New ffdshow build (?)

[Reino]

ffdshow_rev514_20061109_clsid.exe:

-Does one know when ffdshow will be able to play embedded WavPack Audio together with the WavPack Splitter?
-Also does anyone know if ffdshow will ever support quicktime audio such as TWOS,ARAW,etc?
-There seems to be something wrong when trying to play files with embedded SNOW Video (such as the "SoC_[SNOW+Vorbis].mkv" files from Liisachan). There's no video at all (and with this I mean also no ffdshow video decoder in the dsgraph).
-When trying to enhance the video images with warpsharp an ugly green line always shows up (see image below). Is this supposed to happen?
http://img139.imagevenue.com/loc361/th_34576_ffdshow_warpsharp_123_361lo.JPG (http://img139.imagevenue.com/img.php?image=34576_ffdshow_warpsharp_123_361lo.JPG)

[akapuma]

Hello,

I've a problem to play this file: http://akapuma.info/test/Sac-001.mkv (266kB)
This file contains x264 and vorbis in a mkv-container. Player: MPC, WMP6.4. Splitter: haali. WIN2000SP4I asked in the german forum (http://forum.gleitz.info/showthread.php?p=308047). Kurt reported, that's he need 10 seconds, before the movie starts (with 529 CLSID, no problems with 482). So, I repeated my test. MPC starts after 53 seconds.

Best regards

akapuma

[bob0r]

Compiling ffdshow rev 2546: gcc 4.0.3: worked

Compiling ffdshow-tryout rev 538: gcc 4.0.3:

gcc -c -mno-cygwin -mdll -fno-rtti -mthreads -pipe -D_WINGDI_ -DUCLIBCPP -D_GLIBCPP_HAVE_MBSTATE_T -D_WIN32_IE=0x0500 -mmmx -msse -mfpmath=sse -msse2 -w -DNDEBUG -UDEBUG -DFFDEBUG=0 -I. -I.. -Iuclibc++ -Ibaseclasses -I../baseclasses -IimgFilters -I../imgFilters -Implayer -I../mplayer -Isettings -I../settings -Isettings/filters -I../settings/filters -Icodecs -I../codecs -Isubtitles -I../subtitles -Iconvert -I../convert -Idialog -I../dialog -IaudioFilters -I../audioFilters -Icygwin -I../cygwin -Iffmpeg -I../ffmpeg -Iacm -I../acm -Ixiph -I../xiph -Ifilters -I../filters -Imuxers -I../muxers -I"/dx/Include" -L"/dx/MingLib" -ldx9 -I/dx/Include -L/dx/MingLib -ldx9 -O2 -march=pentium-mmx -mtune=i686 -fomit-frame-pointer -finline-functions -finline -frename-registers -fweb -funit-at-a-time -MMD -o ffdshow_all.o ffdshow_all.cpp
In file included from TcpuUsage.cpp:66,
from ffdshow_all.cpp:7:
TcpuUsage.h:3:17: error: pdh.h: No such file or directory


Instead of getting pdh.h from installing Microsoft stuff i dont want to install yet, does anyone mind putting pdh.h online, and possible more files needed?

[clsid]

-There seems to be something wrong when trying to play files with embedded SNOW Video (such as the "SoC_[SNOW+Vorbis].mkv" files from Liisachan). There's no video at all (and with this I mean also no ffdshow video decoder in the dsgraph).SNOW support has been disabled some time ago, since it was extremely buggy.

[clsid]

Instead of getting pdh.h from installing Microsoft stuff i dont want to install yet, does anyone mind putting pdh.h online, and possible more files needed?It is part of the platform SDK. I have uploaded it here (http://www.mytempdir.com/1060418) for you.

[dk75]

Hello,

I've a problem to play this file: http://akapuma.info/test/Sac-001.mkv (266kB)
This file contains x264 and vorbis in a mkv-container. Player: MPC, WMP6.4. Splitter: haali. WIN2000SP4

Problems with newer ffdshow (player hangs, abort via taskmanager required):
533 by Dirk Paehl
529 by CLSID

No problems with older ffdshow:
514 by Dirk Paehl
494 by CLSID

Best regards

akapuma

WinXP SP1
Haali 20061029 aka 1.6.336.23
ffdshow rev 529 clsid generic
ZoomPlayer Standard 5.00 b8
Plays fine except ~10s laaaag at the start

[Eragon4ever]

I think someone should take a look at this (http://forum.doom9.org/showthread.php?t=118111).

