View Full Version : ffdshow and ffdshow_tryout Builds Listing
Pages :
[
1]
2
3
4
5
6
7
8
Inventive Software
9th January 2006, 15:18
Right. Since this thread (http://forum.doom9.org/showthread.php?t=98600) has more than 2000 posts to filter through, I figure we need a new thread that just has one post for each builder, and each builder can update this as necessary. If the builder could state what MMX and SSE support there is, and what compiler(s) has/have been used this will clarify for newbies. If you have a problem with a particular build, post here or in the other thread and we'll get back to you.
If you have a general ffdshow problem that you believe is not the cause of a specifically compiled build, start another thread and don't pollute this one please!
Inventive Software
Slave01
13th January 2006, 13:17
I support the request...It seems that aziendeassociate....link is stopped to 1103 build....the other builds are slower on my sempron 3000+...What is the best build organized by processor...
Thanks
John Slave
Inventive Software
13th January 2006, 13:51
Sempron 3000+, that speaks SSE2 to me. You have lots of choices there!
Milan's "official" daily build, not very recent, but supported. (http://cutka.szm.sk/files/ffdshow-20051129.exe) Dated 2005/11/29
Bob0r's build, (http://files.x264.nl/ffdshow/ffdshow-20051208-gcc4.0.2-sse2-x264.nl.exe) dated 2005/12/08.
Drevil_xxl's build, (http://www.myfilehut.com/userfiles/5120/CollectedBuilds/ffdshow-20051221.zip), dated 2005/12/21.
Lots of choices, take your pick.
Slave01
14th January 2006, 12:11
Sempron 3000+ xp core...only sse...not 64-bit :(. By the way thanks
breez
14th January 2006, 14:28
You can use Milan's and Drevil_xxl's builds, they don't require SSE. I'm using the latter with my Athlon XP and it is a good build (for me it is faster than any GCC compiled build).
Negi
16th January 2006, 06:40
For a "comprehensive" thread, you didn't include many versions of FFDSHOW.
What about SSE2/SSE/X64 build lists? Also, why hasn't there been a new build in a month? Also, put your updates in the first post. That way it's much more organized.
ExtraEye
16th January 2006, 13:20
but i don't think making new build has anything to do with nice Inventive Software person who's trying to help you choose the right builds.
Inventive Software
16th January 2006, 14:56
@Negi: I do the best I can. For the record this was a thread for the builders to post regular updates for their builds, not for me to do a catalogue of every build there has been, as that would take several hours, something I don't have at the moment. In future, I may, but don't hold your breath.
@Slave01: I apologise for getting the specs for your CPU wrong. I was going off the top of my head, and I got it wrong. :(
ICL9 and GCC combined builds are quite fast.
clsid
16th January 2006, 16:55
Build: 2005-12-22 (latest CVS)
Compiler: ICL9 (most files) + GCC 3.4.4 (for libavcodec.dll and mplayer.dll)
Minimum CPU requirement: MMX
Optimized for (used only if supported by CPU): MMX/MMXext/SSE/SSE2/SSE3
Download: click here (http://www.badongo.com/file.php?file=ffdshow+2005-12-22+ICL9-GCC402__2006-01-16_ffdshow-20051222.exe)
Inventive Software
19th January 2006, 13:34
@clsid: I tried that build, and it doesn't open the VFW configuration dialog. The other dialog boxes open fine, just that one.
It says: "ff_vfw.dll requires another module to load".
Thing is, I have an earlier build that works without hassles. Any suggestions?
clsid
19th January 2006, 17:12
It works here.
Dependency Walker (http://www.dependencywalker.com/) might be able to tell you what module is missing.
Inventive Software
22nd January 2006, 17:01
Thanks. I ran "Dependency Walker". It told me I needed "msvcr71.dll". I have about 5 copies of that scattered around my hard disk everywhere except "windows\system".
lithoc
23rd January 2006, 02:13
Do anybody build ffdshow for old machines (eg Pentium-MMX, Pentium 2, older Celeron)?
I still have Pentium-MMX(overclocked) to watch movies sometimes(for sake of torturing it :devil: )
It plays perfectly on Linux + mplayer via VESA 2.0 output.
