ffdshow tryouts project: Discussion & Development

[haruhiko_yamagata]

Originally Posted by ExtraEye

"It is greatly appreciated if anyone write a single page for it. If one don't have skill in html, please post a simple text."
Post a simple text where?

https://sourceforge.net/tracker/?gro...41&atid=867362

[haruhiko_yamagata]

Originally Posted by check

Just looking at your first post:
ffdshow raw filter doesn't work in ZoomPlayer. (reported by midiboy)
This is a ZP 'feature', called smartplay. It seems to remove any raw processing filters, and also includes a blacklist where you can add evil filters (morgan, antifreeze, et al). Disabling it lets everything work as normal.

re: CCCP doing bug reports and helping code. No can do. We test our own version, which means bug reports would simply become more work for us, and we are mostly a bunch of patch monkeys, not Real Coders, so we can't help with dev .

re: documentation. I'd be happy to write some of the ffdshow documentation, where does it belong? I've written a fair amount of the megui documentation, so I know what I'm getting in to.

You may report bugs as long as it is reproducible in the builds from tryouts. We'll fix them if we can.

Writing a documentation for ffdshow is wellcome, but you know I'm not feeling comfortable with CCCP. Before I can act cozy with you,

Open the source code immediately (Am I the copyright holder for the patches?).
Remove the insulting words in your WEB/forum (I know it's not you who wrote that).

[_xxl]

Originally Posted by Yong

can anyone comfirm this?
ffdshow.ax crash if noise filter+SSE2 is enabled.
im using P4 2.4Ghz.
tested with clsdi build, no crash.
i was compiled ffdshow with SSE=1 SSE2=1 flags...

MinGW GCC 4.x.x can't compile SSE2 builds.
That's why you couldn't find them anymore.

[ExtraEye]

Mentar, your "example" with the resizing is kind of silly. I mean, the fact that the user can change the way the video is processed in accordance to his tastes/pc's abilities is definitly not a bad thing. "trying to use software in ways they are not INTENDED to be used" - what?
It's fine to argue here and there and it's fine you want to support cccp for not using ffdshow tryouts but is it possible for you to stop talking so rudely? It's totally unnecessary and doesn't help this discussion. Actually this hole debate *flame* doesn't suit this thread at all - may as well continue them in cccp forums instead of wasting the development thread's pages on it.

[check]

Originally Posted by haruhiko_yamagata

Writing a documentation for ffdshow is wellcome, but you know I'm not feeling comfortable with CCCP. Before I can act cozy with you,

You don't need to act cozy with me to tell me a good place to write up documentation edit: and in any case, I'm not doing this work on behalf of the CCCP. In any case, I've started the thing over on the mewiki (which is turning into a more general AV wiki by the day) here: http://mewiki.project357.com/wiki/Ffdshow_reference
I won't have more time to work on it until sunday, so feel free to add/copy/etc as you see fit. GPL fun: all work posted there is licenced under the GFDL, but I dual licence my first two revisions and release them into the public domain. If you want to start the reference somewhere else, that should be all you need.

[ilpippo80]

@haruhiko
I've tested hw deint in clsid's ffdshow_rev926_20070215_clsid_sse_icl9.exe.
Here's my results using graphedit:

MPV decoder only: ok with VMR7 and VMR9
DScaler only: ok with VMR7 and VMR9
ffdshow(libmpeg2) only:
- with VMR7 hw deint is always off and the image is horizontally stretched (I've never seen this bug before)
- with VMR9 hw deint is always on
MPV+ffdshow(raw): ok with VMR7 and VMR9 (hw deint depends only on MPV setting)
DScaler+ffdshow(raw):
- with VMR7 hw deint depends only on DScaler settings
- with VMR9 hw deint is always off
ffdshow(libmpeg2)+ffdshow(raw): hw deint is always off

[Yong]

Originally Posted by drevil_xxl

MinGW GCC 4.x.x can't compile SSE2 builds.
That's why you couldn't find them anymore.

ok thx for the info
btw discovered 2 new bugs:
1. resize filter, eg, if i wana resize my video from 320x240 to 640x480, picture is resized, but it only show 1/4 on the screen.
same as 640x480 resize to 320x240.
2. resize filter crash if using rgb32 out.

just a quick report.

[HeadBangeR77]

@ clsid:
Your last 926 icl9 sse build works like charm, thanks.

@ haruhiko_yamagata
I cannot reproduce the old bug with auto iDCT & XviD decoding with the above build. Auto works fine for me now.

Btw. Is XviD MMX DCT really always used while encoding?