[bob0r]

@clsid: thanks
So what is the deal with SSE and SSE2?
Which version should i use (put online)

I have compiled:
ffdshow-538-gcc4.0.3-sse2-x264.nl.exe
http://mirror05.x264.nl/ffdshow/force.php?file=./ffdshow-538-gcc4.0.3-sse2-x264.nl.exe

ffdshow-538-gcc4.0.3-sse-x264.nl.exe
http://mirror05.x264.nl/ffdshow/force.php?file=./ffdshow-538-gcc4.0.3-sse-x264.nl.exe

[bob0r]

538 bug report:
BBC.HD.Robin.Hood.S01E03.H.264.1080MBAFF.AC3.5.1-BBC.sample.ts (http://mirror05.x264.nl/public/force.php?file=./BBC.HD.Robin.Hood.S01E03.H.264.1080MBAFF.AC3.5.1-BBC.sample.ts) is not being displayed 16:9, splitters use: MPC internal and Haali's

bug does not exist with:
ffdshow_rev497_20061104_clsid.exe

[kurt]

So what is the deal with SSE and SSE2?
Which version should i use (put online)
I would wait 'till the delay bug is fixed ...

[LoRd_MuldeR]

I would wait 'till the delay bug is fixed ...

Yeah, that makes it nearly unuseable. Better build rev514 for the moment...

[Inventive Software]

What codecs does the massive delay occur with clsid's 529 build? I have that build, and so far, no problems with DivX......

[LoRd_MuldeR]

What codecs does the massive delay occur with clsid's 529 build? I have that build, and so far, no problems with DivX......

I replaced it with last working build. But the delay occured with my h.264 videos (MPC + internal AVI-Splitter + ffdshow + VMR9)

[kurt]

here I get problems with xvid and x264, in avi, mkv, mp4 whatever :)

List of known issues in revision 538:
1) Long delay when opening files. Seems to be caused by the CPU load monitoring code added in rev524.



rev524: Alternative implementation of "CPU Load".
Now it can work on multiprocessor system.
It uses pdh.dll in your HDD in NT, registry hack in Win9x.
Automatic quality control of postprocessing works again.
If "CPU Load" for the heaviest processor >90% and "Queued samples"<8 ...

so it is only multiprocessor related? (I'm sitting behind a coreduo ATM)

[LoRd_MuldeR]

here I get problems with xvid and x264, in avi, mkv, mp4 whatever :)




rev524: Alternative implementation of "CPU Load".
Now it can work on multiprocessor system.
It uses pdh.dll in your HDD in NT, registry hack in Win9x.
Automatic quality control of postprocessing works again.
If "CPU Load" for the heaviest processor >90% and "Queued samples"<8 ...

so it is only multiprocessor related? (I'm sitting behind a coreduo ATM)

Nope. Got the problem on my good ol' AthlonXP (WinXP SP-2)

[Inventive Software]

But the delay occured with my h.264 videos (MPC + internal AVI-Splitter + ffdshow + VMR9)

You should know by now H.264 and AVI don't mix.

I have tested said build on MP4 files with AVC content, and no delay's occured.

[LoRd_MuldeR]

You should know by now H.264 and AVI don't mix.

I have tested said build on MP4 files with AVC content, and no delay's occured.

It always worked before since ages and it works 100% fine with rev514
And please don't start that useless AVI flamewar again now ;)

[Inventive Software]

I'm not starting the AVI flamewar, just using what works. ;)

I still think CPU detection code in the installer would prevent this issue, cause it seems to be related to certain versions of pdh.dll. Is that what you've generally found?

[ilpippo80]

I have the same delay with all the files I've tried it with: avi files, mpeg files, flv files and also using ffdshow for postprocessing only.
I doubt that it is something related to some specific format.

[bob0r]

@kurt

Yup thanks, it was more a general question.
I won't put any build online until i know its working properly.

I just need to know if SSE or SSE2 makes any difference at all.

[fastplayer]

I just need to know if SSE or SSE2 makes any difference at all.
IMHO the SSE2 build is not worth the effort.

[bob0r]

I have noticed ffdshow_rev497_20061104_clsid.exe has SSE2 .dll files (thats what you get when you just run "make")

So i guess those will do, but i am not sure if SSE2 ffdshow.ax still crashes on non SSE2 machines...

[fastplayer]

Oops, I was talking about the ffmpeg/libavcodec part which is hand-optimized for SSE1/2/3DNow etc. Compiling it with SSE1/2 switches won't have any effects on performance. In some cases it even caused instability....

