Aften 0.0.8 is out

[wisodev]

I have builded 64 bit version executable of AFTEN R216 including SSE and SSE3 asm optimization for Windows 64 bit.

Download: BINARIES (http://prdownloads.sourceforge.net/win32builds/aften-svn-r216-win64_sse3-bin.zip?download) | SOURCES (http://prdownloads.sourceforge.net/win32builds/aften-svn-r216-win64_sse3-src.zip?download)

My other builds of AFTEN can be found here (http://win32builds.sourceforge.net/aften/index.html).

[Pookie]

Hi Wisodev :) Welcome to Doom9 forums.

Thanks for your binaries. I like your "FrontEnd" application as well. It looks easy enough to modify the presets and xml so it can work on many different kinds of conversions.

[wisodev]

Hi Wisodev :) Welcome to Doom9 forums.

Thanks for your binaries. I like your "FrontEnd" application as well. It looks easy enough to modify the presets and xml so it can work on many different kinds of conversions.

Thanks.

theForntend is developed with easy usage as a goal (but advanced features are available too ;-)

I am working on new way of handling presets to enable more flexibility, but this is still under development.

About the 64 bit binaries of AFTEN. If anyone has provided some test with my binaries, please send me feedback. I will be grateful for speed and output quality test results (checking if output files are same as 32 bit version is producing).

wisodev

[DarkAvenger]

(checking if output files are same as 32 bit version is producing).


I highly doubt it. Linux and Windows x86_64 uses SSE registers by default - even for scalar operations, while x86 uses i387 by default.

BTW, I would be interested how you compiled the 64 binaries. Did the cmake build system work there, as well? Do you want to try my parallelized version, as well?

[Kurtnoise]

iirc, wisodev uses the ICL compiler...

[DarkAvenger]

cmake works with ICL (at least tested on Linux) as well, or what do you refer to?

[Kurtnoise]

I meant MSVC8 + ICL...Sorry.

[wisodev]

BTW, I would be interested how you compiled the 64 binaries. Did the cmake build system work there, as well? Do you want to try my parallelized version, as well?

Check the build scripts in source package. I am using simple batch scripts for ICL 9.1 and MSVC 8.0 (using cross compiling for 64 bit architectures).

I will try to compile your version this weekend as I will have more free time. I will post results here.

Thanks,
wisodev

[wisodev]

Do you want to try my parallelized version, as well?

I have small problem with your sources, I can't find threading.h header file?

[DarkAvenger]

Oh, all right. I forgot that svn di doesn't list newly added files...too bad it slipped my check. I'll add that file to the zip file and reupload.

OK, should be done, please try again.

[wisodev]

Oh, all right. I forgot that svn di doesn't list newly added files...too bad it slipped my check. I'll add that file to the zip file and reupload.

OK, should be done, please try again.

I downloaded file http://mitglied.lycos.de/darkav/download/aften_parallel3.zip but I don't see any new files. Can you post missing files in post here.

[DarkAvenger]

Ok, now it works, I made sure. Sorry for the hassle. Problem of attaching files here is, they must be made accessible by the admins...which could time...

[wisodev]

Ok, now it works, I made sure. Sorry for the hassle. Problem of attaching files here is, they must be made accessible by the admins...which could time...

OK I have downloaded the archive with threading.h and now I can start building binaries.

Thanks,
wisodev

[wisodev]

I have builded 64 bit version executable of AFTEN R216 including SSE and SSE3 asm optimization for Windows 64 bit using parallelized version by Prakash Punnoor (original sources (http://mitglied.lycos.de/darkav/download/aften_parallel3.zip)).

Download: BINARIES (http://prdownloads.sourceforge.net/win32builds/aften-svn-r216-mt-win64_sse3-bin.zip?download) | SOURCES (http://prdownloads.sourceforge.net/win32builds/aften-svn-r216-mt-win64_sse3-src.zip?download)

Thanks,
wisodev

[ADLANCAS]

how's development coming? (32bit) :)

[tebasuna51]

how's development coming? (32bit) :)
Seems stopped.
After this post (http://forum.doom9.org/showthread.php?p=897950#post897950) Justin is missing. Maybe is discouraged for the little feedback about DRC?.

