Hydrogenaudio Forums > FLAC 1.2.0 released

jcoalson

Jul 24 2007, 18:15

FLAC 1.2.0 is out. There are a few new features and some speedups and fixes, but more importantly, there are some small changes to the decoder to pave the way for possible future compression improvements, so applications developers are encouraged to upgrade (the API has not changed). The changelog has all the details, but in summary:

- automatic SSE OS detection at runtime (so no need to specifically enable SSE at compile time)
- small encoder and decoder speedups
- new --ignore-chunk-sizes option in flac will help with fb2k piped encoding to flac.exe

jamesbaud

Jul 24 2007, 21:26

Awesome! Thanks!

Heliologue

Jul 24 2007, 22:23

You are a gentleman and a scholar.

*tests*

johnsonlam

Jul 24 2007, 22:52

QUOTE(jcoalson @ Jul 25 2007, 08:15)

FLAC 1.2.0 is out.

Thank you Josh.

Then Foobar2000 and Cooledit plugin need to update.

Bourne

Jul 25 2007, 00:06

QUOTE(jcoalson @ Jul 24 2007, 21:15)

- new --ignore-chunk-sizes option in flac will help with fb2k piped encoding to flac.exe

do you mean that the problem about reaching the maximum seek points in the file is gone?

kanak

Jul 25 2007, 00:23

QUOTE(johnsonlam @ Jul 25 2007, 00:52)

Thank you Josh.

Then Foobar2000 and Cooledit plugin need to update.

um why?

Egor

Jul 25 2007, 00:41

QUOTE(Bourne @ Jul 25 2007, 13:06)

QUOTE(jcoalson @ Jul 24 2007, 21:15)

- new --ignore-chunk-sizes option in flac will help with fb2k piped encoding to flac.exe

do you mean that the problem about reaching the maximum seek points in the file is gone?

Yes, though you will need to manually update the parameters string for converter (make a custom preset) and tell fb2k developers it would be great to change the FLAC preset on occasion.

le_canz

Jul 25 2007, 02:42

Thank you Josh !

Seiitsu

Jul 25 2007, 08:00

QUOTE(kanak @ Jul 25 2007, 08:23)

QUOTE(johnsonlam @ Jul 25 2007, 00:52)

Thank you Josh.

Then Foobar2000 and Cooledit plugin need to update.

um why?

Exactly... why? I assume backwards compatibility should not be a problem at this time at least since the decoder was updated for future possible compression improvements.

haregoo

Jul 25 2007, 08:06

v1.2.0 "--ignore-chunk-sizes" seems to not create any seek point. Is it safe for decoder compatibility?

metaflac --list sample1.flac (encoded with %s -o %d)