[clsid]

I have noticed ffdshow_rev497_20061104_clsid.exe has SSE2 .dll files (thats what you get when you just run "make")

So i guess those will do, but i am not sure if SSE2 ffdshow.ax still crashes on non SSE2 machines...
Building libavcodec and libmplayer can be done by just running make. The binary will run fin on CPUs without SSE/SSE2. It will contain both generic and optimized code paths.

Building ffdshow.ax with GCC will generate a binary that is SSE or SSE2 dependent.

[clsid]

I still think CPU detection code in the installer would prevent this issue, cause it seems to be related to certain versions of pdh.dll. Is that what you've generally found?So far all the XP users have the same version. So I don't think this is the cause.

[Episode]

I can reproduce this bug with finnish version of xp, but not with english version.

[fastplayer]

So far all the XP users have the same version. So I don't think this is the cause.
Maybe Haruhiko - who implemented revision 524 - has a newer pdh.dll:
http://support.microsoft.com/kb/917022/en-us

[Px]

Delay also occurs with ffdshow_rev494_20061103_clsid.exe on my XP x64, but not on the first player launch, only when I opening files with already launched player. No dependence with type of launched file, no cpu usage, some weird file operations (bsplayer trying to find xvidcore.dll, avisynth.dll, skl_drv_mpg.dll (???), audxlib.dll), no strange registry keys usage. Later I try to compare registry and file logs in cases of first and second launch.....

[F_L_C]

Just installed ffdshow_rev529_20061113_clsid and have no delay opening xvid/divx avis.

Athlon XP 2400
NV 7800GS (FW 92.91)
WinXP SP2 x86 (fully patched, DX9c Oct. build)
ZP 4.51 (vmr9 renderless)
default ffdshow xvid/divx filters w/ postprocessing, denoise3D, and resize/aspect enabled

[_xxl]

Compiling ffdshow rev 2546: gcc 4.0.3: worked

Compiling ffdshow-tryout rev 538: gcc 4.0.3:

gcc -c -mno-cygwin -mdll -fno-rtti -mthreads -pipe -D_WINGDI_ -DUCLIBCPP -D_GLIBCPP_HAVE_MBSTATE_T -D_WIN32_IE=0x0500 -mmmx -msse -mfpmath=sse -msse2 -w -DNDEBUG -UDEBUG -DFFDEBUG=0 -I. -I.. -Iuclibc++ -Ibaseclasses -I../baseclasses -IimgFilters -I../imgFilters -Implayer -I../mplayer -Isettings -I../settings -Isettings/filters -I../settings/filters -Icodecs -I../codecs -Isubtitles -I../subtitles -Iconvert -I../convert -Idialog -I../dialog -IaudioFilters -I../audioFilters -Icygwin -I../cygwin -Iffmpeg -I../ffmpeg -Iacm -I../acm -Ixiph -I../xiph -Ifilters -I../filters -Imuxers -I../muxers -I"/dx/Include" -L"/dx/MingLib" -ldx9 -I/dx/Include -L/dx/MingLib -ldx9 -O2 -march=pentium-mmx -mtune=i686 -fomit-frame-pointer -finline-functions -finline -frename-registers -fweb -funit-at-a-time -MMD -o ffdshow_all.o ffdshow_all.cpp
In file included from TcpuUsage.cpp:66,
from ffdshow_all.cpp:7:
TcpuUsage.h:3:17: error: pdh.h: No such file or directory


Instead of getting pdh.h from installing Microsoft stuff i dont want to install yet, does anyone mind putting pdh.h online, and possible more files needed?

You need pdh.h and libpdh.a.

[_xxl]

List of known issues in revision 538:

6) ffdshow doesn't work with CrystalPlayer 1.97

[_xxl]

IMHO the SSE2 build is not worth the effort.
Why?Just search "__SSE2__" in ffdshow.

[haruhiko_yamagata]

I still cannot reproduce the delay problem.

To summarize,

The version of pdh.dll does not matter.
When it is delayed, "CPU Load" always shows the error "does not work in vista..." (Is this correct?)

I think the problem is caused by the failure of loading pdh.dll.
If ffdshow try to load windows\system32\pdh.dll and fail, then it serches the environmental variable "PATH". It takes the time.
I'll try explicit full path. But why does loading pdh.dll fail?

[haruhiko_yamagata]

You need pdh.h and libpdh.a.
libpdh.a is not necessary. I think I can remove dependency to pdh.h.

[baka]

6) ffdshow doesn't work with CrystalPlayer 1.97

i second that :devil:

[Peuj]

I think the problem is caused by the failure of loading pdh.dll.
If ffdshow try to load windows\system32\pdh.dll and fail, then it serches the environmental variable "PATH". It takes the time.
I'll try explicit full path. But why does loading pdh.dll fail?

