ffdshow and ffdshow_tryout Builds Listing

[netwolf]

I have a general question regarding SSE/SSE2 etc.:

when you say "requires SSE", does that mean that only SSE is supported (so no SSE2), or does it rather mean that SSE is a minimum requirement, but if your CPU supports more, also SSE2 etc. are used ?

If the latter is true, what's the point in providing SSE2 builds anyway (if you get the same results with SSE builds)?

Or do SSE2 builds still work better on SSE2 capable CPUs than SSE-only builds would do?

thx.

[videomixer9]

requires means that it has to at least have SSE, of course SSE2 intrinsics and other do also work. The only project that doesn't come with specific optimization is the ffdshow.ax that may or may not profit much from automatic optimization done by the compiler. My test show that I had no real performance gain on my SSE2 P4 with specials SSE2 builds, the only differences are there where automatic vectorization kicks in, and for libmplayer and libavcodec which are both the most used workhorses and are usually compiled with GCC this doesn't make much sense as GCC doesn't autovectorize good enough yet. Generally there are not much optimizations for SSE2 or even SSE3 in ffdshow as I see it.

libavcodec always makes use of anything available automatically, even with generic builds that only require MMX.

Take VLC as an example, VLC also uses libavcodec and is compiled usually for generic CPU and still fully uses all optimizations. mplayer is the same, there's only a slight gain cause not every passage of code is fully hand optimized and e.g. memory copy actions can be handled different.

Actually some generic builds seem to be better performing on my P4 just for the decoding part than some optimized builds, so I don't think all these versions are really neccessary.

e.g. I doubt that clsid's builds that only require MMX are worse performing than all other builds. The only thing with the SSE2 thing is that some libs will assume that SSE2 is available and automatically use the optimizations always without any checks, thus on non-SSE2 CPUs they will randomly crash at some point, usually this only happens with libdts or liba52 most often.

[videomixer9]

ffdshow rev. 2535

requires: SSE
compilers: GCC 4.1.1 + MSVC 8
specials: high accuracy tremor, more experimental gcc switches
download: here (http://www.filepoint.de/download/2883-dl-ffdshow-20060428-rev2535_exe) or here (http://rapidshare.de/files/19125295/ffdshow-20060428-rev2535.exe.html) or here (http://www.megaupload.com/de/?d=H9P10Z1M)

changes: vertical and horizontal image flipping

[videomixer9]

ffdshow rev. 2536 (unicode as always)

requires: SSE
compilers: GCC 4.1.1 + MSVC 8
specials: high accuracy tremor, even more experimental gcc switches
download: here (http://www.filepoint.de/download/2895-dl-ffdshow-20060428-rev2536_exe) or here (http://rapidshare.de/files/19132438/ffdshow-20060428-rev2536.exe.html) or here (http://www.megaupload.com/de/?d=GH3BR3ES) (2.90 MB)

changes: DeBand filter

[videomixer9]

ffdshow rev. 2538 MULTITHREADED VERSION

requires: SSE
compilers: GCC 4.1.1 + MSVC 8
specials: high accuracy tremor
download: here (http://bittekeinspam.googlepages.com/ffdshow-20060429-mt-rev2538.exe)

the patch is note exactly made for this revision so it's a unknown to me if it works.

[netwolf]

@videomixer9: thank you for the detailed explanation !
So in the futureI don't have to eagerly wait for a special SSE2 build if there are important changes in SVN and there is "only" an SSE2 version available yet :)

[videomixer9]

ffdshow rev. 2544

requires: SSE
compilers: GCC 4.1.1 + MSVC 8
specials: high accuracy tremor, high accuracy libmad, runtime cpu detection
download: here (http://bittekeinspam.googlepages.com/ffdshow-20060504-rev2544.exe)

Multithreading patch by Haruhiko Yamagata is now officially part of ffdshow. Also new is Aud-X support, audxlib.dll comes with installer, get sure to read the license here (http://bittekeinspam.googlepages.com/AudX_licence.pdf).

[Liisachan]

Multithreading patch by Haruhiko Yamagata is now officially part of ffdshow. Congrats, haruhiko! :cool:

[odyn1982]

Thank you videomixer9 for your great ffdshow builds!:thanks:

[LoRd_MuldeR]

ffdshow rev. 2544
Also new is Aud-X support, audxlib.dll comes with installer

I never used Aud-X before, but I can't see "Aud-X" on the Audio Codecs page of ffdshow. How to enable Aud-X support? Enabled by default?