CODE

METADATA block #0
  type: 0 (STREAMINFO)
  is last: false
  length: 34
  minimum blocksize: 4096 samples
  maximum blocksize: 4096 samples
  minimum framesize: 14 bytes
  maximum framesize: 14295 bytes
  sample_rate: 44100 Hz
  channels: 2
  bits-per-sample: 16
  total samples: 21142128
  MD5 signature: dc7707fe44e414873c67e3597ef44079
METADATA block #1
  type: 3 (SEEKTABLE)
  is last: false
  length: 864
  seek points: 48
    point 0: sample_number=0, stream_offset=0, frame_samples=4096
    point 1: sample_number=438272, stream_offset=863109, frame_samples=4096
    point 2: sample_number=880640, stream_offset=1703274, frame_samples=4096
    point 3: sample_number=1318912, stream_offset=2421506, frame_samples=4096
    point 4: sample_number=1761280, stream_offset=3205408, frame_samples=4096
    point 5: sample_number=2203648, stream_offset=4013462, frame_samples=4096
    point 6: sample_number=2641920, stream_offset=4752195, frame_samples=4096
    point 7: sample_number=3084288, stream_offset=5562112, frame_samples=4096
    point 8: sample_number=3526656, stream_offset=6396950, frame_samples=4096
    point 9: sample_number=3964928, stream_offset=7311730, frame_samples=4096
    point 10: sample_number=4407296, stream_offset=8213471, frame_samples=4096
    point 11: sample_number=4849664, stream_offset=9124752, frame_samples=4096
    point 12: sample_number=5287936, stream_offset=10007487, frame_samples=4096
    point 13: sample_number=5730304, stream_offset=10973161, frame_samples=4096
    point 14: sample_number=6172672, stream_offset=11975115, frame_samples=4096
    point 15: sample_number=6610944, stream_offset=13090217, frame_samples=4096
    point 16: sample_number=7053312, stream_offset=14139121, frame_samples=4096
    point 17: sample_number=7495680, stream_offset=15086376, frame_samples=4096
    point 18: sample_number=7933952, stream_offset=15992142, frame_samples=4096
    point 19: sample_number=8376320, stream_offset=17001045, frame_samples=4096
    point 20: sample_number=8818688, stream_offset=17969089, frame_samples=4096
    point 21: sample_number=9256960, stream_offset=18727284, frame_samples=4096
    point 22: sample_number=9699328, stream_offset=19690182, frame_samples=4096
    point 23: sample_number=10141696, stream_offset=20527314, frame_samples=4096
    point 24: sample_number=10579968, stream_offset=21274576, frame_samples=4096
    point 25: sample_number=11022336, stream_offset=22153031, frame_samples=4096
    point 26: sample_number=11464704, stream_offset=23198098, frame_samples=4096
    point 27: sample_number=11902976, stream_offset=24314989, frame_samples=4096
    point 28: sample_number=12345344, stream_offset=25441145, frame_samples=4096
    point 29: sample_number=12787712, stream_offset=26470938, frame_samples=4096
    point 30: sample_number=13225984, stream_offset=27428645, frame_samples=4096
    point 31: sample_number=13668352, stream_offset=28361327, frame_samples=4096
    point 32: sample_number=14110720, stream_offset=29440696, frame_samples=4096
    point 33: sample_number=14548992, stream_offset=30516648, frame_samples=4096
    point 34: sample_number=14991360, stream_offset=31584719, frame_samples=4096
    point 35: sample_number=15433728, stream_offset=32745588, frame_samples=4096
    point 36: sample_number=15872000, stream_offset=33527464, frame_samples=4096
    point 37: sample_number=16314368, stream_offset=34214583, frame_samples=4096
    point 38: sample_number=16756736, stream_offset=34933179, frame_samples=4096
    point 39: sample_number=17195008, stream_offset=35779963, frame_samples=4096
    point 40: sample_number=17637376, stream_offset=36868855, frame_samples=4096
    point 41: sample_number=18079744, stream_offset=38032471, frame_samples=4096
    point 42: sample_number=18518016, stream_offset=38939203, frame_samples=4096
    point 43: sample_number=18960384, stream_offset=39979350, frame_samples=4096
    point 44: sample_number=19402752, stream_offset=41156540, frame_samples=4096
    point 45: sample_number=19841024, stream_offset=42423269, frame_samples=4096
    point 46: sample_number=20283392, stream_offset=43668945, frame_samples=4096
    point 47: sample_number=20725760, stream_offset=44939333, frame_samples=4096
METADATA block #2
  type: 4 (VORBIS_COMMENT)
  is last: false
  length: 40
  vendor string: reference libFLAC 1.2.0 20070715
  comments: 0
METADATA block #3
  type: 1 (PADDING)
  is last: true
  length: 8192

metaflac --list sample2.flac (encoded with - -o %d)

CODE

METADATA block #0
  type: 0 (STREAMINFO)
  is last: false
  length: 34
  minimum blocksize: 4096 samples
  maximum blocksize: 4096 samples
  minimum framesize: 14 bytes
  maximum framesize: 14295 bytes
  sample_rate: 44100 Hz
  channels: 2
  bits-per-sample: 16
  total samples: 21142128
  MD5 signature: dc7707fe44e414873c67e3597ef44079
METADATA block #1
  type: 3 (SEEKTABLE)
  is last: false
  length: 21924
  seek points: 1218
    point 0: sample_number=0, stream_offset=0, frame_samples=4096
    point 1: sample_number=438272, stream_offset=863109, frame_samples=4096
    point 2: sample_number=880640, stream_offset=1703274, frame_samples=4096
    point 3: sample_number=1318912, stream_offset=2421506, frame_samples=4096
    point 4: sample_number=1761280, stream_offset=3205408, frame_samples=4096
    point 5: sample_number=2203648, stream_offset=4013462, frame_samples=4096
    point 6: sample_number=2641920, stream_offset=4752195, frame_samples=4096
    ~
    ~
    ~
    point 1212: sample_number=534492000, stream_offset=0, frame_samples=0
    point 1213: sample_number=534933000, stream_offset=0, frame_samples=0
    point 1214: sample_number=535374000, stream_offset=0, frame_samples=0
    point 1215: sample_number=535815000, stream_offset=0, frame_samples=0
    point 1216: sample_number=536256000, stream_offset=0, frame_samples=0
    point 1217: sample_number=536697000, stream_offset=0, frame_samples=0
METADATA block #2
  type: 4 (VORBIS_COMMENT)
  is last: false
  length: 40
  vendor string: reference libFLAC 1.2.0 20070715
  comments: 0
METADATA block #3
  type: 1 (PADDING)
  is last: true
  length: 65536

