Kodi Community Forum

Full Version: VAAPI: Nuc, Chromebox, HSW, IVB, Baytrail with Ubuntu 14.04
You're currently viewing a stripped down version of our content. View the full version with proper formatting.
I've checked with 720p content that shows artifacts when lanczos 3 is enabled running intel_gpu_top, it never goes above 40% GPU usage:
http://img854.imageshack.us/img854/5967/yqp9.png . For comparison, GPU usage with bilinear: http://img842.imageshack.us/img842/6375/elqz.png (8%).

Are you sure this system cannot handle it?

This Chromebox seems to be the most powerful of my HTPCs and it's the only one that handles 1080i Deinterlace with no issues.
That looks in deed good. Mesa issue?
1080i, Deinterlace, Lanczos 3: 60% GPU load ( http://img835.imageshack.us/img835/38/n3pk.png ).
1080i, Deinterlace, Bilinear : 20% GPU load ( http://img547.imageshack.us/img547/6420/qv3m.png ).

It appears that Lanczos 3 is really demanding.

As hardware is probably not the issue here, I'll try to poke with software. Mesa is 10.0.3 from ppa:wsnipex/mesa.
> A HD2000 is too slow to do Lanczos 3 upscaling

BTW, it's HD Graphics 3000 according to the specs: http://www.chromestory.com/2012/05/two-c...l-core-i5/ .
I've had several crashes watching the olympics recoded off NBC which is 1080i where I am. I'm watching the recordings directly off the .mpg file, not through the PVR addon.

Here are several crash logs:

http://pastebin.com/cz4SKWDZ
http://pastebin.com/wzBHaj1U
http://pastebin.com/FnkmkQYw
http://pastebin.com/kLUr9QDv

The wife didn't have the patience to wait for me to do much but reset XBMC, but I did save one log file, not much I can see that isn't in the crash log, but here it is:

http://pastebin.com/AVRkZqc5

Needless to say, I wasn't running with debug logging enabled.

Basically XBMC just exited without warning during playback and dumped me back to the desktop. I'm running under xubuntu 13.10 without compositing and an Intel i7-3770. I run with the AE_SINK=ALSA option as the pulse audio sink was giving me problems.
you are running mesa 10.1.
and you didn't follow the howto with xubuntu.
No gdb installed. From what I see - your backend seems highly unstable and spamming the log allover, sorry cannot see anything with the logs. Would also be nice, if you could be more "precise" about the PASink.
wsnipex:
I didn't follow the howto exactly, as I had already installed ubuntu and a mainline XBMC on my new linux box before I found this thread. I had too many other things already set-up on the box to start from scratch with a minimal ubuntu install.

I thought xubuntu would work since you said:

(2014-02-05, 09:14)wsnipex Wrote: [ -> ]xfce would probably work if you disable compositing.

I have to admit that I am unclear on which mesa should be installed for this version of XBMC. I've never tried to do anything to change it from what came with the ubuntu install iso that I recall, except for installing your repository per the howto. But you are right, I have a git version of 10.1 installed. Should I try to downgrade to 10.0.3 which is what apt-cache tells me your repository provides? If so, any advice on the best way to do this given that I can't recall how or why I installed 10.1? For the life of me I can't find what repository or source code 10.1 came from on my system.

fritsch:
The spam in the log seems to be a bug in the cmyth-addon rather than my mythtv backend. It doesn't seem to like my backend having two sources (antennas). It's actually been doing this for a while, even when I had my old mac mini running myth and XBMC. I will have to post on the cmyth forum and see if there is something that can be done, but, other than the spam it seems to be harmless.

The howto didn't mention gdb, and I don't do enough code development to have it installed, but I don't mind installing it if it will help debug the code (a de-bugger helping to debug code - I suppose that's pretty obvious). Would it be worth it given my relative inexperience with it?

I didn't want to hijack this thread with PAsink issues, but I tried using it and, aside form the broken pass-though bug in ubuntu, I had mixed up channel mapping when not using pass-through in XBMC. But only in XBMC, VLC and the built in video player in xubuntu (Parole Media Player) both played a test file correctly. I also noticed that flac and DTS audio tracks sounded distorted with PAsink. Since everything worked fine with the ALSA option, I have been putting off posting in the PAsink thread. It's 4 AM here, so I'll try to post something with logs and a link to a decent test file tomorrow in the proper thread.
Just do: sudo apt-get install gdb and the crashlog will pick it up

Concerning Mesa: Try to remove the ppa with ppa-purge - where you got mesa 10.1 from. I think it's oibaf version of it.

PASink: The channel bug has been fixed - that was a nice one. I forgot to actually set the channel map.

And yeah, don't hurry too much. All is fine - we get it going.

Good Night. Sleep is important - don't want to see you hear again before +8 hours ;-)
So I tried to install gdb, and it says I already have it installed. Is there anything I have to set up to make it work with the crashlog?

I didn't have the oibaf ppa installed, so I still can't figure out where the mesa 10.1 stuff came from. However, installing oibaf and then removing it with ppa-purge worked anyway, and I'm running 10.0.3 now.

You're right, the channels are now correct using Pulse. I did note that it can only do 5.1 channels while ALSA can send a 7.1 channel LPCM stream to my AV receiver. This looks like a limitation of Pulse on my system, as I can only find a 5.1 channel device for my HDMI in the Pulse GUI. I only have the one test file with 7.1 channel audio, so it's not really a limitation in my real use.

One thing that's weird. I tried Pulse in a different user so I wouldn't screw up my mostly working one, and I saw some tearing in the video I was trying out. The same video doesn't tear in my regular user. I checked the obvious settings for de-interlacing, scaling, vertical sync and syncing video to the refresh rate, and they were identical. I'm assuming it's something else different between the two users, so I'll keep looking.
Hello, great work on the ubuntu repo for xbmc posted in the first thread, works like a charm on my intel nuc i3 with thunderbolt.

just one thing that doesnt work as expeced, before i used a raspberry pi, and used ios7 ipad mini, and airplay worked nice somewhat instable at times but it worked..

now it seems that avahi-daemon has a setting i had to fix in /etc/default/avahi-daemon from 1 to 0 (the only row..) and then avahi accually starts on boot.
(using a basic ubuntu server 13.10 with xbmc and everything for it and just made a update-rc.d avahi-daemon defaults)

anyway now on boot (it boots into xbmc automatically) and ios device can see xbmc as a airplay device, i choose it and spotify pauses the playback, and nothing shows on the tv and no sound..

so i just wonder if airplay is broken in these builds, or if its supposed to work so i know if i have to compile from source or where to start Smile

edit i tried a apt-get update and upgrade, and next version of xbmc airplay works again.. so it seems to have been a bug in that build i had,, Smile
Airplay is not touched. It's build like mainline xbmc. Your avahi change looks really odd ... should not be needed at all. Triple check the polkit stuff.
Not sure if it's offtopic here, but is there a way to control overscan via the intel driver/xorg? My Chromebox DP->HDMI output to TV has 5 pixels overscan only on the right side of TV.

I know that it can be calibrated in XBMC and the calibration to 5 pixels for the right edge works just fine, but is there a way to set it globally for the system?

There is a related post at http://ubuntuforums.org/showthread.php?t=1824435 and I get the same errors as in the last comment.
Yes, it can be done with xrandr. But I would not suggest doing this. Set the TV to "Just scan" or "Fit" or something else to get a 1:1 pixel mapping.
Yea, the overscan on my Samsung TV drove me crazy until I finally realized that "Screen Fit" was the setting to get 1:1 pixel mapping, not "16:9"