• 1
  • 85
  • 86
  • 87(current)
  • 88
  • 89
  • 111
OpenELEC Testbuilds for RaspberryPi (Kodi 15.0) Part 1
New OpenELEC Isengard build #0324: RPi / RPi2
(Supercedes previous build)

Code:
# uname -a
Linux rpi512 3.19.2 #1 Tue Mar 24 21:03:39 GMT 2015 armv6l GNU/Linux

# vcgencmd version
Mar 22 2015 15:56:19
Copyright (c) 2012 Broadcom
version 6d87a21b1e7a0744092d1dc481f3ae3c56bdff1a (clean) (release)

# lsb_release
OpenELEC (Milhouse) - Version: devel-20150324210249-#0324-g115800e [Build #0324]

# vcdbg log msg 2>&1 | grep DTOK
001663.228: Kernel trailer DTOK property says yes

# Kernel device tree status: Enabled

Based on tip of OpenELEC master (115800e4, changelog) and tip of XBMC master (1fb19a3a, changelog) with the following modifications: Build Highlights:
  1. ffmpeg threading reworked
  2. dcadec (DTS-HD) updated
  3. omxplayer stereoupmix and fixed behave more like dvdplayer
  4. Fast deinterlace
Build Details:
  1. OpenELEC:
    • libva-driver-intel: VPP: Make sure the store buffer is allocated (115800e4)
  2. XBMC:
    • [AE/CA/OSX] - support devices with up to 72 channels (64 unknown + 8 mapped) (PR:6797, 2 commits, 3 files changed)
    • [project readme.md] update quick links (PR:6796, 1 commit, 1 file changed)
    • rework frame threading for ffmpeg (PR:6771, 2 commits, 5 files changed)
    • [stringutils] fix sizestring alignment (PR:6786, 1 commit, 1 file changed)
    • [pvr.iptvsimple] Sync with v1.11.0 (API 1.9.6) (PR:6789, 1 commit, 1 file changed)
    • fix frametime for active vsync (PR:6709, 2 commits, 4 files changed)
  3. dcadec:
    • Document 16-bit big-endian bitstream format requirement. (f8e09437)
    • Implement strict decoding mode. (5f4740a0)
    • Enable strict decoding by default in dcadec. (9bd837e4)
    • Be more strict about XLL bit depth. (12c142af)
    • Remove redundant X96 and XCH header validation checks. (fe23d004)
    • Count core channels in a more robust way. (87ed40e4)
  4. newclock4:
    • New commits in this build:
      • [OMXAudio] Make stereoupmix and fixed behave more like dvdplayer (daf0fc0c)
      • squash: [omxplayer] Handle failures from EmptyThisBuffer (539083d2)
      • squash: [RBP] Make vsync event persistant (d074a905)
      • [mmal] Explicitly use fast deinterlace for HD and reduce buffers allocated by 2 (7f7e0a11)
    • Commits no longer in build:
      • guilib: only update scrollinfo if frametime did change (2dfe88ae)
      • fix frametime for active vsync (d528de96)
      • dvdplayer - rework frame therading for ffmpeg (48f02a03)
      • vaapi: reduce refs by one after rework of frame threading (e94d5355)
  5. Additional commits/pull requests/changes not yet merged upstream:
Texture Cache Maintenance Utility: Preload your texture cache for optimal UI performance. Remotely manage media libraries. Purge unused artwork to free up space. Find missing media. Configurable QA check to highlight metadata issues. Aid in diagnosis of library and cache related problems.
Thank you. So, I finally started testing with my Pi2 with build #0324, and...

- 23.976 is perfect, as is 59.940, 24, 60
- Haven't tested mpeg2 yet (OTA HD)
- Dolby TrueHD & ATMOS decode perfectly to 8 channel. Sounds as good as AVR doing it
- DTS-MA, same story, decodes fine
- MVC MKV's play great in full, frame-packed 3D (tested rip of 'Exodus: Gods and Kings' and 'Avatar')

Two screenshots to share (the dropped frames is likely due to SMB vs. NFS?):
- ATMOS test
- 3D MVC MKV test

Issues I'm seeing are stuttering (over wired SMB) with above & high bitrate test clips (Birds @ 90Mbps, & Jellyfish @ 100Mbps). I tried NFS, but it was worse. I have a Synology 1812+, and not sure how to use NFS properly. I believe I need to set Proto to UDP & rsize to 32768... but, don't know how to do this with my RPi2.
[H]i-[d]eft [M]edia [K]een [V]ideosaurus
My Family Room Theater
(2015-03-25, 02:04)hdmkv Wrote: Issues I'm seeing are stuttering (over wired SMB) with above & high bitrate test clips (Birds @ 90Mbps, & Jellyfish @ 100Mbps). I tried NFS, but it was worse. I have a Synology 1812+, and not sure how to use NFS properly. I believe I need to set Proto to UDP & rsize to 32768... but, don't know how to do this with my RPi2.

I think that's too high a bitrate for the VideoCore 4 GPU in the Pi/Pi2.

It's not really a network or CPU issue on the Pi2, as neither are being maxed out (although CPU would most likely become the bottleneck on the Pi1, before the GPU).

You can run "bcmstat.sh cxpd10" (included as standard in my builds) and you'll see the network is not being saturated by Jellyfish @ 90Mbps.

I'm seeing an average of about 8-9MB/s over SMB for Jellyfish @ 90Mbps, with Pi2 1GHz CPU load @ 11% (35% for the single core handling playback and presumably network IO). This stutters, badly.