metaflac --list sample3.flac (encoded with --ignore-chunk-sizes - -o %d)

CODE

METADATA block #0
  type: 0 (STREAMINFO)
  is last: false
  length: 34
  minimum blocksize: 4096 samples
  maximum blocksize: 4096 samples
  minimum framesize: 14 bytes
  maximum framesize: 14295 bytes
  sample_rate: 44100 Hz
  channels: 2
  bits-per-sample: 16
  total samples: 21142128
  MD5 signature: dc7707fe44e414873c67e3597ef44079
METADATA block #1
  type: 4 (VORBIS_COMMENT)
  is last: false
  length: 40
  vendor string: reference libFLAC 1.2.0 20070715
  comments: 0
METADATA block #2
  type: 1 (PADDING)
  is last: true
  length: 8192

And I found the last FLAC takes more CPU cycle to seek with long flac files (55 minutes - 400MB).

skamp

Jul 25 2007, 08:23

QUOTE(haregoo @ Jul 25 2007, 16:06)

v1.2.0 seems to not create any seek point. Is it safe for decoder compatibility?

metaflac --list sample1.flac (encoded with %s -o %d)

CODE

METADATA block #1
  type: 3 (SEEKTABLE)
  is last: false
  length: 864
  seek points: 48
    point 0: sample_number=0, stream_offset=0, frame_samples=4096
    point 1: sample_number=438272, stream_offset=863109, frame_samples=4096
    point 2: sample_number=880640, stream_offset=1703274, frame_samples=4096
    point 3: sample_number=1318912, stream_offset=2421506, frame_samples=4096

I don't understand - what are those, if not seekpoints?

haregoo

Jul 25 2007, 08:31

QUOTE(skamp @ Jul 25 2007, 23:23)

I don't understand - what are those, if not seekpoints?

FLAC - documentation
http://flac.sourceforge.net/documentation_...t_overview.html

QUOTE

Other blocks allow for padding, seek tables, tags, cuesheets, and application-specific data. There are flac options for adding PADDING blocks or specifying seek points. FLAC does not require seek points for seeking but they can speed up seeks, or be used for cueing in editing applications.

My question also is solved. >Is it safe for decoder compatibility?

skamp

Jul 25 2007, 08:36

QUOTE(haregoo @ Jul 25 2007, 16:31)

FLAC - documentation
http://flac.sourceforge.net/documentation_...t_overview.html

I still don't understand what you meant. I have read the format documentation many times, and as far as I know, those are seekpoints... Am I so tired that I am missing something?

haregoo

Jul 25 2007, 08:42

QUOTE(skamp @ Jul 25 2007, 23:36)

QUOTE(haregoo @ Jul 25 2007, 16:31)

FLAC - documentation
http://flac.sourceforge.net/documentation_...t_overview.html

I still don't understand what you meant. I have read the format documentation many times, and as far as I know, those are seekpoints... Am I so tired that I am missing something?

Sorry, I made mistake. What I meant was:
v1.2.0 "--ignore-chunk-sizes" seems to not create any seek point.

Dr. Oviri

Jul 25 2007, 09:41

Great Josh

LANjackal

Jul 25 2007, 09:56

Thanks! I upgraded

. Also, good work on getting an executable installer out at the same time too

.

Fandango

Jul 25 2007, 10:35

Wow, two new versions of lossless codecs released within half a day.

That calls for an updated comparison chart...