[videomixer9]

no clue, I missed it there too but I guess it is enabled per default via mp3 decoding as it is an extension to mp3. Supposedly the dll should appear in the list in the about dialog, whatever ... nothing here, no clue what works.

[LoRd_MuldeR]

no clue, I missed it there too but I guess it is enabled per default via mp3 decoding as it is an extension to mp3.

Found it :D

It is available as a decoder for MP3. You can select it instead of "mp3lib" or "libmad". Don't know if it's good to use Aud-X decoder, when your MP3 audio is not Aud-X though...

[videomixer9]

maybe it upsamples to surround ?_?

ffdshow rev. 2545

requires: SSE
compilers: GCC 4.1.1 + MSVC 8
specials: high accuracy tremor, high accuracy libmad, runtime cpu detection
download: here (http://bittekeinspam.googlepages.com/ffdshow-20060504-rev2545.exe)

software scaler cosmetics? whatever, this also comes with audxlib.def as I noticed that there are prolly load errors without it, gotta hate this external stuff. Aud-X may still crash if it's used as regular mp3 audio decoder all the time, at least it crashed for me after a while.

[Avish]

maybe it upsamples to surround ?_?

ffdshow rev. 2545

requires: SSE
compilers: GCC 4.1.1 + MSVC 8
specials: high accuracy tremor, high accuracy libmad, runtime cpu detection
download: here (http://bittekeinspam.googlepages.com/ffdshow-20060504-rev2545.exe)

software scaler cosmetics? whatever, this also comes with audxlib.def as I noticed that there are prolly load errors without it, gotta hate this external stuff. Aud-X may still crash if it's used as regular mp3 audio decoder all the time, at least it crashed for me after a while.Does videomixer9's "specials" aslo available in other's builts?

[videomixer9]

ffdshow rev. 2546

requires: SSE
compilers: GCC 4.1.1 + MSVC 8
specials: high accuracy tremor, high accuracy libmad, runtime cpu detection, aud-x libs incl.
download: here (http://bittekeinspam.googlepages.com/ffdshow-20060505-rev2546.exe)

LPCM seems to be broken for some unclear reason, it seems to be the case in other builds too so I don't think it's an compiler error. LPCM is not hard on the CPU and doesn't need advanced decoding. The decoder is embedded into the ffdshow.ax. If you use e.g. MPC just use the internal LPCM decoder, it's uncompressed anyways thus quality is equal on any decoder too.

[IgorC]

Maybe it will be better to suscribe ffdshow new revision as sticky here http://forum.doom9.org/showthread.php?t=89979

and don't make a new post each time.

Just advice.

[Liisachan]

As many of you already knew, recently celtic_druid restarted to build ffdshow too.
http://esby.free.fr/CelticDruid/
http://ffdshow.faireal.net/

[clsid]

SVN: 2006-05-05 (rev. 2546)
Compilers used:
- MSVC71 (ffdshow.ax, etc)
- GCC 3.4.5 (libavcodec.dll, mplayer.dll, kerneldeint.dll)
Patches:
- high accuracy tremor
- high accuracy libmad
Minimum CPU requirement: MMX
Download: click here (http://rapidshare.de/files/19968328/ffdshow-20060505-rev2546.exe.html)

[igor1st]

SVN: 2006-05-05 (rev. 2546)

Great build, thank you.
But for me any ffdshow build since rev2544 crashes on seeking. And this occurs without enabling multithreaded video decoding in options.
So I proceed with rev2541 and wait until Milan will fix this.

[MatMaul]

Great build, thank you.
But for me any ffdshow build since rev2544 crashes on seeking. And this occurs without enabling multithreaded video decoding in options.
So I proceed with rev2541 and wait until Milan will fix this.

yeah !
I have the same problem, but a lot of person don't have this problem.
You have this problem with MPC?
Can you post your cpu type please ?

[Gord]

Mine did the same thing until I changed the MP3 decoder.

libmad and aud-x don't crash it... mp3lib does.

