The Mighty Power Of Love"
   PERFORMER "Tavares"
   INDEX 00 40:46:30
   INDEX 01 40:47:50
  TRACK 11 AUDIO
   TITLE "One Step Away"
   PERFORMER "Tavares"
   INDEX 00 44:41:65
   INDEX 01 44:43:02
  TRACK 12 AUDIO
   TITLE "The Ghost Of Love"
   PERFORMER "Tavares"
   INDEX 00 48:56:45
   INDEX 01 48:57:50
  TRACK 13 AUDIO
   TITLE "The Love I Never Had"
   PERFORMER "Tavares"
   INDEX 00 54:45:47
   INDEX 01 54:46:20
  TRACK 14 AUDIO
   TITLE "I Wanna See You Soon"
   PERFORMER "Tavares"
   INDEX 00 58:13:37
   INDEX 01 58:14:30
  TRACK 15 AUDIO
   TITLE "Bein' With You"
   PERFORMER "Tavares"
   INDEX 00 62:00:55
   INDEX 01 62:01:70
  TRACK 16 AUDIO
   TITLE "More Than A Woman"
   PERFORMER "Tavares"
   INDEX 00 65:45:62
   INDEX 01 65:46:55
i haven't encountered this problem with FLAC and this all is done on Windows 7 RC.

[!--sizeo:1--][span style=\"font-size:8pt;line-height:100%\"][!--/sizeo--]Moderation: Replaced "quote" with "codebox".  Please learn how to do this for future posts.[/size]
  • Last Edit: 19 June, 2009, 12:57:29 PM by greynol

  • sld
  • [*][*][*][*][*]
TAK 1.1.2 - Beta release
Reply #8
(I don't know how to make the box scrollable):

Code: [Select]
[codebox][/codebox]
  • Last Edit: 19 June, 2009, 12:11:48 PM by sld

  • Alexxander
  • [*][*][*][*]
TAK 1.1.2 - Beta release
Reply #9
Some more information. Please split off this part of thread because as one can read in the following this has nothing to do with TAK. Sorry for messing up this thread.

I ripped an other CD to TAK and the same problem ocurred so I ripped the same album to FLAC image to check. The same behaviour with FLAC showed up! At first I was surprised but then I realised the difference: I ripped album image to TAK and let EAC tag the file, but in past when I ripped album to one FLAC image I didn't let EAC tag the file at all (no tagging instructions in EAC external compressor commandline arguments). In the latter case embedding cuesheet never was a problem because album image had no tags.

So after this I used foobar2000's Remove Tags option and indeed, after doing so embedding a cuesheet was done correctly. I've spent some more time deleting individual tag entries (through Properties dialogue) and embedding the cuesheet afterwards. For now, I've come to the conclusion that foobar2000 displays native FLAC tags or TAK apev2 tags when present. If tag isn't present it is retrieved from cuesheet. Can someone confirm this? In foobar2000 is there indeed some kind of priority in reading and displaying tags?

  • alvaro84
  • [*][*][*]
TAK 1.1.2 - Beta release
Reply #10
Hi! I've just tested the new beta for decoding performance under foobar and I encountered a strange phenomenon.
For the first I tested its speed against the last (1.1.1) decoding library while playing another TAK in the meantime and found a huge leap in performance: 378x vs 298x... for a brief moment I was very happy about it, but that 298 looked like a great regression towards the first releases, I had remembered way-over-300 values... and a test run without background music showed me that the old library is capable of 369x speed on the same test material. I repeated the test a few  times and concluded that 1.1.1 is indeed sensitive to another instance. Looks like a glitch in threading, I don't know if it's in foobar (0.9.6.7, sorry, I was too lazy to upgrade to the very last version) or on tak_deco_lib's side...
Anyway, it seems to be gone in 1.1.2 beta.
The results are:

1.1.1 w/ background music - 298x
1.1.2 w/ background music - 378x
1.1.1 w/o background music - 369x
1.1.2 w/o background music - 379x

(running on a Conroe C2D 6420@3.33GHz(8*416), 832MHz cl4 ddr2, offboard nVidia gfx, PCI x-fi fatal1ty)
And on the top of that it still has some performance improvement and is smaller. Keep up the good work!

  • TBeck
  • [*][*][*][*][*]
  • Developer
TAK 1.1.2 - Beta release
Reply #11
I just tried new TAK 1.1.2 Beta with EAC v0.99pb5 and checked with foobar2000 v0.9.6.8 (with foo_input_tak.dll TAK Decoder 0.4.3 and tak_deco_lib.dll v.1.1.2).
...
All seems to work just fine, I intentionally used latin characters in track title, album title and artist name.