Synthetic Soul

Jul 25 2007, 12:16

QUOTE(Fandango @ Jul 25 2007, 17:35)

Wow, two new versions of lossless codecs released within half a day.

That calls for an updated comparison chart...

Tell me about it.

I'm hoping to run my tests this weekend, when I know the PC will be free.

jcoalson

Jul 25 2007, 12:36

seekpoints are not required but can improve seek times. with --ignore-chunk-sizes, a seektable is currently not added. in the future I plan to have the encoder go back and add one after encoding is finished and it knows the total number of samples. a workaround in the meantime is to add a seektable afterwards with metaflac.

QUOTE(Seiitsu @ Jul 25 2007, 09:00)

Exactly... why? I assume backwards compatibility should not be a problem at this time at least since the decoder was updated for future possible compression improvements.

a 1.2.0 decoder will be able to decode any 1.2.x streams, so upgrading now means not having to when 1.2.x encoder improvements come out. the FLAC 1.1.x series spanned 4.5 years.

QUOTE(LANjackal @ Jul 25 2007, 10:56)

Thanks! I upgraded

. Also, good work on getting an executable installer out at the same time too

.

yes, this should be the norm now that I have my own nsi script.

QUOTE(Fandango @ Jul 25 2007, 11:35)

Wow, two new versions of lossless codecs released within half a day.

That calls for an updated comparison chart...

speaking of which, FLAC and WavPack encode at the same speed by default but have different ratings on the wiki.

molnart

Jul 25 2007, 12:38

Thanks for the new version! I'm just wondering: why that version numbering jump (1.1.x -> 1.2.x) ?

xmixahlx

Jul 25 2007, 13:14

QUOTE(molnart @ Jul 25 2007, 11:38)

Thanks for the new version! I'm just wondering: why that version numbering jump (1.1.x -> 1.2.x) ?

that's teh unix system (MAJOR).(MINOR).(REVISION)

it usually points to compatibility issue(s)

later

ernstblaauw

Jul 25 2007, 13:15

Great, however, every time I try to download the Windows version, I get a file with size 0. Has anyone successfully downloaded the 1.2.0 Windows version?

mr budzone

Jul 25 2007, 13:19

<3

john33

Jul 25 2007, 13:53

QUOTE(ernstblaauw @ Jul 25 2007, 20:15)

Great, however, every time I try to download the Windows version, I get a file with size 0. Has anyone successfully downloaded the 1.2.0 Windows version?

There's a compile on Rarewares as well, if that's easier.

Eli

Jul 25 2007, 14:00

Josh,
Thank you for the new release!!!

Any chance of the RAW CD TOC support in upcoming revisions?

jcoalson

Jul 25 2007, 14:19

QUOTE(ernstblaauw @ Jul 25 2007, 14:15)

Great, however, every time I try to download the Windows version, I get a file with size 0. Has anyone successfully downloaded the 1.2.0 Windows version?

looks like a problem with one of the mirrors, I fixed the download link on download.html

QUOTE(Eli @ Jul 25 2007, 15:00)

Any chance of the RAW CD TOC support in upcoming revisions?

it's on the TODO, can't assign a chance just yet but it's non-zero!

funkyblue

Jul 25 2007, 16:26

I cannot download from any sourceforge mirror. I am getting a file not found 404 message for the Windows installer.
Can you please fix.

Thanks

jcoalson

Jul 25 2007, 16:38

I don't know what the problem is with sourceforge, I think they might be checking the referrer now or something, try copy-pasting this link into the address bar, it just worked for me:

http://downloads.sourceforge.net/flac/flac-1.2.0a.exe

jcoalson

Jul 25 2007, 16:48

if not, download it via the sourceforge file section: http://sourceforge.net/project/showfiles.p...lease_id=527434

funkyblue

Jul 25 2007, 16:52

I have tried that, but all mirrors are coming up with a 404 error.
Thanks

crazee_canuck

Jul 25 2007, 17:26

QUOTE(funkyblue @ Jul 25 2007, 17:52)

I have tried that, but all mirrors are coming up with a 404 error.
Thanks

I was able to download via the mirror "superb-east.dl.sourceforge.net ( McLean, Virginia - North America )"

Try that one.

funkyblue

Jul 25 2007, 17:29

Nope, still dead.
Are you able to email me the exe file?

scott at aceblue.com

Thanks

greynol

Jul 25 2007, 17:33

I've given up on posting the links. Just try different mirrors, cancel on the popup and instead right-click on the direct link and choose save link as...