[igor1st]

yeah !
I have the same problem, but a lot of person don't have this problem.
You have this problem with MPC?
Can you post your cpu type please ?
Athlon XP 3200+, MPC rev609.
Seeking crashes reproduces easily by seeking with hotkeys on few steps (one step == 10s for me). Crashes with seeking by 'seek bar' are much more rare.

BTW, not ffdshow himself crashes - problem with libavcodec. And this confirms by switching the decoder (for example from lavc to xvid for xvid video).

libmad and aud-x don't crash it... mp3lib does.
I try default ffdshow configuration and my tweaked one (btw, I always use libmad). Problem exist since rev2544.

[Gord]

I try default ffdshow configuration and my tweaked one (btw, I always use libmad). Problem exist since rev2544.

Changing the mp3 decoder lib was what fixed it for me anyway... might be another issue though.

[MatMaul]

it's not mp3lib for me too : I use libmad

[3dsnar]

Hi,
@ Videomixer:
Is it possible to distribute the audxlib.dll
with the installer? (I think it is not available by default).
Please let me know, 3d

[videomixer9]

it's in the installer already, aud-x should work per default with my builds. Check About -> Version details, it should say audxlib.dll present there.

[Liisachan]

seek->crash ocasionaly occurs to me, even with ffdshow MP3 decoder disabled.

[MacAddict]

Seeking in MPC has been crashing the past few days for me as well, just randomly. I'm only playing VOB files _without_ any sort of MP3 decoding though. MPC build is 609.

[3dsnar]

it's in the installer already, aud-x should work per default with my builds. Check About -> Version details, it should say audxlib.dll present there.
OK. I have downloaded FFDshow (rev2546-SSE2) from www.free-codecs.com and the DLL is not in the installer. Why?
------
The mp3 Aud-X stream contains stream indicator (once in a while string 'Aud-X1.0' is present in the stream. Based on this it is possible to distinguish regular mp3 from aud-x mp3. Would it be possible to provide for the user some 'activate mp3 5.1 decoding' switch (instead of the library choice, as an alternative mp3 decoding engine). So if the switch would be ON, and after detecting the 'Aud-X1.0' string, Aud-X DLL would be automatically used. I hope you know what I mean.
Please let me know what you think.
:thanks:

[videomixer9]

cause that is not my build. ask celtic_druid for that.

[3dsnar]

OK, how about the option that I proposed?

[videomixer9]

ask milan for that, he's still the author of ffdshow, use sourceforge feature requests page for that or submit a patch.

http://sourceforge.net/tracker/?group_id=53761&atid=471492

[kurt]

I also have problems wit celtic_druid's latest Build (ffdshow-rev2546-SSE2.exe). I can't play xvid files anymore, libvacodec doesn't seem to be registered...

with rev2529 (sse2) there are no such problems...

[celtic_druid]

Yeah, I didn't include it.

libavcodec doesn't need registering. I'll have a look when I get home. I know at least that AVC decoding works and that uses libavcodec.

[3dsnar]

Yeah, I didn't include it.

libavcodec doesn't need registering. I'll have a look when I get home. I know at least that AVC decoding works and that uses libavcodec.
How about adding audxlib.dll to the installer (it is only less than 200 kB, when compressed)?

[kurt]

Yeah, I didn't include it.

libavcodec doesn't need registering. I'll have a look when I get home. I know at least that AVC decoding works and that uses libavcodec.
after reinstalling it again, it works now. It might be because of my strange system. Sorry for the inconvenience :p

[chipzoller]

I hope this is the right place to post this, if not please move it.

I wanted to test out vobsub support in ffdshow and so unregistered vsfilter.dll from my system and shifted all the vobsub decoding to ffdshow but found it doesn't work 100% of the time. BTW, I searched this thread and found no mention of this problem. Apparently, it really doesn't like it (in my MKV with vobsubs) if you seek ahead then seek back...the ahead seek will usually display subs but seeking back won't. Is there something wrong here in ffdshow, or is its support experimental? I'd like to see if I can get this working so I can limit my software packages down to just 2 in order to redistro. my encodes.

[Liisachan]

Generally sub support of ffdshow is experimental, at least for SSA/ASS/ASS2. Not very sure about Vobsub.

[foxyshadis]

Does FFDShow use code from vsfilter? If it does, it must be quite old, since many things don't render correctly with it. (It'd be nice if it was as simple as updating the code, but I suppose it probably isn't.)