[DarkAvenger]

He is still waiting for getting his internet access back...

[jruggle]

He is still waiting for getting his internet access back...
Indeed I am still waiting. Development has not stopped though. Just in the last week I've been doing some work on Aften. I reimplemented a completely new exponent strategy decision algorithm. Some of the bit allocation parameters have also been tweaked. The big news is that I have channel coupling about 90% working! Right now it decodes without error, but one channel ends up with noise for some reason. I hope to get that fixed in the next day or two. I do need my internet back to continue the work on DRC for the online references, but that won't be too long hopefully.

-Justin

[vlada]

Hi Justin,
I have one question. I'm working on a GUI application which should be able to open almost any video file and convert it to DVD-Video. It will be using AviSynth as the basic engine. I'd like to use Aften to compress a sound to AC3. Can I use AviSynth script as source for Aften? If not, is AviSynth input in you to-do list? Thank you.

[tebasuna51]

@vlada
sh0dan is working to encode ac3 inside AviSynth script:
http://forum.doom9.org/showthread.php?p=925907#post925907

and i think he want use Aften library http://forum.doom9.org/showthread.php?p=920344#post920344

[Pookie]

vlada -

bepipe.exe --script "import(^avisynth.avs^)" - | aften.exe -v 0 - -b 256 ac3file.ac3

Too bad Bepipe needs .NET 2.0. sh0dan to the rescue? ;)

[Mug Funky]

avs2wav :search:

"c:\path\to\avs2wav.exe" "%~1" - | "c:\path\to\aften\aften.exe" -b 224 -s 1 -m 1 -w 45 -dnorm 27 - "%~dpn1.aftend.ac3"

that's the line i use. it's suitable for putting in a batch file.

[edit]

@ jruggle - good news on the channel coupling! once that and DRC are going this will be my preferred encoder, at least for my own stuff (you can't use a non dolby certified encoder on a DVD and be entitled to put the double-D logo on the disc)

[tebasuna51]

avs2wav :search:

"c:\path\to\avs2wav.exe" "%~1" - | "c:\path\to\aften\aften.exe" -b 224 -s 1 -m 1 -w 45 -dnorm 27 - "%~dpn1.aftend.ac3"

avs2wav don't work for me.
I found avs2wav-v1.1.zip who say me:
avs2wav v1.1 by Jory Stone <jcsston@toughguy.net>, stdout patch by kassandro
Input: d:\Programa\Audio\0\Audio.avs
Output: -
Scanning for Audio Stream...

Gaining with a factor of 1.0000 ...
but the output file is always empty.

And kassandro's avs2wav.rar who say me:
Don't found MSVCP70.dll

Bepipe works without problems for me.

[Mug Funky]

MSVCP7x dlls can be found with virtualdub, and a few other programs. if a system search doesn't turn it up, try google or something. it does suck to need runtimes though...

once you find these dll's, either put them somewhere in path, or just in the directory with the program that requires them.

[tebasuna51]

@Mug Funky, thanks for your answer but still don't work for me.

For don't interfere with Aften thread, I reopen the old thread
Wanted: avs2wav (http://forum.doom9.org/showthread.php?p=926685#post926685) to continue this topic if any is interested.

[jruggle]

Hi all,
Just wanted to vent my enthusiasm. :) I got channel coupling working. I'll be able to apply the changes within the next week (as soon as AT&T gets my DSL turned on). It's pretty basic right now because I don't have much to go on as far as the best encoding algorithm to calculate the coupling channel and coupling coefficients. The spec gives a basic idea, but the implementation is not quite so straight-forward.

The plan is to eventually have an adaptive coupling strategy for each frame, but for now I just picked one for stereo and one for multi-channel based on some samples from DVD's. It will just take more analysis of professionally-encoded AC3 files to figure out the best way forward.

On that note, I also got my AC3 frame analyzer working 100% now. It's in the basic preliminary stages, but will eventually be able to do fun stuff like modify metadata, DRC, etc... of existing AC3 files.

-Justin

[BigDid]

Hi all,
Just wanted to vent my enthusiasm. :)
...
On that note, I also got my AC3 frame analyzer working 100% now. It's in the basic preliminary stages, but will eventually be able to do fun stuff like modify metadata, DRC, etc... of existing AC3 files.