THANKS JOSH!!!

bilbo

Jul 25 2007, 18:00

Can't seem to be able to download anything from Sourceforge at this point!

funkyblue

Jul 25 2007, 18:04

zip worked, exe still dead...

I have tried with my a diffrent net connection, and the exe will not work there as well.

Thanks greynol.

Seiitsu

Jul 25 2007, 18:15

QUOTE(jcoalson @ Jul 25 2007, 20:36)

QUOTE(Seiitsu @ Jul 25 2007, 09:00)

Exactly... why? I assume backwards compatibility should not be a problem at this time at least since the decoder was updated for future possible compression improvements.

a 1.2.0 decoder will be able to decode any 1.2.x streams, so upgrading now means not having to when 1.2.x encoder improvements come out. the FLAC 1.1.x series spanned 4.5 years.

While I do see the point in upgrading the decoders since it will become necessary to do so in the future once the upgrades kick in I assume it will be some time before any improvements will kick in to allow for developers to update their software and hardware manufacturers to update firmwares to support 1.2.0

Hence I mostly meant it as that it doesn't necessarily have to be done right now, since it sounded as if it was necessary for 1.2.0 to be useful, but I do see the point for it eventually.

fj4

Jul 25 2007, 19:01

Thanks for your hard work, Josh!

jcoalson

Jul 25 2007, 22:15

QUOTE(Seiitsu @ Jul 25 2007, 19:15)

While I do see the point in upgrading the decoders since it will become necessary to do so in the future once the upgrades kick in I assume it will be some time before any improvements will kick in to allow for developers to update their software and hardware manufacturers to update firmwares to support 1.2.0

actually I've been working with hardware people since Jan (because it takes longer for firmware to propagate). at this point it's more the software apps but these are easier to upgrade especially since the API hasn't changed.

probedb

Jul 26 2007, 02:47

Excellent

Now I'd best dig out that script someone posted to run through and reconvert all my FLAC files from 1.1.4.....

Josef Pohm

Jul 26 2007, 02:54

Flac 1.2.0 vs 1.1.4 on a Prescott 2.8ghz UNDERCLOCKED to 1.85ghz.
Ratio is related to my so called SetF.
Average improvement: 20/25% in encoding speed, 5/10% in decoding speed.

CODE

  Mode   Ratio    Enc(1.1.4)Dec    Enc(1.2.0)Dec
|| -0 || 70,03 || 41,7x | 50,5x || 47,9x | 54,3x ||
|| -1 || 68,24 || 39,8x | 46,8x || 45,0x | 51,4x ||
|| -2 || 68,09 || 33,6x | 48,5x || 38,0x | 52,4x ||
|| -3 || 66,35 || 31,6x | 46,1x || 41,4x | 48,7x ||
|| -4 || 64,15 || 24,1x | 44,5x || 34,1x | 47,4x ||
|| -5 || 64,00 || 19,1x | 44,5x || 26,1x | 48,5x ||
|| -6 || 64,00 || 18,6x | 44,4x || 25,1x | 48,0x ||
|| -7 || 63,91 ||  8,3x | 44,6x || 10,1x | 48,4x ||
|| -8 || 63,65 ||  6,0x | 44,3x ||  7,0x | 47,4x ||

culturevulture

Jul 26 2007, 02:56

Great work Josh!

Seiitsu

Jul 26 2007, 04:15

QUOTE(jcoalson @ Jul 26 2007, 06:15)

QUOTE(Seiitsu @ Jul 25 2007, 19:15)

While I do see the point in upgrading the decoders since it will become necessary to do so in the future once the upgrades kick in I assume it will be some time before any improvements will kick in to allow for developers to update their software and hardware manufacturers to update firmwares to support 1.2.0

actually I've been working with hardware people since Jan (because it takes longer for firmware to propagate). at this point it's more the software apps but these are easier to upgrade especially since the API hasn't changed.