Maybe people who have the problem could use ProcessExplorer (http://www.microsoft.com/technet/sysinternals/ProcessesAndThreads/ProcessExplorer.mspx)
(freeware tool, nothing to install simply execute the exe)
Menu "Find" -> "Find Handle or DLL..."

To confirm the dll is loaded or not during playback.

[Peuj]

List of known issues in revision 538:

Maybe you could had this to the list http://forum.doom9.org/showthread.php?t=118111

Thanks

[_xxl]

6)CrystalPlayer 1.97 doesn't work with ffdshow!
CrystalPlayer is protected by Aspack 2.12b, so I CAN'T enable debug.I'll blacklist Crystal.exe.

[_xxl]

Maybe you could had this to the list http://forum.doom9.org/showthread.php?t=118111
Thanks
Yes, libdts seems to be broken for 44100 Hz and 1024 samples.
EDIT:
Same problem with Gabest's MpaDecFilter Filter.

[baka]

6)CrystalPlayer 1.97 doesn't work with ffdshow!
CrystalPlayer is protected by Aspack 2.12b, so I CAN'T enable debug.I'll blacklist Crystal.exe.

ye, always the best solution :(

why not to contact crystal player team? :P

[G_M_C]

Turn on postprocessing......

Yeah, but it is still pixellated on the edges, XviD shows well-defined edges. When the decoding of H264 by ffdshow produces ruff edges, i cant use it for my transcodes ...

[KoD]

CrystalPlayer also doesn't load VSfilter so it can't display subtitles. It has its own way of building filter graphs and forcing things that is not compatible with some filters. Really something that needs to be corrected by the CrystalPlayer devs and not anyone else.

G_M_C : on the ffdshow video decoder config -> Decoder options -> switch IDCT to auto and see if you still have that problem. You could also play around with Error concealment option and Error resilience.

[_xxl]

Maybe people who have the problem could use ProcessExplorer (http://www.microsoft.com/technet/sysinternals/ProcessesAndThreads/ProcessExplorer.mspx)
(freeware tool, nothing to install simply execute the exe)
Menu "Find" -> "Find Handle or DLL..."
To confirm the dll is loaded or not during playback.
http://i9.tinypic.com/43cqalh.jpg

[baka]

CrystalPlayer also doesn't load VSfilter so it can't display subtitles. It has its own way of building filter graphs and forcing things that is not compatible with some filters. Really something that needs to be corrected by the CrystalPlayer devs and not anyone else.

so can we do something in crystal player to use ffdshow normally? some option or whatever?

[KoD]

Baka, short answer: no. Long answer: CrystalPlayer's code restricts the way a rendering graph is built so nothing can be done. Take it with its developers.

[_xxl]

so can we do something in crystal player to use ffdshow normally? some option or whatever?
No.It doesn't work with ffdshow since rev8.Other media players are working so the problem is in CrystalPlayer.

[netwolf]

Delay also occurs with ffdshow_rev494_20061103_clsid.exe on my XP x64, but not on the first player launch, only when I opening files with already launched player. No dependence with type of launched file, no cpu usage, some weird file operations (bsplayer trying to find xvidcore.dll, avisynth.dll, skl_drv_mpg.dll (???), audxlib.dll), no strange registry keys usage. Later I try to compare registry and file logs in cases of first and second launch.....
I *sometimes* have the same problem, or at least a very similar one: when playing a file and opening another one, the second file does not play properly, but I only see the picture for a second (or less) and the player stops.
Restarting the file gives the same result.

When I close the player (ZoomPlayer) and open the same file again, it plays without problems.

WRT the delay problem: here it doesn't seem to be busy finding a system file. There is no HDD activity during the delay, it seems to be just sitting and waiting...

[fastplayer]

Why?Just search "__SSE2__" in ffdshow.
I was referring to the ffmpeg/libavcodec part. Benchmarks showed no improvements when using various compiler switches. In fact videomixer's generic MMX build turned out to be faster than any of those "optimized" builds.
As for ffdshow itself, I think you can answer the question better than me :)

[Inventive Software]

In 543 it looks like a GCC bug. Take it up with GCC developers, cause the fatal error in that was an "internal compiler error".

542: Coding issues. Something for the team to look at I think.

@bob0r: Have you tried compiling with GCC 3.4.6?

[_xxl]

It works for me.Tested with mingw GCC 4.0.3.
http://img149.imagevenue.com/loc502/th_89905_Capture_122_502lo.jpg (http://img149.imagevenue.com/img.php?image=89905_Capture_122_502lo.jpg)