-Justin
Happy new year and all my best, glad to read the good news, I'll be waiting for it.

Did

[Mug Funky]

will eventually be able to do fun stuff like modify metadata, DRC, etc... of existing AC3 files

ach! now you're just teasing :). i'm not sure if a tool like this exists anywhere, commercial or otherwise, but it's certainly going to be an extremely useful tool.

happy new year and thanks heaps for keeping this going.

btw, the channel coupling looks good. analysing existing DVDs seems like a good way to do things. one thing though - try checking out movies of the whole spectrum (particularly budget). i've noticed that lower budget movies have rather predictable, generic 5.1 mixes that could help you.

[raquete]

encoding with Aften have more presence and quality in the audio than using softencode in my taste but i'm having only one issue: authoring musics with dvdlab,gui for dvdauthor or with a single audiodvdcreator give too big click between tracks.
if i use the poor ac3 encoder inside audiodvdcreator the sound is worse(less levels/presence) comparing with Aften but don't have clicks or issues,the passages are clean and soft.
was hard to find what could be the reason,only after encoding with another ac3 encoder i could compare the results.
if someone want samples(lots of people don't like)i can host with clicks(from aften)and clean(from audiodvdcreator) results authored in video_ts folder.
anyone have one hint to help me?

[Mug Funky]

i've not encountered clicks using spruce's AV sync track to join audio, nor from binary merging ac3's created by aften.

could you go over your workflow a little? programs etc.

clicks between assets could be wav headers being encoded as audio, or it could be something else.

[raquete]

clicks between assets could be wav headers being encoded as audio...i was clear and was happening.authoring the same sources using Aften(14-07-06)have clicks but not using advdc.
after change to Aften-0.04(2006-Aug-06)no more clicks...seems the solution and deserve more tests,i'm doing!
it all takes me to one question: i'm using the GUI.what Aften version i have to use?
regards

[raquete]

Mug Funky,
i authored 4 full albums in "Gui for dvdauthor" using Aften-0.04 and all is fine,no more clicks. ;)
can you tell me what Aften version are you using and if this version works fine with AftenGUI?
thanks so much!

[Mug Funky]

i think it's .05 - i just noticed it doesn't tell me the version number with aften -h

file is 60,928 bytes, created on wed 23/08/06, 10:56:44 am

i don't know if it works with aftenGUI - i only do 224kbps stereo encodes with it, so i don't need a GUI :)

[tebasuna51]

The last version in http://sourceforge.net/projects/aften (Justin compiles) is:
aften.exe v0.05 21/08/2006 58368 bytes

But after Justin was add the DRC support at 23/10/2006 with daily builds.
Now I can't access Justin compiles (http://jbr.homelinux.org/aften/daily/).

There are KurtNoise/DarkAvenger compiles at http://kurtnoise.free.fr/index.php?dir=Aften/ :
...
aften-0.04 06/08/2006 44032 bytes
aften-0.05 23/08/2006 52736 bytes
aften-0.05_rev214 30/10/2006 679676 bytes
aften-0.05_rev216 23/11/2006 127261 bytes
...
AftenGUI-1.2 for Aften 0.04
AftenGUI-1.3 for Aften 0.05 (+ DRC)
...
Between 3/10 and 30/10 there was some rev with bugs, see http://forum.doom9.org/showthread.php?p=893510#post893510

To see differences between version you can see the blog (http://aftenblog.blogspot.com/).
For work with DRC we can use AftenGUI-1.3 and 0.05_rev216.

[raquete]

first....thank you both for answers!
a long time i call for one Aften little guide...as it "never came" i have one (or two) more questions/doubts: i always adjust DRC to "none" because i encode only musics in 5.1. i have all Aften's versions here and all AftenGUI's saved.
please tell me what version i have to use with the GUI.
thanks.

[tebasuna51]

You can use:
AftenGUI-1.2 for Aften 0.04 or 0.05 (before 23/10/2006)
AftenGUI-1.3 for Aften 0.05 (DRC, 23/10/2006 and news)

the first rev with DRC I have is v0.05_rev185 (23/10/2006) from KurtNoise site, but is buggy (until rev_211). The first ok is rev_212 (29/10/2006) but the recommended is rev_216.