Oh, that's cool, then I guess we might see more improvement in flac in the near future than I expected. Nice work Josh.

beto

Jul 26 2007, 08:59

QUOTE(Josef Pohm @ Jul 26 2007, 05:54)

Flac 1.2.0 vs 1.1.4 on a Prescott 2.8ghz UNDERCLOCKED to 1.85ghz.
Ratio is related to my so called SetF.
Average improvement: 20/25% in encoding speed, 5/10% in decoding speed.

CODE

  Mode   Ratio    Enc(1.1.4)Dec    Enc(1.2.0)Dec
|| -0 || 70,03 || 41,7x | 50,5x || 47,9x | 54,3x ||
|| -1 || 68,24 || 39,8x | 46,8x || 45,0x | 51,4x ||
|| -2 || 68,09 || 33,6x | 48,5x || 38,0x | 52,4x ||
|| -3 || 66,35 || 31,6x | 46,1x || 41,4x | 48,7x ||
|| -4 || 64,15 || 24,1x | 44,5x || 34,1x | 47,4x ||
|| -5 || 64,00 || 19,1x | 44,5x || 26,1x | 48,5x ||
|| -6 || 64,00 || 18,6x | 44,4x || 25,1x | 48,0x ||
|| -7 || 63,91 ||  8,3x | 44,6x || 10,1x | 48,4x ||
|| -8 || 63,65 ||  6,0x | 44,3x ||  7,0x | 47,4x ||

which compile did you use? thanx.

Josef Pohm

Jul 26 2007, 10:22

QUOTE(beto @ Jul 26 2007, 16:59)

QUOTE(Josef Pohm @ Jul 26 2007, 05:54)

Flac 1.2.0 vs 1.1.4 on a Prescott 2.8ghz UNDERCLOCKED to 1.85ghz.

which compile did you use? thanx.

This one.

jcoalson

Jul 26 2007, 11:58

QUOTE(probedb @ Jul 26 2007, 03:47)

Excellent

Now I'd best dig out that script someone posted to run through and reconvert all my FLAC files from 1.1.4.....

no need this time, compression will be the same.

DOS386

Jul 26 2007, 12:06

QUOTE

FLAC 1.2.0 is out. There are a few new features and some speedups and fixes, but more importantly, there are some small changes to the decoder to pave the way for possible future compression improvements

QUOTE(jcoalson @ Jul 26 2007, 17:58)

QUOTE(probedb @ Jul 26 2007, 03:47)

Excellent

Now I'd best dig out that script someone posted to run through and reconvert all my FLAC files from 1.1.4.....

no need this time, compression will be the same.

YES. Compressed with 1.1.4 and 1.2.0 and result files were identical except version and date string.

QUOTE

I cannot download from any sourceforge mirror.

No download problems for me.

I wonder whether in future there is a chance for source simplifications or a separate minimal UN-FLAC - couldn't find any so far

... and the full source package is not really tiny or simple ...

bawjaws

Jul 27 2007, 06:50

QUOTE(jcoalson @ Jul 25 2007, 20:15)

QUOTE(Seiitsu @ Jul 25 2007, 19:15)

While I do see the point in upgrading the decoders since it will become necessary to do so in the future once the upgrades kick in I assume it will be some time before any improvements will kick in to allow for developers to update their software and hardware manufacturers to update firmwares to support 1.2.0

actually I've been working with hardware people since Jan (because it takes longer for firmware to propagate). at this point it's more the software apps but these are easier to upgrade especially since the API hasn't changed.

I don't suppose there's been any communication with Apple to ensure the FLAC support they will (apparently) be shipping with Leopard has the necessary support for this. It'd be a shame for them to ship something that had just become outdated. Particularly as they seem to 'lose focus' on a lot of projects, so any future updates are hardly guaranteed.

AngelGR

Jul 27 2007, 07:27

Is there an updated libFLAC.dll for CDex?

john33

Jul 27 2007, 08:23

QUOTE(AngelGR @ Jul 27 2007, 14:27)

Is there an updated libFLAC.dll for CDex?

I've just added the dlls to the 'Lossless' page at Rarewares.

Edit: So they're available for when the API has been sorted in CDex!!

imre_herceg

Jul 27 2007, 09:33

How can you use the new libFLAC.dll with CDex? I use CDex version 1.70b2 unicode version, and with new libFLAC.dll it crashes, just as it did with libFLAC.dll 1.1.4.
Any suggestions?