Kodi Community Forum

Full Version: LibreELEC Testbuilds for RaspberryPi (Kodi 17.0)
You're currently viewing a stripped down version of our content. View the full version with proper formatting.
(2016-09-11, 17:06)popcornmix Wrote: [ -> ]
(2016-09-11, 16:09)ElectricPim Wrote: [ -> ]Are there any plans to improve VP9 support?

VP9 is only really used by Google, who also provide all files in H.264 format, so no, VP9 optimisation is very low priority.
Searching a popular torrent site for HEVC gives almost 100k hits. Searching for VP9 gets 2...

I guess most of that is offered in h264 too, but doesn't that imply that the actual legal supply of vp9 content is bigger then of hevc content?

I got the impression that the actual neon optimizations for hevc for rpi came from the raspberry/kodi developers not so much from ffmpeg developers, but maybe I'm wrong.
New LibreELEC.tv Krypton build #0911: RPi / RPi2
(Supercedes previous build)

Code:
# uname -a
Linux rpi512 4.7.3 #1 Sun Sep 11 21:18:10 BST 2016 armv6l GNU/Linux

# vcgencmd version
Sep  7 2016 14:50:28
Copyright (c) 2012 Broadcom
version 1f1070f16206004c8f2172144ec8f167db6c742f (clean) (release)

# lsb_release
LibreELEC (Milhouse) - Version: devel-20160911211547-#0911-g04b7a18 [Build #0911]

# vcdbg log msg 2>&1 | grep DTOK
002391.018: Kernel trailer DTOK property says yes

# Kernel device tree status: Enabled

Based on tip of LibreELEC.tv master (04b7a18a, changelog) and tip of XBMC master (feec0abf, changelog) with the following modifications: Build Highlights:
  1. Minors
Build Details:
  1. LibreELEC.tv:
    • scripts/unpack: clean unpatched packages (PR:722, 1 commit, 1 file changed)
  2. XBMC:
    • AESinkAudioTrack: Track wraparound (PR:10431, 1 commit, 2 files changed)
    • [Estuary] bugfixes (PR:10442, 11 commits, 23 files changed)
    • [cmake] linux: the kodi wrapper script is arch dependent (PR:10449, 1 commit, 1 file changed)
    • [skin] Don't include the themes folder in final copy as it's packaged… (PR:10452, 1 commit, 1 file changed)
    • [cmake] Fix skin packaging for windows (PR:10454, 1 commit, 1 file changed)
(2016-09-11, 22:08)ElectricPim Wrote: [ -> ]I guess most of that is offered in h264 too, but doesn't that imply that the actual legal supply of vp9 content is bigger then of hevc content?

It implies that VP9 is not widely used. We spent large amounts of engineering time on optimising HEVC.
While that could be done for VP9 it would benefit almost no one. Whereas spending the time on, say 10-bit HEVC would benefit many more users.

There is also a big technical issue with VP9. With standard encode settings the arithmetic decode of a frame is dependent on the arithmetic decode of the previous frame.
That means that multithreaded decode is very limited compared with HEVC, where we can be decoding a number of different frames concurrently.
A multi-core software decoder will always be more capable with HEVC than with VP9 (when both codecs use the default settings of encoder).
Quote:I am having the exact same problem. Only when switching from a 1080i H264 channel to a 720p h264 channel. With OMX enabled everything is good.
when a channel stutters you can also switch the deinterlacing method and then go back to "auto". this seems to disable the falsely running deinterlacing.
The last working build (with the mentioned fix) is 0726. After that i believe it was removed because it was causing crashes for someone.
the problem is that @popcornmix has not been able to reproduce this with his setup. therefore it is hard to debug.
But so far i have not come up with a way to reproduce this with a recording so that popcornmix can debug this.

Is there a way to force OMX only for LiveTV an keep using MMAL for the rest?