DigitalDivide
23rd January 2006, 02:25
Edited and deleted:
clsid
24th January 2006, 00:05
Build: 2006-01-23
Compiler: ICL9 (most files) + GCC 3.4.5 (for libavcodec.dll and mplayer.dll)
Minimum CPU requirement: MMX
Optimized for (used only if supported by CPU): MMX/MMXext/SSE/SSE2/SSE3
Download: click here (http://www.badongo.com/file.php?file=ffdshow+CVS+build+2006-01-23+ICL9-GCC3.4.5__2006-01-23_ffdshow-20060123.exe)
bob0r
25th January 2006, 00:07
Do anybody build ffdshow for old machines (eg Pentium-MMX, Pentium 2, older Celeron)?
I still have Pentium-MMX(overclocked) to watch movies sometimes(for sake of torturing it :devil: )
It plays perfectly on Linux + mplayer via VESA 2.0 output.
Try clsid's build, it should work.
Mirrored on http://files.x264.nl/ffdshow/ffdshow-20060123_clsid.exe
GCC version out once it can be compiled :o
LoRd_MuldeR
28th January 2006, 15:32
Try clsid's build, it should work.
Mirrored on http://files.x264.nl/ffdshow/ffdshow-20060123_clsid.exe
Some filters are broken in this build (e.g. Resize-Sharpen)
clsid
28th January 2006, 18:56
For old CPUs I recommend Milan's own builds (ICL8.1). Those are more stable than ICL9 builds. Hopefully he will put a new one online sometime soon, since his most current one dates from november 29th.
LoRd_MuldeR
28th January 2006, 19:15
Try clsid's build, it should work.
Mirrored on http://files.x264.nl/ffdshow/ffdshow-20060123_clsid.exe
GCC version out once it can be compiled :o
Can you make a new SSE build like "ffdshow-20051221-gcc4.0.2-sse-x264.nl.exe" from the current version?
cc979
6th February 2006, 23:17
Can you make a new SSE build like "ffdshow-20051221-gcc4.0.2-sse-x264.nl.exe" from the current version?
i think current cvs, is broke for gcc
Inventive Software
7th February 2006, 12:21
If it makes more sense, use a stable version of GCC for compiling. Most often, the CVS version is not worth it when compiling, as they don't give such noticable speed increases with program running time.
cc979
7th February 2006, 13:09
could someone upload the last compile-able version, as i've lost it .. thanks
Inventive Software
7th February 2006, 13:35
GCC or ffdshow?
cc979
7th February 2006, 16:45
GCC or ffdshow?
ffdshow i you can, thanks
zambelli
20th February 2006, 09:50
Does anyone know why the x264.nl hosted build of ffdshow hasn't been updated since 12-21-05? What are they waiting for?
bob0r
20th February 2006, 11:35
Does anyone know why the x264.nl hosted build of ffdshow hasn't been updated since 12-21-05? What are they waiting for?
ffdshow fails to compile with gcc 4.0.2, Milan updated ffdshow, we reported the bug, but he never went active again.
http://sourceforge.net/tracker/?group_id=53761&atid=471489
> http://sourceforge.net/tracker/index.php?func=detail&aid=1413410&group_id=53761&atid=471489
( also http://sourceforge.net/tracker/index.php?func=detail&aid=1425654&group_id=53761&atid=471489 )
Trying gcc-4.1.0-20060219 now
ftp://ftp.nluug.nl/mirror/languages/gcc/prerelease-4.1.0-20060219/
Its not a snapshot, but a true prerelease it seems.
It has compiled on mingw.
ffdshow fails:
gcc-4.0.2
http://files.x264.nl/ffdshow/ffdshow_output_20_feb_2006_gcc-4.0.2.txt
gcc-4.1.0.20060219
http://files.x264.nl/ffdshow/ffdshow_output_20_feb_2006_gcc-4.1.0.20060219.txt
cc979
20th February 2006, 17:34
ffdshow fails to compile with gcc 4.0.2, Milan updated ffdshow, we reported the bug, but he never went active again.
http://sourceforge.net/tracker/?group_id=53761&atid=471489
> http://sourceforge.net/tracker/index.php?func=detail&aid=1413410&group_id=53761&atid=471489
( also http://sourceforge.net/tracker/index.php?func=detail&aid=1425654&group_id=53761&atid=471489 )
Trying gcc-4.1.0-20060219 now
ftp://ftp.nluug.nl/mirror/languages/gcc/prerelease-4.1.0-20060219/
Its not a snapshot, but a true prerelease it seems.