[videomixer9]

it shouldn't use even one line of code from vsfilter

[chipzoller]

Vobsub support in ffdshow seems to also be experimental at best...I sure hope this area gets some attention soon.

[Inventive Software]

Because NSIS is quite configurable, it should be quite easy to have a page for all the free codecs, then all the patent-risky ones, both for audio and video. I'll have a look tonight, been ages since I dabbled in NSIS!

Oh, in case this gets too off-topic, can we keep it focused pleeeaase? First post in this thread has the link for the unofficial thread (now at more than 1000) for bug reports and the like. Here it is again in case you've forgotten / can't be bothered to find it (;)): ffdshow bug report / feature request thread (http://forum.doom9.org/showthread.php?t=98600). This thread is for the builders to post what they have enabled, and to keep a log of all the builds made. Videomixer9 has done an incredible job with the regularity of his builds, and his work is VERY MUCH APPRECIATED!!

[3dsnar]

Because NSIS is quite configurable, it should be quite easy to have a page for all the free codecs, then all the patent-risky ones, both for audio and video. I'll have a look tonight, been ages since I dabbled in NSIS!

Oh, in case this gets too off-topic, can we keep it focused pleeeaase? First post in this thread has the link for the unofficial thread (now at more than 1000) for bug reports and the like. Here it is again in case you've forgotten / can't be bothered to find it (;)): ffdshow bug report / feature request thread (http://forum.doom9.org/showthread.php?t=98600). This thread is for the builders to post what they have enabled, and to keep a log of all the builds made. Videomixer9 has done an incredible job with the regularity of his builds, and his work is VERY MUCH APPRECIATED!!
I am sorry. I have deleted my request and posted it
where you suggested.:o

[videomixer9]

ffdshow64 rev. 2546

requires: SSE3
compilers: ICL 9.1.022
specials: high accuracy tremor, high accuracy libmad, libavcodec 6ch vorbis fix
download: here (http://bittekeinspam.googlepages.com/ffdshow64-20060512-rev2546-icl91.exe)

pure ICL9.1 build incl. libavcodec and libmplayer

[LotharZ]

Im geting a strange error with your new compilation.

http://img125.imageshack.us/img125/6272/ffdshowerror1cl.th.jpg (http://img125.imageshack.us/my.php?image=ffdshowerror1cl.jpg)

http://img89.imageshack.us/img89/7746/aviproperties5tn.th.jpg (http://img89.imageshack.us/my.php?image=aviproperties5tn.jpg)

Edit: Problem, seems that comes from the AC3 audio stream.

[videomixer9]

hm I don't have an 64bit CPU and thus no win64 so I really cannot test it. Wonder what this is about, I checked the projects and there was no duplicate linking to libguide ... wonder how that happens ...

libguide is used for automatic parallelization, liba52 uses this amongst others, this error may occur only randomly I was told somewhere else so it shouldn't make any problems all the time, in fact chances to get it are pretty low and I bet most people will never see it. there is nothing to change with the way of using it, everything is done automatic and I just added libguide.lib to other dependencies.

[LotharZ]

Oops my bad, isnt the 64bits build, I was speaking about the "ffdshow-20060512-rev2546-icl91.exe (3.96 MB)" compiled with ICL9.1.022 + GCC 4.1.1.

And isnt a random problem, always happens with AC3 files. Its the first time that happens with your builds.

[videomixer9]

That's odd, I didn't see any problems and so far noone else seems to had this one, I usually test an AC3, DTS, OGG and AAC movie before releasing a build. I tried multiple demos I had and had no problem, I also got that same file you got there and it works fine. I found that tremor can be crashed this way using parallelization, need to investigate some more on it and maybe turn it off, though it's odd if it's just happens on one system. Wonder if it only happens on specific cpu, e.g. your sempron if you still got that an win2003 iirc.

However I silently replace files sometimes and I did with this one too, check if your build date of liba52 is 20:01.

[LotharZ]

Yeah, its May 12 2006 20:01.32

Don't worry too much, probably its only a conjuntion of ffdshow options and my system, I'll try with your next build.

[videomixer9]

please try this (http://bittekeinspam.googlepages.com/ffdshow-20060513-rev2546-icl91.exe) build, auto parallelization is off for liba52, libdts and tremor, not that it parallelized anything that worthy anyways. However auto-parallelization manages to fuck up the code.