If you don't need DRC you can use AftenGUI-1.2 with last Justin version aften.exe v0.05 21/08/2006 58368 bytes (works fine for me).

[raquete]

woo...thanks,now is very clear!

[wisodev]

I've builded latest Aften binaries from SVN sources at revision 224. The binaries can be downloaded from my website (http://win32builds.sourceforge.net/aften/index.html).

[raquete]

@ tebasuna51
i can feel but i still can't proove that aften 0.05 results are "rispids" but clean with 0.04 version.to feel is needed to relax and listen one know and clean but powerfull source like king crimson or seamless....mfsl or from hdcd sources if possible.
if you don't mind can you do few tests please?

[tebasuna51]

@raquete
My ears can't appreciate any diference (I'm sure the problem is in my ears).

Checking the waveforms I see little differences between original and encodes with 0.04, 0.05. Maybe there are problems using 0.05 and mid/sid rematrixing (default).

Try using 0.05 and independent L+R channels (-m 0), seems work better for high bitrates.

BTW Justin is working in channel coupling now. Maybe there are changes in next version.

[raquete]

ok,i will try independent L+R channels.
waiting for the news too!
thanks for your tests.

[Chainmax]

...
BTW Justin is working in channel coupling now. Maybe there are changes in next version.

Great news, I will finally ditch Scenarist's encoder and switch to AftenGUI as soon as a version with channel coupling is released :). Thanks for all the hard work Justin, and thanks for the compiles and updates, tebasuna51 http://smilies.vidahost.com/otn/wink/thumb.gif.

[jruggle]

BTW Justin is working in channel coupling now. Maybe there are changes in next version.

Yes, version 0.06 is forthcoming, and maybe it will solve the problem...if we're lucky...otherwise I'll try to reproduce your issue & see if I can figure it out.

Version 0.06 will pretty much be an SVN snapshot once a couple small issues get sorted out. It will not have channel coupling. Prakash is still working on parallel encoding, and we want to get that applied first since the channel coupling will be a large commit affecting several files. Both of those changes will be post-v0.06. The Changelog (http://aften.svn.sourceforge.net/viewvc/*checkout*/aften/Changelog?revision=240) in SVN gives a good idea of what will differ between 0.05 and 0.06.

There are many other good things on the TODO list as well. I've recently been playing around with adapting twolame's psychoacoustic model 3 (model 1 from the mpeg spec) for use with Aften. I've got the psych model working...now I just have to implement delta bit allocation in order to take advantage of it. Anyway, there is so much I'm working on it's a bit overwhelming. I'm just taking it one step at a time as far as integration to svn, but in the meantime I'm testing out more features.

-Justin

[Mug Funky]

if you're into psy-models, you might want a look at the Musepack one. it's a codec derived from mp2 (32 subbands and all that), but manages to score slightly more efficient than much more advanced codecs when aiming for transparency.

not sure how much of it would be applicable to ac3 though.

good to hear about the channel coupling :)

[jruggle]

ok,i will try independent L+R channels.
waiting for the news too!
thanks for your tests.

The only thing of consequence I can find that changed between 0.04 and 0.05 which might be affecting output is, indeed, the stereo rematrixing. I cannot hear any differences though...

However, if I am guessing correctly, the problem might be occuring when exponents are being reused from the previous block, where one of the two blocks uses rematrixing and the other does not. This may lead to some inaccurate exponents, which can sometimes be heard in the output. I've had similar issues come up in the course of the psy-model tests. I believe the newest exponent strategy routine, which has been added since 0.05, would resolve the issue since it uses a least-error method rather than a fixed threshold.

The only thing I can think of as to why 0.04 did not manifest the same problems is that the rematrixing decision algorithm was straight from the spec and probably did not do rematrixing as often. Therefore, the issue may have still been present, but would have been less noticeable.

edit: Did some testing...0.05 and later rematrixing had a bug which chose to do rematrixing WAY too much (95% vs. 17%). I just changed this in SVN so it will be fixed in 0.06 as well.

Then again, I may be wrong about all of this. :)

-Justin

P.S. If all goes as planned, Aften 0.06 will be coming out on 1/28.

