LAV Filters - DirectShow Media Splitter and Decoders

[ryrynz]

Nev, with the mixer enabled duplicating stereo to 7 channels drops the volume down, can I have LAV retain the volume?

[nevcairiel]

Originally Posted by ryrynz

Nev, with the mixer enabled duplicating stereo to 7 channels drops the volume down, can I have LAV retain the volume?

LAV does not support upmixing at this time.
If you play stereo and have 7 channels selected, you just get stereo in the two front channels, and a lot of silent channels. The stereo signal should be untouched.

[jq963152]

Originally Posted by nevcairiel

Only Ivy Bridge does 4K decoding, if you're running it on your 2500k from your signature, no 4K for you.

NVIDIA (via CUVID) only supports it with VP5 (VDPAU Feature Set D). AMD claims to support it with the 7xxx series, but in my tests, it just failed miserably.

Not sure if it has been posted in here already (?), but just came across the following article:

Originally Posted by anandtech.com/show/6270/ivy-bridge-gets-4k-display-support-in-october

http://www.anandtech.com/show/6270/i...ort-in-october

[...]

This October Intel will be providing a driver update for Ivy Bridge that will enable 4K x 2K resolution support as well as hardware accelerated 4K video decode.

[...]

[ryrynz]

Originally Posted by nevcairiel

The stereo signal should be untouched.

Ah, it's the receiver doing that then.

Originally Posted by nevcairiel

LAV does not support upmixing at this time.

Possible implementation soonish?

[nevcairiel]

Originally Posted by jq963152

Not sure if it has been posted in here already (?), but just came across the following article:

4k decoding already works just fine on Ivy Bridge, and has for a few month already.

[jq963152]

Originally Posted by nevcairiel

4k decoding already works just fine on Ivy Bridge, and has for a few month already.

And what could the driver update they mentioned for next month be for then ?

As you wrote it would already work: could you possibly report about the performance?

Are maxed out AVC HP@L5.2 "4K" videos with 5 reference frames and 300 Mbps at 60 fps working?

[andyvt]

Originally Posted by jq963152

And what could the driver update they mentioned for next month be for then ?

The driver adds 4K output, current drivers are limited to 1080p when not using DP. Unfortunately IVB will require two DP outs to do this, so the feature is more about press than usefulness.

Originally Posted by jq963152

As you wrote it would already work: could you possibly report about the performance?

What metrics are you interested in? CPU utilization is ~4% on an i7-3770K.

Originally Posted by jq963152

Are maxed out AVC HP@L5.2 "4K" videos with 5 reference frames and 300 Mbps at 60 fps working?

Got a sample?

[jq963152]

Thanks for your reply.

Originally Posted by andyvt

What metrics are you interested in? CPU utilization is ~4% on an i7-3770K.

With what kind of sample did you test?

Also came across the following post:

http://forum.doom9.org/showpost.php?...&postcount=133

Would "GPU Load" be the same as "VPU Load" in that case?

Originally Posted by andyvt

Got a sample?

Only came across the following sample:

http://forum.doom9.org/showthread.php?t=163825&page=2

But MediaInfo apparently says the following about that video:

Format                                   : AVC
Format/Info                              : Advanced Video Codec
Format profile                           : High@L5.1
Format settings, CABAC                   : Yes
Format settings, ReFrames                : 4 frames
Codec ID                                 : avc1
Codec ID/Info                            : Advanced Video Coding
Duration                                 : 51s 50ms
Bit rate                                 : 150 Mbps
Maximum bit rate                         : 204 Mbps
Width                                    : 3 840 pixels
Height                                   : 2 160 pixels
Display aspect ratio                     : 16:9
Frame rate mode                          : Constant
Frame rate                               : 59.940 fps
Color space                              : YUV
Chroma subsampling                       : 4:2:0
Bit depth                                : 8 bits
Scan type                                : Progressive
Bits/(Pixel*Frame)                       : 0.302
Stream size                              : 919 MiB (100%)
Writing library                          : x264 core 120 r2120 0c7dab9
Encoding settings                        : cabac=1 / ref=3 / deblock=1:0:0 / analyse=0x3:0x113 / me=hex / subme=7 / psy=1 / psy_rd=1.00:0.00 / mixed_ref=1 / me_range=16 / chroma_me=1 / trellis=1 / 8x8dct=1 / cqm=0 / deadzone=21,11 / fast_pskip=1 / chroma_qp_offset=-2 / threads=6 / sliced_threads=0 / nr=0 / decimate=1 / interlaced=0 / bluray_compat=0 / constrained_intra=0 / bframes=3 / b_pyramid=2 / b_adapt=1 / b_bias=0 / direct=1 / weightb=1 / open_gop=0 / weightp=2 / keyint=250 / keyint_min=25 / scenecut=40 / intra_refresh=0 / rc_lookahead=10 / rc=abr / mbtree=1 / bitrate=150000 / ratetol=1.0 / qcomp=0.60 / qpmin=0 / qpmax=69 / qpstep=4 / ip_ratio=1.40 / aq=1:1.00

So at least according to MediaInfo and according to the following Wikipedia page:

http://en.wikipedia.org/wiki/H.264/MPEG-4_AVC#Levels

this sample does not appear to be "fully maxed out" AVC spec wise.

With "fully maxed out", the following would be meant:

--level 5.2 --preset placebo --bitrate 300000 --vbv-bufsize 300000 --vbv-maxrate 300000

Or would:

--level 5.2 --preset placebo --crf 0 --vbv-bufsize 300000 --vbv-maxrate 300000

be better ?