For what it's worth, tried build 0726 and indeed no stuttering at all on RPi2. On later builds I experience the same as described above (tvheadend client).
Hi there,

HD Homerun driver addon is not working.

Running the userspace-driver.sh spits back the following:

tvserver:~/.kodi/addons/driver.dvb.hdhomerun/bin # ./userspace-driver.sh
modprobe: FATAL: Module dvb_hdhomerun not found in directory /lib/modules/4.7.3
modprobe: FATAL: Module dvb_hdhomerun_fe not found in directory /lib/modules/4.7.3
./userspace-driver.sh: line 12: userhdhomerun: not found
tvserver:~/.kodi/addons/driver.dvb.hdhomerun/bin # grep: /var/log/dvbhdhomerun.log: No such file or directory

Works fine on Libreelec Jarvis

Cheers.
(2016-09-10, 18:38)kieranc Wrote: [ -> ]Unfortunately, I'm still having this problem, even after a full wipe. Whenever I leave the system for more than a few hours, I come back to the blue spinning circle in the middle of the screen, with no way to use the system without rebooting.
I'm 99% sure the problem is related to my PS3 bluetooth remote, if i disable bluetooth, the problem goes away. I have the 'dim' screensaver set to 10 minutes which I believe makes the remote suspend, but the problem doesn't occur after 10 minutes, only after a couple of hours.
I can save hours of debug log if necessary but I guess it's going to be tricky to find the relevant entry, is there anything else I can do which might help?

I've tested this with bluetooth enabled but the screensaver disabled, the problem does not occur, but the bluetooth remote doesn't go into standby.
With bluetooth and the screensaver enabled, the remote goes into standby when the screensaver kicks in (10 minutes) but after a couple of hours I get the perpetual blue spinning circles and have to reboot to use the system.
Any thoughts?
(2016-09-11, 21:37)MONSTA Wrote: [ -> ]Log http://sprunge.us/XNQM
CEC - off, resample - low, threshold - 2

Code:
Thread 1 (Thread 0x61cb23a0 (LWP 584)):
#0  __GI_raise ([email protected]=6) at ../sysdeps/unix/sysv/linux/raise.c:55
#1  0x750cc990 in __GI_abort () at abort.c:89
#2  0x75104004 in __libc_message ([email protected]=2, fmt=0x751b3fbc "*** Error in `%s': %s: 0x%s ***\n") at ../sysdeps/posix/libc_fatal.c:175
#3  0x7510a3c0 in malloc_printerr (action=<optimized out>, str=0x751b4490 "malloc(): smallbin double linked list corrupted", ptr=<optimized out>, ar_ptr=<optimized out>) at malloc.c:5007
#4  0x7510cc68 in _int_malloc ([email protected]=0x73000010, [email protected]=13) at malloc.c:3387
#5  0x7510e680 in __GI___libc_malloc (bytes=13) at malloc.c:2914
is generally heap corruption (e.g. writing beyond end of buffer).

Do you have a way of reproducing this reliably? E.g. with a specific file?
(2016-09-12, 13:43)popcornmix Wrote: [ -> ]
(2016-09-11, 22:08)ElectricPim Wrote: [ -> ]I guess most of that is offered in h264 too, but doesn't that imply that the actual legal supply of vp9 content is bigger then of hevc content?

It implies that VP9 is not widely used. We spent large amounts of engineering time on optimising HEVC.
While that could be done for VP9 it would benefit almost no one. Whereas spending the time on, say 10-bit HEVC would benefit many more users.
It is used by YouTube, apparently with DASH streaming. I doubt YouTube will ever use hevc. It is supported in Chrome, Edge, Firefox, and Opera and Android 4.4+ (and JW Player)

(2016-09-12, 13:43)popcornmix Wrote: [ -> ]There is also a big technical issue with VP9. With standard encode settings the arithmetic decode of a frame is dependent on the arithmetic decode of the previous frame.
That means that multithreaded decode is very limited compared with HEVC, where we can be decoding a number of different frames concurrently.
A multi-core software decoder will always be more capable with HEVC than with VP9 (when both codecs use the default settings of encoder).