A lower bit rate file, Jellyfish @ 60Mbps, has broadly similar network throughput and CPU load, but no stutter, so my opinion is that 60-70Mbps is the bitrate limit for the VC4 GPU and this limit is reached first before any other component in the system (ie. CPU, network).

Overclocking the GPU (h264 etc.) might help a bit, but you're unlikely to see files with these kind of torture-test bit rates in normal viewing so the increased risk of instability probably isn't worth it (IMHO).
Texture Cache Maintenance Utility: Preload your texture cache for optimal UI performance. Remotely manage media libraries. Purge unused artwork to free up space. Find missing media. Configurable QA check to highlight metadata issues. Aid in diagnosis of library and cache related problems.
Thanks, I'll try overclocking. I'm not sure, but don't some 3D BD's reach bitrate peaks of around 90Mbps? Also, are there instructions somewhere on how to properly use NFS with the values I posted above?
[H]i-[d]eft [M]edia [K]een [V]ideosaurus
My Family Room Theater
(2015-03-25, 02:49)hdmkv Wrote: Thanks, I'll try overclocking. I'm not sure, but don't some 3D BD's reach bitrate peaks of around 90Mbps? Also, are there instructions somewhere on how to properly use NFS with the values I posted above?

Maximum Blu-Ray video bitrate is 40Mb/s (60Mb/s for MVC).
(2015-03-25, 02:43)Milhouse Wrote: ...so my opinion is that 60-70Mbps is the bitrate limit for the VC4 GPU and this limit is reached first before any other component in the system (ie. CPU, network).

Overclocking the GPU (h264 etc.) might help a bit, but you're unlikely to see files with these kind of torture-test bit rates in normal viewing so the increased risk of instability probably isn't worth it (IMHO).

Yes agreed, this was my conclusion when I did my testing a while back over both NFS and SMB. 70Mbps was the limit over NFS with an Overclock.

(2015-03-25, 03:43)wrxtasy Wrote: Yes agreed, this was my conclusion when I did my testing a while back over both NFS and SMB. 70Mbps was the limit over NFS.

You can play 80Mb/s jellyfish over NFS with an OS mount.
The inbuilt libnfs is a bit slower.
(2015-03-25, 01:03)Milhouse Wrote: Build Highlights:
  1. Fast deinterlace
Sorry. What does this mean?
Generally, deinterlacing in RPi looks strange. Like some pulsating in motion. One frame is just great, next creepy.
http://dropmefiles.com/Q9WDw
Some Omxplayer trouble with build #0323 and #0324 (Rpi1):
player hangs after skipping backward, then I get a black screen trying to stop it.

See the log:
http://pastebin.com/zjzsGqcA

Thanks

P.s.

Build #0321 is fine.
with #0323 I have serious issues when mmal renderer is used. As soon as there is any UI overlay playback is choppy with lots of drops. This doesn't happen with omxplayer. I tried with GUI update limit to 24fps and without any limit - same effect.

MVC-3D playback is smooth with mmal and omxplayer
(2015-03-25, 02:49)hdmkv Wrote: Thanks, I'll try overclocking. I'm not sure, but don't some 3D BD's reach bitrate peaks of around 90Mbps? Also, are there instructions somewhere on how to properly use NFS with the values I posted above?

Look here for the NFS options:

http://forum.kodi.tv/showthread.php?tid=...pid1946803

There's info on how to do it through systemd on the OE wiki:

http://wiki.openelec.tv/index.php/Mounti...mple_mount
Hi,

To test the Master HD Sound, what must i change in the Settings? I ve pssthrough activated and the i have a DD and DTS captable receiver. But in the Hobbit with Master HD it shows only DTS in my Receiver which is a DTS HD Receiver.

Thx for Info Wink

@popcornmix:
for Info: TAB 3D Movies are still with no picture in this build with option "Use Full HDMI modes for 3D" activated.
The picture still not there if i turn off 3D on my TV while playing.


Greetz
J.
Detailed a few posts back in the 323 release announcement.
If I have helped you or increased your knowledge, click the 'thumbs up' button to give thanks :) (People with less than 20 posts won't see the "thumbs up" button.)
(2015-03-25, 11:13)twelvebore Wrote: Look here for the NFS options:

http://forum.kodi.tv/showthread.php?tid=...pid1946803

There's info on how to do it through systemd on the OE wiki:

http://wiki.openelec.tv/index.php/Mounti...mple_mount
Thanks... over my head, but will read up and try it out.
[H]i-[d]eft [M]edia [K]een [V]ideosaurus
My Family Room Theater
(2015-03-25, 07:47)MONSTA Wrote: Sorry. What does this mean?
Generally, deinterlacing in RPi looks strange. Like some pulsating in motion. One frame is just great, next creepy.
http://dropmefiles.com/Q9WDw

There are two deinterlace algorithms available:
advanced - works like yadif2 but can only handle up to DVD resolution
bob (fast) - simpler but works at 1080i.

advanced produces a sharper image, but may have more motion artifacts. You can switch between them with OSD video menu.
If you have a sample that doesn't deinterlace well, then upload (e.g. to google drive or dropbox) and I'll see if anything is wrong.

As advanced doesn't work with 1080i, we fall back to bob on the firmware side. However doing that is less efficient that requesting the right
algorithm from Kodi (where we can also request fewer buffers, which saves 3M of GPU memory).
omxplayer has always done this, but this update makes dvdplayer/mmal do the same thing.
  • 1
  • 85
  • 86
  • 87(current)
  • 88
  • 89
  • 111

Logout Mark Read Team Forum Stats Members Help
OpenELEC Testbuilds for RaspberryPi (Kodi 15.0) Part 112