And for framesize/framerate the maximum AVC specs would allow appears to be 3840x2160@60fps or 4096x2160@60fs according to the following Wikipedia page:

http://en.wikipedia.org/wiki/H.264/MPEG-4_AVC#Levels

?

4096x2304@60fps apparently is out of AVC specs and only "allowed" up to around 56 fps?

So is there any 3840x2160@60fps or 4096x2160@60fps sample out there which is encoded with:

--level 5.2 --preset placebo --bitrate 300000 --vbv-bufsize 300000 --vbv-maxrate 300000

or with:

--level 5.2 --preset placebo --crf 0 --vbv-bufsize 300000 --vbv-maxrate 300000

?

[nevcairiel]

crf 0 is lossless, the hardware cannot decode that.
Anyway, please keep stuff about hardware performance and whatnot out of the LAV thread, it adds way too much noise.

[jq963152]

Originally Posted by nevcairiel

crf 0 is lossless, the hardware cannot decode that.

But what about this part:

--vbv-bufsize 300000 --vbv-maxrate 300000

?

Wouldn't that automatically take care about bitrate not exceeding HP@L5.2 constraints?

And if not, then what about the following command line:

--level 5.2 --preset placebo --bitrate 300000 --vbv-bufsize 300000 --vbv-maxrate 300000

?

Any sample encoded with that out there ?

Or have you maybe even tried if that would work ?

[jq963152]

PS:

Originally Posted by nevcairiel

Anyway, please keep stuff about hardware performance and whatnot out of the LAV thread, it adds way too much noise.

Oh, apparently you edited your post.

Posted it in a more appropriate (?) thread then, see:

http://forum.doom9.org/showthread.ph...53#post1591353

[Pat357]

Originally Posted by nevcairiel

4k decoding already works just fine on Ivy Bridge, and has for a few month already.

The article is not about decoding, but about getting the decoded frames on a real 4K display.
It seems we still need 2 HDMI or 2 Display Port cables to "drive"a 4k display (IvyB).
Haswell should be able to do this with only 1 HDMI cable or 1 DP cable. (something about fully supporting the HDMI v1.4a spec).

[nevcairiel]

Originally Posted by Pat357

The article is not about decoding, but about getting the decoded frames on a real 4K display.
It seems we still need 2 HDMI or 2 Display Port cables to "drive"a 4k display (IvyB).
Haswell should be able to do this with only 1 HDMI cable or 1 DP cable. (something about fully supporting the HDMI v1.4a spec).

I know that, i just corrected his mis-interpretation of the article.

[jq963152]

Originally Posted by nevcairiel

I know that, i just corrected his mis-interpretation of the article.

How can you call this a "mis-interpretation" if they apparently are writing the following (especially see the highlighted (big/bolded/italic) part):

Originally Posted by anandtech.com/show/6270/ivy-bridge-gets-4k-display-support-in-october

http://www.anandtech.com/show/6270/i...ort-in-october

[...]

This October Intel will be providing a driver update for Ivy Bridge that will enable 4K x 2K resolution support as well as hardware accelerated 4K video decode.

[...]

?

[pie]

One thing about using lavfilters that is a real show stopper for me is that the splitter simply stops playback of audio when the cpu is under high load - I use a pretty old laptop so this becomes annoyingly obvious whenever I'm simultaneously web browsing (or doing anything else that moderately taxes the cpu) and watching a video.

What this equates to is jerky audio (stops and starts), making the video absolutely unwatchable - in this case the Haali splitter handles this much better as the audio generally does not stop at all (although it may lag behind the video; this is actually ok and much more watchable as it is less noticeable; it usually also catches up when the cpu is under less load too)

[cyberbeing]

Originally Posted by pie

One thing about using lavfilters that is a real show stopper for me is that the splitter simply stops playback of audio when the cpu is under high load
...
in this case the Haali splitter handles this much better...

LAV Splitter uses a Base Priority of 7 (Below Normal) for the Audio Decoder.

Web Browsers normally would have a Base Priority of 8 (Normal).

Haali Splitter uses a Base Priority of 9 (Above Normal) for the Audio Decoder.


If you boost the priority of everything through your media player (MPC-HC has a check-box for this), I suspect it would resolve your audio dropouts with LAV Splitter.

[nevcairiel]

Note that thread priorities are always relative to the process, not the system. If you have performance issues, you might as well try to increase the priority of the whole player.

[cyberbeing]

Was there a particular reason why LAV Splitter lowers the priority of the audio decoder thread while Haali Splitter, MPC-HC's Splitter, AV Splitter, and others all raise the priority?

[nevcairiel]

A 20 year old PC could decode audio, so i simply gave priority to video. Once you don't have enough performance to do both, something bad will happen, no matter how its setup.
I could probably remove that code and make it have all the same priority.

PS:
Everyone else doing something in this DirectShow world most likely just means they copied from each other, not that there is a real reason to it.

[pie]

Originally Posted by nevcairiel

A 20 year old PC could decode audio, so i simply gave priority to video. Once you don't have enough performance to do both, something bad will happen, no matter how its setup.

While it's true that pretty much any PC can decode audio, when you combine it with video and you prioritise audio below video, you're going to notice audio dropping first (when starved of resources) which is arguably bad because audio dropping is more noticeable than video lagging.

It would be good if you did remove the difference in priority ^^