Thank you!

Yes, i also like positive (non-bug) reports about the new feature. Has anyone tried to embed cuesheet files with the -tt command?

I noticed when that when the DOS Window pops up when takc.exe is launched after ripping, it just shows the temporarily filename (determined by EAC) and some dots to show the progress, i.e. no percentage or any other kind of number that could indicate a correct/succesfull compression was done.

TAK will show the compression ratio when it is done. But probably you can't see this, because the window is beeing closed too fast.

Other applications can check TAK's exit code. It's 0 if anything went well.

If you want to know, if your specific application checks the exit code, you may add an invalid paramter to your command line to force TAK to return an error code.

Slightly faster decoding on my Athlon 83.9x vs 85.4x

Unexpected but nice!

I repeated the test a few  times and concluded that 1.1.1 is indeed sensitive to another instance. Looks like a glitch in threading, I don't know if it's in foobar (0.9.6.7, sorry, I was too lazy to upgrade to the very last version) or on tak_deco_lib's side...
Anyway, it seems to be gone in 1.1.2 beta.

Fine!

To be honest, i don't know what exactly has caused the performance drop in 1.1.1...

Im sure it has to do with the cpu's L1 (or maybe L2) cache utilization. Although i am caring a lot about optimal caching, there can be a lot of specific interaction with other applications, that is not easy to control.

(running on a Conroe C2D 6420@3.33GHz(8*416), 832MHz cl4 ddr2, offboard nVidia gfx, PCI x-fi fatal1ty)
And on the top of that it still has some performance improvement and is smaller. Keep up the good work!

Thank you!

  Thomas

  • Alexxander
  • [*][*][*][*]
TAK 1.1.2 - Beta release
Reply #12
Has anyone tried to embed cuesheet files with the -tt command?

I used
Code: [Select]
F:\Temp>takc -e -tt "cuesheet=@PCM.cue" "Patrick Cowley - Menergy.wav"

and takc -fi showed
Code: [Select]
F:\Temp>takc -fi "Patrick Cowley - Menergy.tak"
=== Patrick Cowley - Menergy.tak ==============================

  File size:                   209.62 MB
  Header size:                   0.11 KB
    Unused:                      0.00 KB
  Compression:                  64.57 %
  Samples per channel:       85101828
  File duration:              1929.75 sec
  Frame duration:                 125 ms
  Seek table:              Not available
  Audio format:            PCM, 44100 Hz, 16 Bits, 2 Channels
  Encoder:                 V 1.1.2, -p2
  Wave file meta data:     Header 44, Footer 0 Bytes
  MD5:                     Not available
  APEv2-Tag:               Yes / 1 Items / 0.68 KB
  Status:                  Ok

But how can I see the tag(s) content? Foobar2000 Properties dialogue is empty and says no embedded cuesheet is present, but this latter probably has something to do with the tak foobar component.

On the other hand: How can I tag an already compressed file?

  • TBeck
  • [*][*][*][*][*]
  • Developer
TAK 1.1.2 - Beta release
Reply #13
Has anyone tried to embed cuesheet files with the -tt command?

I used
Code: [Select]
F:\Temp>takc -e -tt "cuesheet=@PCM.cue" "Patrick Cowley - Menergy.wav"

and takc -fi showed
[code]F:\Temp>takc -fi "Patrick Cowley - Menergy.tak"
  APEv2-Tag:           Yes / 1 Items / 0.68 KB

Thank you for testing!

On the other hand: How can I tag an already compressed file?

I don't think this would be the task of the compressor (would be some kind of "reinventing the wheel"). Better use a dedicated application like Mp3Tag or foobar (i hope, that cuesheet support can be implemented sooner or later).

I still would like to receive more feedback regarding the new tagging functionality. If more people can confirm, that it is working well, i may prepare the final release.

  Thomas

  • Alexxander
  • [*][*][*][*]
TAK 1.1.2 - Beta release
Reply #14
Not much people replying  It seems there's not much interest anymore in TAK?

  • carpman
  • [*][*][*][*][*]
  • Developer
TAK 1.1.2 - Beta release
Reply #15
Maybe it's because Thomas did such a good job of previous versions -  there are too many content TAK users.

C.
  • Last Edit: 30 June, 2009, 01:39:14 PM by carpman
PC = TAK + LossyWAV  ::  Portable = Lame MP3

  • Dologan
  • [*][*][*][*]
  • Members (Donating)
TAK 1.1.2 - Beta release
Reply #16
Great stuff. Keep up the wonderful work with TAK.