PS. * Using words of French origin seems to be complicated.

[haruhiko_yamagata]

Originally Posted by ilpippo80

@haruhiko
I've tested hw deint in clsid's ffdshow_rev926_20070215_clsid_sse_icl9.exe.
Here's my results using graphedit:

MPV decoder only: ok with VMR7 and VMR9
DScaler only: ok with VMR7 and VMR9
ffdshow(libmpeg2) only:
- with VMR7 hw deint is always off and the image is horizontally stretched (I've never seen this bug before)
- with VMR9 hw deint is always on
MPV+ffdshow(raw): ok with VMR7 and VMR9 (hw deint depends only on MPV setting)
DScaler+ffdshow(raw):
- with VMR7 hw deint depends only on DScaler settings
- with VMR9 hw deint is always off
ffdshow(libmpeg2)+ffdshow(raw): hw deint is always off

Oops! True. Please wait.

[ExtraEye]

drevil_xxl: Exactly the point. Nothing is coming out of this - he even made a negative post in which he asks to be ignored and that's probably the best idea too. Let's just move on.

[foxyshadis]

All CCCP discussion moved to http://forum.doom9.org/showthread.php?t=122341

Please keep it there, this thread is discussion for the development of tryouts only.

[cc979]

just be using the resizer in ffdshow rev926, i got it set to 720/576 bicubic and resize allways and 0% border set

thing is - viewing the video on my desktop with normal 100% zoom using overlay/vm7 renderless/vmr9 renderless/haali- normal viewspace, but when in fullscreen there is about an inch of black border on the sides - is that normal ?

[haruhiko_yamagata]

Originally Posted by haruhiko_yamagata

Oops! True. Please wait.

Oops again, I replied too quickly. I tested with wrong settings.
All the case you pointed works OK for me.

[ilpippo80]

That's bad...
Which wrong settings? Maybe I had something wrong too.

[haruhiko_yamagata]

Resize was checked.

[clsid]

Originally Posted by cc979

just be using the resizer in ffdshow rev926, i got it set to 720/576 bicubic and resize allways and 0% border set

thing is - viewing the video on my desktop with normal 100% zoom using overlay/vm7 renderless/vmr9 renderless/haali- normal viewspace, but when in fullscreen there is about an inch of black border on the sides - is that normal ?

That can happen if you have "keep aspect ratio" enabled.

[foxyshadis]

Originally Posted by check

You don't need to act cozy with me to tell me a good place to write up documentation edit: and in any case, I'm not doing this work on behalf of the CCCP. In any case, I've started the thing over on the mewiki (which is turning into a more general AV wiki by the day) here: http://mewiki.project357.com/wiki/Ffdshow_reference
I won't have more time to work on it until sunday, so feel free to add/copy/etc as you see fit. GPL fun: all work posted there is licenced under the GFDL, but I dual licence my first two revisions and release them into the public domain. If you want to start the reference somewhere else, that should be all you need.

I think that's a great idea and a great place for it, since you already have a well-known wiki set up. Thanks!

[ilpippo80]

ok, my bad, I just rechecked, I didn't understand that I had to disconnect and reconnect ffdshow in the graph after modifying the settings! also the bug with ffdshow(libmpeg2) only with VMR7 went away!
Everything works as it should now, thanks haruhiko!

[ilpippo80]

ehm, well, the truth is that there's another small bug, I noticed it during my tests...
with gabest's mpeg2 splitter and vmr9, with ffdshow mpeg2 decoding (but also with dscaler, not with mpv decoder) after seeking the field order seems to be wrong and the hw deinterlaced video becomes "bumpy" (goes 2 frames forward and 1 backward).
I can't tell if it works well with ms mpeg2 splitter because it doesn't allow me to seek at all in graphedit.
I don't know if you can reproduce this and if you have any interest in fixing it...

[haruhiko_yamagata]

Originally Posted by ilpippo80

ehm, well, the truth is that there's another small bug, I noticed it during my tests...
with gabest's mpeg2 splitter and vmr9, with ffdshow mpeg2 decoding (but also with dscaler, not with mpv decoder) after seeking the field order seems to be wrong and the hw deinterlaced video becomes "bumpy" (goes 2 frames forward and 1 backward).
I can't tell if it works well with ms mpeg2 splitter because it doesn't allow me to seek at all in graphedit.
I don't know if you can reproduce this and if you have any interest in fixing it...

http://support.microsoft.com/kb/924523/en-us
I can't reproduce untill now.
Is this related? If it is not reproducible in VMR7, it's possible.
Which version of direct X do you use?