It has compiled on mingw.
ffdshow fails:
gcc-4.0.2
http://files.x264.nl/ffdshow/ffdshow_output_20_feb_2006_gcc-4.0.2.txt
gcc-4.1.0.20060219
http://files.x264.nl/ffdshow/ffdshow_output_20_feb_2006_gcc-4.1.0.20060219.txt
have tried to re-compile gcc-4.1 with gcc-4.1
clsid
24th February 2006, 21:38
Build: 2006-01-27 (newest CVS)
Compiler: ICL 9.0.029 (most files) + GCC 3.4.5 (libavcodec.dll, mplayer.dll)
Minimum CPU requirement: MMX
Download: click here (http://rapidshare.de/files/14051213/ffdshow-20060127.exe.html)
cc979
2nd March 2006, 04:35
Build: 2006-01-27 (newest CVS)
Compiler: ICL 9.0.029 (most files) + GCC 3.4.5 (libavcodec.dll, mplayer.dll)
Minimum CPU requirement: MMX
Download: click here (http://rapidshare.de/files/14051213/ffdshow-20060127.exe.html)
what dx sdk do you use ?
zambelli
2nd March 2006, 07:28
It'd be nice if somebody ran performance and functionality tests on these new builds as they become available. It wouldn't need to be anything elaborate, just a few basic functionality tests and some perf numbers on basic decoding and such. I feel like there are too many competing builds of ffdshow popping up and we have absolutely no bar and no way of measuring them up.
seehowyouare
2nd March 2006, 10:53
I feel like there are too many competing builds of ffdshow popping up and we have absolutely no bar and no way of measuring them up.
IDEA - If all the builds were distributed via Bit Torrent (the main audience anyway) it would
1) be "one" place to get your latest codec
2) natural attrition might just kill off "bad" builds
3) NO more searching forums and threads and ..
4) Reduces storage/bandwidth needs for builders.
Lots of disadvantages. For example, the super-beta-mega-quality-best-ever-XviD-codec.exe file will always be in circulation because of the filename, even if it is a "bad" build.
clsid
2nd March 2006, 18:11
what dx sdk do you use ?
The one included in VS.net. I also have the oct2005 one, but that should only be used in case something isn't found in the standard VS libs.
IDEA - If all the builds were distributed via Bit Torrent (the main audience anyway)
I thought they were distributed via Bittorrent already, and several different builds in one package.
zambelli
2nd March 2006, 20:45
several different builds in one package.
I don't think that's helpful at all; in fact, I think that's confusing. Once again, I think there should be some sort of performance baseline for these builds. Just compiling Ffdshow and advertising its more recent build date shouldn't be any guarantee of a build's quality or performance value.
I don't think that's helpful at all; in fact, I think that's confusing. Once again, I think there should be some sort of performance baseline for these builds. Just compiling Ffdshow and advertising its more recent build date shouldn't be any guarantee of a build's quality or performance value.
Of course. Trouble is that for performance issues, proper test should be done on those segments on CPU kingdom :p
Celeron and P3 (SSE1 only)
Athlon XP (SSE1 only)
P4 (SSE2)
Sempron and Athlon 64 (SSE2)
Athlons prior XP (no SSE at all).
It's safe to assume that for each group above the the ultimate proper build is different.
GCC believed to produce better results for AMD, ICL for Intel. Then it's the question of additional CPU instruction sets. Many builds won't work on casual Athlons at all. Also for AMD it should be separate builds for SSE1 and SSE2 capable processors. Don't forget there're different versions of gcc and ICL as well.
I tell all that to show that hardly anyone would volunteer to do proper testing. And w/o proper testing it all remains speculative and dependent on specific systems, as it is now.
clsid
3rd March 2006, 11:59
ICL can be patched so that it works good on AMD processors as well.
GCC compilations need at least SSE. ICL and MSVC compilations also work without SSE.
Most cpu hungry routines in ffdshow/libavcodec are already hand optimized in assembler code. So making a build that is optimized for a specific CPU instruction set won't give huge performance boosts.
cc979
3rd March 2006, 13:00
The one included in VS.net. I also have the oct2005 one, but that should only be used in case something isn't found in the standard VS libs.
cheers, i've been using dec2005 but had a look at feb2006 some interesting dx10 things in it
Inventive Software
3rd March 2006, 13:37
If people want to run tests and post results, may I suggest a new thread? This thread's primarily for ffdshow builders to report new builds. I would run some tests, but my system isn't great, and I have a lack of varied material.