[Kurtnoise]

@DarkAvenger : for makefile creation with cmake, I've got this
-- Using YASM/NASM
-- Performing Test HAVE_NASM_VISIBILITY
-- Performing Test HAVE_NASM_VISIBILITY - Failure
What is it exactly ?



And before to release the 0.06, why not fix also these warnings :
[ 61%] Building C object CMakeFiles/aften_static.dir/libaften/x86/x86_sse_mdct_common_init.obj
c:/temp/aften/libaften/x86/x86_sse_mdct_common_init.c: In function 'sse_mdct_ctx_init':
c:/temp/aften/libaften/x86/x86_sse_mdct_common_init.c:143: warning: 'XMM2' may be used uninitialized in this function
c:/temp/aften/libaften/x86/x86_sse_mdct_common_init.c:143: warning: 'XMM0' may be used uninitialized in this function
c:/temp/aften/libaften/x86/x86_sse_mdct_common_init.c:169: warning: 'XMM2' may be used uninitialized in this function
c:/temp/aften/libaften/x86/x86_sse_mdct_common_init.c:169: warning: 'XMM0' may be used uninitialized in this function
[ 66%] Building C object CMakeFiles/aften_static.dir/libaften/x86/x86_sse3_mdct_dummy.obj
[ 71%] Building ASM object CMakeFiles/aften_static.dir/libaften/x86/x86_cpu_caps_detect.obj
Linking C static library libaften_static.a
[ 71%] Built target aften_static
Scanning dependencies of target aften_wav
[ 76%] Building C object CMakeFiles/aften_wav.dir/aften/wav.obj
Linking C static library libaften_wav.a
[ 76%] Built target aften_wav
Scanning dependencies of target aften_exe
[ 80%] Building C object CMakeFiles/aften_exe.dir/aften/aften.obj
Linking C executable aften.exe
[ 80%] Built target aften_exe
Scanning dependencies of target wavfilter
[ 85%] Building C object CMakeFiles/wavfilter.dir/util/wavfilter.obj
[ 90%] Building C object CMakeFiles/wavfilter.dir/libaften/filter.obj
Linking C executable wavfilter.exe
[ 90%] Built target wavfilter
Scanning dependencies of target wavinfo
[ 95%] Building C object CMakeFiles/wavinfo.dir/util/wavinfo.obj
c:/temp/aften/util/wavinfo.c: In function 'wavinfo_print':
c:/temp/aften/util/wavinfo.c:314: warning: format '%ld' expects type 'long int', but argument 2 has type 'int64_t'
Linking C executable wavinfo.exe
[ 95%] Built target wavinfo
Scanning dependencies of target wavrms
[100%] Building C object CMakeFiles/wavrms.dir/util/wavrms.obj
Linking C executable wavrms.exe
[100%] Built target wavrms

I use gcc 4.1.1 & nasm 0.98

[DarkAvenger]

@DarkAvenger : for makefile creation with cmake, I've got this

What is it exactly ?


Unimportant on Windows, probably I should deactivate that test in Windows. (On linux all symbols are public/visible by default in a shared lib, so I detect whether nasm/yasm can hide them.)



And before to release the 0.06, why not fix also these warnings :


I use gcc 4.1.1 & nasm 0.98

The warnings on mdct are wrong. gcc (nor icc, but there you can deactivate them) is not smart enough.

The last warning in wavinfo was fixed by Justin.


BTW, does anybody have an intel Mac to test compilng aften in MacOS X?

[Kurtnoise]

gcc is not smart enough...funny. :D


btw, thanks for the channels mapping.

[yuvi]

BTW, does anybody have an intel Mac to test compilng aften in MacOS X?

It fails as such with the cmake generated makefile (the Xcode generation fails internally to cmake due to CMAKE_EXE_LINKER_FLAGS not being set during the nasm visibility test):