  • Firehawk
  • [*]
TAK 1.1.2 - Beta release
Reply #17
finally tags by default?
honestly.. that was the reason which kept me from using TAK (didn't want to use this wapet thingy)
will start testing tomorrow

  • Gow
  • [*][*][*]
TAK 1.1.2 - Beta release
Reply #18
One step closer to 2.0. 
Zune 80, Tak -p4 audio library, Lossless=Choice

  • Skymmer
  • [*][*][*]
TAK 1.1.2 - Beta release
Reply #19
Nice to see the new TAK version but let's look closer to it. First of all I have been quite interested with following phrase from the changelog:

Quote
BTW: It doesn't make sense to go higher than -p2m when compressing LossyWav-files...


So let's see how it looks on practise. In this and all the following tests I've used the Chick Corea - The Ultimate Adventure album being converted to one WAV file or its variant passed through LossyWAV beta 1.1.3e at default settings.
At one test another album have been used.
Timer v8.00 used for measurements and output file is redirected to another HDD to avoid impact from IO-system.

Test between 2m, 3m and 4m levels for lossyWAV variant
TAK v1.1.2 beta 1 at -p# -fsl512 -wm0
Code: [Select]
Chick Corea - The Ultimate Adventure (lossyWAV)
------------------------------------
2m = 129.436 = 253 239 535
3m = 185.518 = 253 359 228
4m = 208.372 = 253 359 228

Chick Corea - Return To Forever (lossyWAV)
-------------------------------
2m =  83.269 = 150 480 212
3m = 117.042 = 150 382 348
4m = 132.707 = 150 382 348


In both cases 4m preset produce the same sized output file as 3m but slower so more correctly will be to say:
It doesn't make sense to go higher than -p3m when compressing LossyWav-files... Anyway, thanks for tip 
More interesting will be to see if previous issue with bad behaviour on some -fsl values is gone.
First of all the brief test on lossyWAV variant.

Code: [Select]
Chick Corea '2006 - The Ultimate Adventure (Lossy)
--------------------------------------------------
takc104 -e -p5m -fsl512 -wm0 -sts3     138.360         253 372 934
takc110 -e -p4m -fsl512 -wm0 -sts0     190.513         253 408 503
takc111 -e -p4m -fsl512 -wm0           208.103         258 427 682
takc112 -e -p4m -fsl512 -wm0           208.372         253 359 228 (still slower but best ratio)


Unexpected results. v1.1.2 beta 1 still provides bad time rankings but somehow managed to provide best compression.
Here is another test table for lossless variant.

Code: [Select]
Chick Corea '2006 - The Ultimate Adventure (Lossless)
-----------------------------------------------------
takc104 -e -p5m -fsl512 -wm0 -sts3     139.929      519 996 238
takc110 -e -p4m -fsl512 -wm0 -sts0     193.307      519 875 895
takc111 -e -p4m -fsl512 -wm0           201.356      522 506 351
takc112 -e -p4m -fsl512 -wm0           206.617      520 875 280 (still slower and worse ratio than 104)

takc104 -e -p5m -fsl1024 -wm0 -sts3    207.017      512 660 380
takc110 -e -p4m -fsl1024 -wm0 -sts0    197.920      512 648 369
takc111 -e -p4m -fsl1024 -wm0          209.851      512 915 726
takc112 -e -p4m -fsl1024 -wm0          210.802      512 446 995 (still slower but best ratio)

takc104 -e -p5m -fsl2048 -wm0 -sts3    257.169      508 215 882
takc110 -e -p4m -fsl2048 -wm0 -sts0    203.414      508 167 598
takc111 -e -p4m -fsl2048 -wm0          219.218      508 407 219
takc112 -e -p4m -fsl2048 -wm0          218.873      508 246 729 (faster)

takc104 -e -p5m -fsl4096 -wm0 -sts3    263.727      505 789 477
takc110 -e -p4m -fsl4096 -wm0 -sts0    191.189      505 777 818
takc111 -e -p4m -fsl4096 -wm0          206.334      505 679 778
takc112 -e -p4m -fsl4096 -wm0          205.392      505 651 357 (faster and best ratio)

takc104 -e -p5m -fsl8192 -wm0 -sts3    285.178      504 967 305
takc110 -e -p4m -fsl8192 -wm0 -sts0    208.951      504 801 487
takc111 -e -p4m -fsl8192 -wm0          222.858      504 767 462
takc112 -e -p4m -fsl8192 -wm0          221.311      504 762 999 (faster and best ratio)

takc104 -e -p5m -fsl16384 -wm0 -sts3   290.972      504 605 505
takc110 -e -p4m -fsl16384 -wm0 -sts0   213.665      504 827 649
takc111 -e -p4m -fsl16384 -wm0         228.194      504 612 442
takc112 -e -p4m -fsl16384 -wm0         229.043      504 612 442 (faster)


Still strange behaviour on -fsl values of 512 and 1024 regarding the time. Also strange behaviour on -fsl values of 4096 and 8192 regarding the size.
Hope it will help 
Gabber, Jazz and IDM

TAK 1.1.2 - Beta release
Reply #20
I just concern about the decode performance's difference with FLAC in foobar2000. 

