1080i native playback
#16
sorry for the noob question, with new FernetMenta's vdpau branch can a GTS450 take benefits from that? I've no issue with vdpau and 1080i temporal/spatial deinterlacing, just curious if I can boost, if it is possibile (Big Grin), the overall quality.
Reply
#17
Just tried to compile, but there seems to be a problem in VDPAU.cpp:

Code:
In file included from VDPAU.h:58:0,
                 from VDPAU.cpp:26:
/home/dirk/xbmc/fm-build/xbmc/xbmc/utils/ActorProtocol.h: In constructor ‘Actor::Protocol::Protocol(std::string, CEvent*, CEvent*)’:
/home/dirk/xbmc/fm-build/xbmc/xbmc/utils/ActorProtocol.h:86:19: warning: ‘Actor::Protocol::outDefered’ will be initialized after
/home/dirk/xbmc/fm-build/xbmc/xbmc/utils/ActorProtocol.h:78:15: warning:   ‘std::string Actor::Protocol::portName’
/home/dirk/xbmc/fm-build/xbmc/xbmc/utils/ActorProtocol.h:63:3: warning:   when initialized here
VDPAU.cpp: In member function ‘virtual long int VDPAU::CDecoder::Release()’:
VDPAU.cpp:244:1: warning: no return statement in function returning non-void
VDPAU.cpp: In member function ‘VDPAU::CVdpauRenderPicture* VDPAU::COutput::ProcessMixerPicture()’:
VDPAU.cpp:2780:17: warning: unused variable ‘vdp_st’
VDPAU.cpp: In member function ‘bool VDPAU::COutput::EnsureBufferPool()’:
VDPAU.cpp:2939:34: warning: comparison between signed and unsigned integer expressions
VDPAU.cpp: In member function ‘bool VDPAU::COutput::GLInit()’:
VDPAU.cpp:3175:3: warning: no return statement in function returning non-void
VDPAU.cpp: At global scope:
VDPAU.cpp:3176:3: error: expected unqualified-id before ‘return’
VDPAU.cpp:3177:1: error: expected declaration before ‘}’ token
Reply
#18
You need glew-dev >= 1.5.5. Not sure whether there is a package for for Ubuntu 10.04. I have compiled it myself for 10.10
http://glew.sourceforge.net/
Reply
#19
robo989 Wrote:I assume this question has now been answered so hope this isn't interfereing, if so please disregard my post Smile

Very curious on this de-interlacing, I have a GT430 which I thought was the ideal card for de-interlacing but it isn't capable of smooth playback using temporal\spatial full on satellite HD DVB-S\S2 sources in the UK of around 9-10mbit. I get roughly 40fps instead of 50fps.

You say the 520M can do this OK?

I specifically avoided getting a 520 model due to hearing numerous reports about it's appalling de-interlacing abilities, even though having the new VP engine and thus able to decode standard progressive material at an extremely high fps limit.

Does the 520M not have these de-interlacing shortcomings?

I'm very interested, as obviously I'm sure you're aware there is no low profile, single slot, fan-less GT430 available. But there is a 520 that has those featues, ideal for a HTPC, if the de-interlacing issues are resolved\not present in the 520M version?

Thanks, much appreciated.
That's strange. I also have a GT430 in my HTPC and it can deinterlace 1080i 40Mbit streams just fine.
Reply
#20
LB06 Wrote:That's strange. I also have a GT430 in my HTPC and it can deinterlace 1080i 40Mbit streams just fine.

I'm talking purely "temporal\spatial", not temporal or spatial. It can de-interlace fine in those modes, but not the combined one.

Are you saying yours can?

Ta
Reply
#21
FernetMenta Wrote:I have got a gt220 in my box in the living room and it does smooth playback on 1080i50 using temporal/spatial.
The problem is not your hardware, it has more power than my gt220 or a ion2. The problem is the single threaded de-interlacing (vdpau) in XBMC which limits performance. A gt220 for example needs 16-17ms only for de-interlacing of 1080i (sometimes more). There is not much room for delays in a 20ms window a frame needs to be displayed.
I am working on improvement for quire a while. You can try my vdpau branch:
https://github.com/FernetMenta/xbmc/tree/vdpau
or try the above mentioned openelec build.

Thanks for that, much appreciated insight.
Reply
#22
robo989 Wrote:I'm talking purely "temporal\spatial", not temporal or spatial. It can de-interlace fine in those modes, but not the combined one.

Are you saying yours can?

Ta
Well, yes. But not in combination with Lanczos3 optimized video scaling (which is unnecessary with 1080i anyway).
Reply
#23
LB06 Wrote:Well, yes. But not in combination with Lanczos3 optimized video scaling (which is unnecessary with 1080i anyway).

What is Lancosz3 opt. doing wrong with 1080i files?
What whould you suggest to use?
Reply
#24
You do not have to scale 1080i to a 1080 resolution. It is already the correct size after deinterlacing has been done.
First decide what functions / features you expect from a system. Then decide for the hardware. Don't waste your money on crap.
Reply
#25
fritsch Wrote:You do not have to scale 1080i to a 1080 resolution. It is already the correct size after deinterlacing has been done.

Sorry, i missed that it was the same resolution Blush
Reply
#26
(2012-02-25, 10:04)FernetMenta Wrote:
Quote:Is an interleaved passthrough mode possible?
No, it is not.

Is there a reason why it's not possible to output 1080i from XBMC?

I have configured the correct metamodes in xorg.conf but XBMC doesn't want to switch to 1080i even when deinterlacing method is set to "Off". I have a GT520 and it seems that no matter the content, the best deinterlacing method that works is "Temporal/Spatial (half)". I don't see anything wrong with the picture but if the TV can deinterlace, why not let it do its job?
Reply
#27
Strange, I've not experienced any problems with high-bitrate 1080i streams on my standard clocked passively cooled GT430. Using full spatial/temporal deinterlacing.
Reply
#28
I don't really know what's going on. If I choose the "wrong" deinterlacer, the video is basically smooth but plays faster once every second so it's obviously unwatchable. The framerate jumps between 35-45 FPS when playing 50i content (I assume it's trying to create 50 frames but can't keep up). Is there a way to check GPU usage on Linux while playing? I only have one screen so I can't launch nvidia-xconfig (which I think displays it), plus launching it locks my TV at 1080p50 (don't know why).
Reply
#29
sorry to overlook the original problem...but you really should not use 1080i to begin with.

1080i (where the i stands for interlaced) means that half of the pixels are virtual pixels so you only get 540 "real" pixels...it kinda tricks your tv into thinking that the video is 1080 when its not...interlaced video also is horrible at displaying fast movement, you will see a lot of blurring/trailing in 1080i that you do not get in 720p

Im doing my best to make this easy to understand because marketing has confused the subject...but 1080i is a LOWER quality video than 720p...720p is only surpassed by 1080p (and soon to come, 4320p)

anyway...if you are using 1080i, you would be better off converting your video to 720p...or re-ripping from the source to 1080p because any kind of progressive format is better than interlaced.

PS, using a deinterlacer on an interlaced video is prob a bad idea too, and may be why you see the skiping
Reply
#30
This is wrong!
1080i is not lower quality than 720p. 720p needs to be upscaled to 1080, 1080i needs to be de-interlaced. Both methods can be done with higher or lower quality.
I use vdpau temporal/spatial and the picture is perfectly fine.

Nevertheless some users want to have the tv do the de-interlacing. Check this thread:
http://forum.xbmc.org/showthread.php?tid=130059&page=2
Reply

Logout Mark Read Team Forum Stats Members Help
1080i native playback0