Recommended encoding settings for VP9 nowadays support multi-threading decoding.
http://wiki.webmproject.org/ffmpeg/vp9-encoding-guide

My experience is that decoding VP9 is much faster on x86 than hevc. It's a pity it isn't on ARM architecture. Encoding VP9 though is much slower. I compiled ffmpeg with vp9 support on a rpi2, it is super slow (0.1fps), but much better quality per bitrate then hardware h264 encoding (10fps).

1080p hevc seems the limit for an RPI3, but when VP9 decoding is really 55% faster it feels like a nice option.
Hopefully we will see AV1, and hardware support some day.

I certainly appreciate all your and others efforts on optimising Kodi / HEVC RPI, and I'm not advocating it should shift to VP9, I just wondered, I have the feeling VP9 is underestimated.
(2016-09-12, 18:28)loggio Wrote: [ -> ]Hi there,

HD Homerun driver addon is not working.

Running the userspace-driver.sh spits back the following:

tvserver:~/.kodi/addons/driver.dvb.hdhomerun/bin # ./userspace-driver.sh
modprobe: FATAL: Module dvb_hdhomerun not found in directory /lib/modules/4.7.3
modprobe: FATAL: Module dvb_hdhomerun_fe not found in directory /lib/modules/4.7.3
./userspace-driver.sh: line 12: userhdhomerun: not found
tvserver:~/.kodi/addons/driver.dvb.hdhomerun/bin # grep: /var/log/dvbhdhomerun.log: No such file or directory

Works fine on Libreelec Jarvis

Cheers.

For now it doesn't look like the HD Homerun add-on is compatible with LibreELEC 8.0. The drivers don't build against the latest 4.7.x kernel which is why they are missing, the error is that the HD Homerun add-on continues to be offered for LE8 when it shouldn't be. It's being looked into, there's a feeling it's no longer required, it will either be fixed or removed.
(2016-09-12, 22:31)Milhouse Wrote: [ -> ]
(2016-09-12, 18:28)loggio Wrote: [ -> ]Hi there,

HD Homerun driver addon is not working.

Running the userspace-driver.sh spits back the following:

tvserver:~/.kodi/addons/driver.dvb.hdhomerun/bin # ./userspace-driver.sh
modprobe: FATAL: Module dvb_hdhomerun not found in directory /lib/modules/4.7.3
modprobe: FATAL: Module dvb_hdhomerun_fe not found in directory /lib/modules/4.7.3
./userspace-driver.sh: line 12: userhdhomerun: not found
tvserver:~/.kodi/addons/driver.dvb.hdhomerun/bin # grep: /var/log/dvbhdhomerun.log: No such file or directory

Works fine on Libreelec Jarvis

Cheers.

For now it doesn't look like the HD Homerun add-on is compatible with LibreELEC 8.0. The drivers don't build against the latest 4.7.x kernel which is why they are missing, the error is that the HD Homerun add-on continues to be offered for LE8 when it shouldn't be. It's being looked into, there's a feeling it's no longer required, it will either be fixed or removed.

Thanks for your response.

I believe this to be true for tvheadend as it recognises the HDhomerun tuners (must have internal support for hdhomerun devices)
But for the VDR backend the drivers are needed for the backend to use the HDhomerun tuners.
Hopefully this can be fixed as there seems to be a lot of users that use hdhomerun devices.