  • TBeck
  • [*][*][*][*][*]
  • Developer
TAK 1.1.2 - Beta release
Reply #21
finally tags by default?
honestly.. that was the reason which kept me from using TAK (didn't want to use this wapet thingy)
will start testing tomorrow

Did it work for you?

One step closer to 2.0. 

That's true, but although i am regulary working on the optimized codec, i can't see a release date earlier than maybe the end of this year...

Nice to see the new TAK version but let's look closer to it. First of all I have been quite interested with following phrase from the changelog:

Quote
BTW: It doesn't make sense to go higher than -p2m when compressing LossyWav-files...


So let's see how it looks on practise. In this and all the following tests I've used the Chick Corea - The Ultimate Adventure album being converted to one WAV file or its variant passed through LossyWAV beta 1.1.3e at default settings.
At one test another album have been used.
...
In both cases 4m preset produce the same sized output file as 3m but slower so more correctly will be to say:
It doesn't make sense to go higher than -p3m when compressing LossyWav-files... Anyway, thanks for tip 

Thank you for testing!

My recommendation is based upon the average result of several quite heterogenous file sets. But surely there will be cases where higher presets are advantageous.

More interesting will be to see if previous issue with bad behaviour on some -fsl values is gone.
First of all the brief test on lossyWAV variant.
...
Unexpected results. v1.1.2 beta 1 still provides bad time rankings but somehow managed to provide best compression.

Looks good for me.

Here is another test table for lossless variant.
...
Still strange behaviour on -fsl values of 512 and 1024 regarding the time. Also strange behaviour on -fsl values of 4096 and 8192 regarding the size.

For me this is a non-issue because the -fsl does only make real sense in combination with LossyWav.

Hope it will help 

Oh yes! 

I just concern about the decode performance's difference with FLAC in foobar2000. 

Maybe this is for you:

While i was waiting for more testers of the new tagging funcionality (this seems to become the longest beta test of any TAK version...), i already worked on V 1.1.3. I have achieved speed improvements of up to 7 percent for both encoding and decoding! 

And i have deceided to make those optimizations part of the 1.1.2 release. Therefore i intend to release a beta 2 within the next days.

  Thomas

TAK 1.1.2 - Beta release
Reply #22
Are you thinking about releasing the encoder library too?
As I wrote in the official 1.1.1 release thread, I wrote two filters, for Audition and nero, obviously decoding only.
I hope you will consider about releasing the encoding library as soon as possible; is 1.1.2 beta2 too soon?

  • TBeck
  • [*][*][*][*][*]
  • Developer
TAK 1.1.2 - Beta release
Reply #23
While i was waiting for more testers of the new tagging funcionality (this seems to become the longest beta test of any TAK version...), i already worked on V 1.1.3. I have achieved speed improvements of up to 7 percent for both encoding and decoding! 

Well, here are some results for my Intel Pentium Dual Core 2 GHz and AMD Sempron 2,2 GHz. Speed improvements over V1.1.1 in percent:

Code: [Select]
         Intel Core        AMD Sempron
         Enco     Deco     Enco     Deco
-----------------------------------------
-p0     10,38     8,50     4,05    3,33
-p1      7,86     7,77     2,47    2,87
-p2      6,76    10,83     1,98    6,03
-p3     -2,59     7,06     1,07    4,12
-p4      1,57     8,56     0,88    3,39
-p4m     1,10              0,96
-----------------------------------------

Now i will prepare the beta 2 release.

edit: Updated according to the results of the Beta 2.
  • Last Edit: 18 July, 2009, 06:29:14 AM by TBeck

  • TBeck
  • [*][*][*][*][*]
  • Developer
TAK 1.1.2 - Beta release
Reply #24
Are you thinking about releasing the encoder library too?
As I wrote in the official 1.1.1 release thread, I wrote two filters, for Audition and nero, obviously decoding only.
I hope you will consider about releasing the encoding library as soon as possible; is 1.1.2 beta2 too soon?

Oh yes, that's definitely too soon!

It involves design, coding, internal and external testing and a considerable update of the SDK documentation.

But i really appreciate your work! I will see, what i can do. No promises, because i don't know how much spare time will be left for TAK in the next months.

  Thomas