Kodi Community Forum
Linux VAAPI: Nuc, Chromebox, HSW, IVB, Baytrail with Ubuntu 14.04 - Printable Version

+- Kodi Community Forum (https://forum.kodi.tv)
+-- Forum: Support (https://forum.kodi.tv/forumdisplay.php?fid=33)
+--- Forum: General Support (https://forum.kodi.tv/forumdisplay.php?fid=111)
+---- Forum: Linux (https://forum.kodi.tv/forumdisplay.php?fid=52)
+---- Thread: Linux VAAPI: Nuc, Chromebox, HSW, IVB, Baytrail with Ubuntu 14.04 (/showthread.php?tid=165707)



RE: vaapi-sse4: Deinterlacing Testing - plinkyplonky - 2014-09-28

Getting pausing issues on 1080i 50hz videos which are fairly high bitrate - 19mbps. Authored from FInal Cut Pro from my camcorder. It's not happening all the time but pretty regularly. It will pause for about 3 seconds, sometimes the sound continues and then it catches up the video and continues playing. When it does play the deinterlacing is very good on my Chromebox 2955U. (Having recently moved from a N2820 NUC which sadly has a crippled GPU for deinterlacing).

Debug logs as follows (glitch occurred at 12:17:33)
$ dmesg -T | pastebinit
http://paste.ubuntu.com/8448925/
$ cat ~/.xbmc/temp/xbmc.log | pastebinit
http://paste.ubuntu.com/8448935/
$ cat /var/log/Xorg.0.log | pastebinit
http://paste.ubuntu.com/8448936/
$ DISPLAY=:0 vainfo | pastebinit
http://paste.ubuntu.com/8448944/
$ dpkg -l |grep mesa | pastebinit
http://paste.ubuntu.com/8448952/
$ id | pastebinit
http://paste.ubuntu.com/8448953/
$ aplay -L | pastebinit
http://paste.ubuntu.com/8448954/


RE: vaapi-sse4: Deinterlacing Testing - fritsch - 2014-09-28

Please provide a Debug Log with just that non working sample playing.

Quote:[Sun Sep 28 12:17:37 2014] [drm] stuck on render ring
[Sun Sep 28 12:17:37 2014] [drm] GPU crash dump saved to /sys/class/drm/card0/error
[Sun Sep 28 12:17:37 2014] [drm] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.
[Sun Sep 28 12:17:37 2014] [drm] Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel
[Sun Sep 28 12:17:37 2014] [drm] drm/i915 developers can then reassign to the right component if it's not a kernel issue.
[Sun Sep 28 12:17:37 2014] [drm] The gpu crash dump is required to analyze gpu hangs, so please always attach it.
[Sun Sep 28 12:17:44 2014] [drm] stuck on render ring
[Sun Sep 28 12:17:44 2014] [drm:i915_set_reset_status] *ERROR* render ring hung inside bo (0x2900000 ctx 1) at 0x29002c4

This one sucks though.

Try to upgrade to kernel 3.16.3, please via ubuntu mainline.


RE: vaapi-sse4: Deinterlacing Testing - illiac4 - 2014-09-29

Is it something wrong with builds or. am i just too fast?

The following packages have unmet dependencies.
xbmc : Depends: xbmc-bin (< 2:14.0~git20140928.0500-c350d22-0trusty.1~) but 2:14.0~git20140929.0500-795877c-0trusty is to be installed
E: Unable to correct problems, you have held broken packages.


P.S.:
OK I was too fast it has not build yet. Now it is ok.


RE: vaapi-sse4: Deinterlacing Testing - sveni_lee - 2014-09-29

(2014-09-27, 13:16)fritsch Wrote:
Quote:10:24:49 T:140281793988352 DEBUG: CPythonInvoker(13, /home/xbmc/.xbmc/addons/script.metadata.actors/resources/lib/service.py): th

Fix that addon or remove it.

Thanks, that helped for me... afer removing the addon, no more crashes...

(2014-09-28, 14:41)fritsch Wrote: This patch should fix the python issue of sveni_lee and orduek: https://github.com/xbmc/xbmc/pull/5425 - when jenkins is happy I will push it.

So I should update again?


RE: vaapi-sse4: Deinterlacing Testing - fritsch - 2014-09-29

Vorwärts immer - rückwärts nimmer!


RE: vaapi-sse4: Deinterlacing Testing - sveni_lee - 2014-09-29

:-))

"Vorwärts Leute... wir müssen zurück"


RE: vaapi-sse4: Deinterlacing Testing - Frank-NL - 2014-10-01

Libva-1.4.0 is now in OE 4.2. So Gotham will have the vaapi deint features as well?


RE: vaapi-sse4: Deinterlacing Testing - fritsch - 2014-10-01

No. OE will only have VAAPI-BOB

Libva-1.4.0 is not enough for mcdi and madi, you need 8 additional patches, see: https://github.com/fritsch/libva-intel-driver/commits/ppa14

Everything with the name "vebox" in it is needed. I think the "nightly builds" of OpenELEC have the relevant code, but ask sraue - he knows best.

Edit: I just verified the code. It only has SW Deinterlacing with either Yadif or OpenGL Bob. And no automatically switching between VAAPI Render and SWRender in OE 4.2
OpenELEC master has a two days old patch for all VPP methods and they provide the relevant driver with the patches https://github.com/OpenELEC/OpenELEC.tv/blob/master/packages/multimedia/libva-intel-driver/patches/libva-intel-driver-001-upstream.patch


RE: vaapi-sse4: Deinterlacing Testing - -DDD- - 2014-10-01

Gotham have deint (yadif) since dec 2013.
Or do you mean madi DI with vpp?


RE: vaapi-sse4: Deinterlacing Testing - FernetMenta - 2014-10-01

If anything vpp related shows up in Gotham it is without my knowledge, means that behavior is undefined Smile


RE: vaapi-sse4: Deinterlacing Testing - Frank-NL - 2014-10-01

Thanks fritsch. Still enjoying your yadif work for a while! Smile


RE: vaapi-sse4: Deinterlacing Testing - fritsch - 2014-10-01

Btw. now after 4.2 is released - the next OE release will be 5.0 which will feature Helix. So the next alpha / beta / gammas will most likely have all the VPP work.


RE: vaapi-sse4: Deinterlacing Testing - Frank-NL - 2014-10-01

Yes looking forward to OE5.0! Just waiting for the betas and I'll switch.


RE: vaapi-sse4: Deinterlacing Testing - robo989 - 2014-10-01

(2014-09-22, 11:00)robo989 Wrote: Aftert recent sudo apt-get update/upgrade to Alpha4 the options for the MADI/MCDI deinterling have vanished leaving the usual bob/weave/deinterlace etc...
Any ideas on that? Thanks

(2014-09-22, 11:41)fritsch Wrote: Provide what the first page wants, thanks.

Have done so and then the MCDI/MADI modes started working on haswell.
Have just done and update/upgrade and the modes are gone again Huh


RE: vaapi-sse4: Deinterlacing Testing - fritsch - 2014-10-01

Provide what the first page wants, thanks.