Thanks again.
New LibreELEC.tv Krypton build #0912: RPi / RPi2
(Supercedes previous build)

Code:
# uname -a
Linux rpi512 4.7.3 #1 Mon Sep 12 23:27:54 BST 2016 armv6l GNU/Linux

# vcgencmd version
Sep 12 2016 19:09:28
Copyright (c) 2012 Broadcom
version b804f093e774e56dc1d441e0aed8eff5ac66b0ae (clean) (release)

# lsb_release
LibreELEC (Milhouse) - Version: devel-20160912232619-#0912-gd683a66 [Build #0912]

# vcdbg log msg 2>&1 | grep DTOK
002392.373: Kernel trailer DTOK property says yes

# Kernel device tree status: Enabled

Based on tip of LibreELEC.tv master (d683a662, changelog) and tip of XBMC master (4c91ab92, changelog) with the following modifications: Build Highlights:
  1. New firmware - bitmapped warning icons and deinterlace fix
  2. MMALRender: Ignore unexpected timestamp difference in framerate calculation
Build Details:
  1. Firmware (Sep 12):
    • firmware: arm_display: Add bitmapped icons for warning conditions See: link
    • firmware: deinterlace: Avoid frame doubling with progressive frames. See: link
  2. LibreELEC.tv:
    • u-boot: build second u-boot package for imx6 project (PR:704, 1 commit, 5 files changed)
    • moonlight: update addon (PR:725, 1 commit, 4 files changed)
    • emby: update to 3.0.7100 (PR:724, 1 commit, 2 files changed)
  3. XBMC:
    • win32: fix eac3 for wasapi, 2nd round (PR:10453, 1 commit, 1 file changed)
    • [cmake] linux: install missing peripheral and vfs addon headers (PR:10457, 1 commit, 2 files changed)
    • CGUIDialogContentSettings: fix crash when choosing a scraper and no scraper is set yet (PR:10447, 1 commit, 1 file changed)
    • [Estuary][PVR] Guide window: add vertical scrollbar to channel list. (PR:10450, 2 commits, 4 files changed)
  4. inputstream.mpd:
  5. newclock5:
    • New commits in this build:
      • MMALRender: Ignore unexpected timestamp difference in framerate calculation (1a415e79)
  6. Additional commits/pull requests/changes not yet merged upstream:
    • Added: [pkg] PR:10459: AE: CActiveAEResampleFFMPEG check for resampling being active before calling swr_set_compensation
The 1080i to 720p channel switching seem to be fixed now.
(2016-09-10, 16:11)bmonster Wrote: [ -> ]That's interesting, I'm only using OMX at the moment, mmal get my pi3 so hot and I feel I get a better picture with OMX.
Strange that changing my skin seems to have sorted it, will try a few other films in a bit.

Edit:

Yer I still have the sub problem, its on 90% of what I play, guessing the other 10% could a different sub type.

I also have the same problem since a couple of days. Subtiltes are displayed until a new sub-title starts. Using .srt-subtitles ...
(2016-09-12, 21:40)popcornmix Wrote: [ -> ]Do you have a way of reproducing this reliably? E.g. with a specific file?

Always (24>25 fps). Absolutely any mkv. The last one was bd remux 40 gb. Before that varius rips 4-15 gb. No matter from ntfs or ext4 (if it's important).
m2ts or avi didn't tested, but sure get the same Smile
I noticed last night and previous night when playing a 5.1 audio show that after a pause and play there is a chance where the audio channels don't come back properly. For example on 9/11 I paused the show we were watching to talk a second and hit play again but then the center channel which carried the voices was missing, we heard audio from the other channels but the actors had no voices. Last night on 9/12 we paused the show to talk and upon hitting play the actors sounded like they were really far away at the end of a tunnel. On both occasions hitting stop and play again resolved the issue, I was even able to skip back and go over the same scene to ensure there was no video corruption.

I recently switched on this machine from an AMD APU setup to a RPi3 due to advances in technology and not sure if this is only limited to the RPi or if anyone else has experienced this as well. All my other RPi systems are on 2.0 systems straight into the TV so this is a new experience for me in the RPi space with 5.1 audio. Here is a debug log from last night.

http://sprunge.us/QPfT