ExtraEye
3rd March 2006, 16:28
how do i run tests? i have an athlon 64 3000+ if you need me to test it
zilexa
8th March 2006, 21:45
Why not just use the newest version on the ffdshow page?
Just search Google: ffdshow wiki
and you will find the page!
Getting FFDShow (http://ffdshow.sourceforge.net/tikiwiki/tiki-index.php?page=Getting+ffdshow)
Liisachan
9th March 2006, 00:56
milan's builds are not the newest, but actually I myself am using 2005-11-15 not -24 nor -29. Life would be too easy if the newest was always the best, but if you like to test the newest versions, here (http://ffdshow.faireal.net/)'s a random pack. It's torrented because bw is expensive and I'm not rich, only have 500GB/mo while need 2000.
Yong
9th March 2006, 11:09
Can someone pls tell me where to download "mingw compatible" GCC 4.xx binary? i wan to try to compile ffdshow too. ;)
i tried this one
http://oss.netfarm.it/mplayer/misc/gcc-mingw32-4.2.0.svn.20060122.tar.gz
but it seem lack some libs...
cc979
9th March 2006, 22:46
Can someone pls tell me where to download "mingw compatible" GCC 4.xx binary? i wan to try to compile ffdshow too. ;)
i tried this one
http://oss.netfarm.it/mplayer/misc/gcc-mingw32-4.2.0.svn.20060122.tar.gz
but it seem lack some libs...
for ffdshow use cvs, tortoisecvs is pretty easy use too all info is in https://sourceforge.net/
Liisachan
13th March 2006, 10:35
By the way, I'd really appreciate it if you would compile ff_tremor in high-accuracy mode, like celtic_druid did.
tremor is not good in default compile, I can easily abx at 15/15, meaning Ogg Vorbis decoded thru ffdshow/tremor is damaged in quality. (That's why I can't respect any codec packs that use ffdshow for Vorbis decoding. They don't have ears -_- )
Egh
13th March 2006, 21:23
By the way, I'd really appreciate it if you would compile ff_tremor in high-accuracy mode, like celtic_druid did.
tremor is not good in default compile, I can easily abx at 15/15, meaning Ogg Vorbis decoded thru ffdshow/tremor is damaged in quality. (That's why I can't respect any codec packs that use ffdshow for Vorbis decoding. They don't have ears -_- )
There's also an option to decode vorbis thru libavc, but in last (january) cvs it's seems a bit broken (most likely within libavc itself).
clsid
3rd April 2006, 14:57
SVN: 2006-04-03
Compilers used:
- MSVC71 (ffdshow.ax, etc)
- GCC 3.4.5 (libavcodec.dll, mplayer.dll, kerneldeint.dll)
Minimum CPU requirement: MMX
Download: click here (http://rapidshare.de/files/17103116/ffdshow-svn20060403.exe.html)
Liisachan
3rd April 2006, 15:44
Thanks very much clsid!!!
@Egh
libavcodec is ok, but tremor is the default for vorbis, isnt it? So you'd have to manually change the setting. Ordinary ffdshow users don't know that. I'd like ppl to enjoy the real quality of Ogg Vorbis.
LoRd_MuldeR
3rd April 2006, 20:59
SVN: 2006-04-03
Compilers used:
- MSVC71 (ffdshow.ax, etc)
- GCC 3.4.5 (libavcodec.dll, mplayer.dll, kerneldeint.dll)
Minimum CPU requirement: MMX
Download: click here (http://rapidshare.de/files/17103116/ffdshow-svn20060403.exe.html)
What are the changes ???
videomixer9
3rd April 2006, 21:32
most important is that it has updated parts in libavcodec and updated libmpeg2. The h264 code was also updated but there seems to be no major speed improvement, and actually it seems libavcodec goes a different direction there, actually even VLC got recently more problems with 1080 content on my box, i'm still waiting for them to beat coreavcs performance.
LoRd_MuldeR
3rd April 2006, 21:49
most important is that it has updated parts in libavcodec and updated libmpeg2. The h264 code was also updated but there seems to be no major speed improvement, and actually it seems libavcodec goes a different direction there, actually even VLC got recently more problems with 1080 content on my box
Thanks for info :)
i'm still waiting for them to beat coreavcs performance.
Good luck :D
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.