[ 71%] Building ASM object CMakeFiles/aften_static.dir/libaften/x86/x86_cpu_caps_detect.o
/usr/bin/nasm -f elf -I/Users/yuvi/aften/default -I/Users/yuvi/aften -I/Users/yuvi/aften/libaften -I/Users/yuvi/aften/aften -I/Users/yuvi/aften/libaften/x86 -DHAVE_CONFIG_H -DAFTEN_BUILD_LIBRARY -o CMakeFiles/aften_static.dir/libaften/x86/x86_cpu_caps_detect.o /Users/yuvi/aften/libaften/x86/x86_cpu_caps_detect.nasm
Linking C static library libaften_static.a
/usr/bin/cmake -P CMakeFiles/aften_static.dir/cmake_clean_target.cmake
/usr/bin/cmake -E cmake_link_script CMakeFiles/aften_static.dir/link.txt --verbose=1
/usr/bin/ar cr libaften_static.a "CMakeFiles/aften_static.dir/libaften/a52enc.o" "CMakeFiles/aften_static.dir/libaften/bitalloc.o" "CMakeFiles/aften_static.dir/libaften/bitio.o" "CMakeFiles/aften_static.dir/libaften/crc.o" "CMakeFiles/aften_static.dir/libaften/dynrng.o" "CMakeFiles/aften_static.dir/libaften/window.o" "CMakeFiles/aften_static.dir/libaften/mdct.o" "CMakeFiles/aften_static.dir/libaften/exponent.o" "CMakeFiles/aften_static.dir/libaften/filter.o" "CMakeFiles/aften_static.dir/libaften/util.o" "CMakeFiles/aften_static.dir/libaften/x86/x86_cpu_caps.o" "CMakeFiles/aften_static.dir/libaften/x86/x86_sse_mdct_dummy.o" "CMakeFiles/aften_static.dir/libaften/x86/x86_sse_mdct_common_init.o" "CMakeFiles/aften_static.dir/libaften/x86/x86_sse3_mdct_dummy.o" "CMakeFiles/aften_static.dir/libaften/x86/x86_cpu_caps_detect.o"
/usr/bin/ranlib libaften_static.a
/usr/bin/cmake -E cmake_progress_report /Users/yuvi/aften/default/CMakeFiles 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

[...]

[ 80%] Building C object CMakeFiles/aften_exe.dir/aften/aften.o
/usr/bin/gcc -Wno-switch -Wextra -Wdeclaration-after-statement -Wbad-function-cast -Wfloat-equal -Wdisabled-optimization -std=gnu99 -pedantic -Wall -Wpointer-arith -Wredundant-decls -Wformat -Wunused -fvisibility=hidden -funroll-loops -fomit-frame-pointer -O3 -DNDEBUG -I/Users/yuvi/aften/default -I/Users/yuvi/aften -I/Users/yuvi/aften/libaften -I/Users/yuvi/aften/aften -I/Users/yuvi/aften/libaften/x86 -DHAVE_CONFIG_H -DAFTEN_BUILD_LIBRARY -o CMakeFiles/aften_exe.dir/aften/aften.o -c /Users/yuvi/aften/aften/aften.c
Linking C executable aften
/usr/bin/cmake -P CMakeFiles/aften_exe.dir/cmake_clean_target.cmake
/usr/bin/gcc -Wno-switch -Wextra -Wdeclaration-after-statement -Wbad-function-cast -Wfloat-equal -Wdisabled-optimization -std=gnu99 -pedantic -Wall -Wpointer-arith -Wredundant-decls -Wformat -Wunused -fvisibility=hidden -funroll-loops -fomit-frame-pointer -O3 -DNDEBUG -headerpad_max_install_names -fPIC "CMakeFiles/aften_exe.dir/aften/aften.o" -o aften -L/Users/yuvi/aften/default -laften_wav -laften_static -lm
/usr/bin/ld: Undefined symbols:
__alDetectx86CPUCaps
collect2: ld returned 1 exit status
make[2]: *** [aften] Error 1
make[1]: *** [CMakeFiles/aften_exe.dir/all] Error 2
make: *** [all] Error 2

The problem is that you're specifying "-f elf" to nasm, where Mac OS X uses Mach-O. I don't know cmake well enough to fix it myself, and unfortunately I won't have access to an Intel Mac again for some time to test further.

[HeadBangeR77]

Sorry to interrupt that developement-compilation discussion with a noob-like question. ;)
Which revision of Aften & its GUI would you recommend? I should add I won't be using DRC at all. And the 2nd one: which container could handle AC3 VBR + XviD/x264?

Hope I didn't bother too much. ;)
<...and he disappeared into thin air>

cheers,
HDBR77