• 1
  • 18
  • 19
  • 20(current)
  • 21
  • 22
  • 146
OpenELEC Testbuilds for RaspberryPi (Kodi 17.0)
(2015-12-22, 23:46)popcornmix Wrote:
(2015-12-22, 23:25)bill_orange Wrote: Okay....progress. Build 1207 does not have the problem and build 1208 and subsequent locks up. The sequence of events to get a lock up is:

If it was #1208, then PR8497 and the pulse audio changes look like the only music related changes.

Can PR8497 be excluded in a subsequent build? Or perhaps a fork?
(2015-12-22, 23:25)bill_orange Wrote: Here are some sample tracks: https://onedrive.live.com/redir?resid=F1...file%2czip

I've unzipped the two albums and have played them from Music > Files (over smb://)

Both albums play without any problems, with no lockups.

However the Hillbilly Holiday tracks all give the following output so this seems to be some sort of encoding issue:
Code:
23:21:56 6572.665527 T:1579152288    INFO: ffmpeg[5E1FF3A0]: [mp3] overread, skip -7 enddists: -6 -6
23:21:56 6572.687988 T:1579152288    INFO: ffmpeg[5E1FF3A0]: [mp3] overread, skip -4 enddists: -1 -1
23:21:56 6572.765137 T:1579152288    INFO: ffmpeg[5E1FF3A0]: [mp3] overread, skip -5 enddists: -4 -4
23:21:56 6572.787598 T:1579152288    INFO: ffmpeg[5E1FF3A0]: [mp3] overread, skip -4 enddists: -2 -2
23:21:56 6572.810547 T:1579152288    INFO: ffmpeg[5E1FF3A0]: [mp3] overread, skip -4 enddists: -1 -1

Will see about reverting PR8497 in a test build.
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.
(2015-12-23, 01:24)Milhouse Wrote:
(2015-12-22, 23:25)bill_orange Wrote: Here are some sample tracks: https://onedrive.live.com/redir?resid=F1...file%2czip

I've unzipped the two albums and have played them from Music > Files (over smb://)

Both albums play without any problems, with no lockups.

However the Hillbilly Holiday tracks all give the following output so this seems to be some sort of encoding issue:
Code:
23:21:56 6572.665527 T:1579152288    INFO: ffmpeg[5E1FF3A0]: [mp3] overread, skip -7 enddists: -6 -6
23:21:56 6572.687988 T:1579152288    INFO: ffmpeg[5E1FF3A0]: [mp3] overread, skip -4 enddists: -1 -1
23:21:56 6572.765137 T:1579152288    INFO: ffmpeg[5E1FF3A0]: [mp3] overread, skip -5 enddists: -4 -4
23:21:56 6572.787598 T:1579152288    INFO: ffmpeg[5E1FF3A0]: [mp3] overread, skip -4 enddists: -2 -2
23:21:56 6572.810547 T:1579152288    INFO: ffmpeg[5E1FF3A0]: [mp3] overread, skip -4 enddists: -1 -1

Will see about reverting PR8497 in a test build.

Interesting.

Using >> Files the problem does not occur. At least to my layman's eyes, that makes PR8497 even more likely to be the problem.
@bill_orange: Here are two RPi2 test builds. Try them and report which works best, assuming there is any difference at all.

#1212y: Reverted PR8497
#1212z: Reverted PulseAudio -> ALSA Bridge (PR:4509 and PR:4482)
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.
Hello,
with #1222 i notice stuttering video and artefacts at beginnig to start the "Tina" DVD.
Buffering is also little higher than build in build 1221.

Sample _Test.7z is on GoogleDrive, you know.
I don't find the option to enable 8 chanel pcm decoding for hd audio. Is this automatically done if bitstreaming is disabled?
(2015-12-23, 10:40)Caramba69 Wrote: I don't find the option to enable 8 chanel pcm decoding for hd audio. Is this automatically done if bitstreaming is disabled?

Automatically enabled on Pi2, disabled on Pi1.
Bitstreaming takes precedence though, so disable DTS passthrough if you want 8 channel audio.
(2015-12-23, 10:18)doldi Wrote: Hello,
with #1222 i notice stuttering video and artefacts at beginnig to start the "Tina" DVD.
Buffering is also little higher than build in build 1221.

Sample _Test.7z is on GoogleDrive, you know.

Here is a little Video, recorded from screen with my Microsoft WebCam about Buffering and pixeld video (12MB):

Problemvideo
(2015-12-22, 18:39)polo_joe Wrote:
(2015-12-21, 23:32)popcornmix Wrote:
(2015-12-21, 21:50)polo_joe Wrote: I'm still having artifacts when watching hd livetv channels with advanced deinterlace. Even if I watch 720p channels there are artifacts.
Any news on this front?

I did add some settings you could try tweaking so see if they help. Use this firmware:
https://dl.dropboxusercontent.com/u/3669...re_arb.zip

(Use start_x.elf and fixup_x.dat and replace your OpenELEC start.elf/fixup.dat).

This has a new command you can run from ssh. It allows setting the priority of different blocks on the Pi.
Code:
$ vcgencmd arbiter
usage: arbiter [ status | algorithm <arbiter> <credit|priority|hybrid|delayed> | set <block> <priority> ]
$ vcgencmd arbiter status
SYSTEM_UNCACHED: alg:hybrid
      SYSTEM_L2: alg:hybrid
     SYSTEM_SRC: alg:hybrid
    SYSTEM_PERI: alg:credit
     SYSTEM_DMA: alg:credit
   VPU_UNCACHED: alg:hybrid
         VPU_L2: alg:hybrid
            CAM: alg:credit
    HVS: pri:12 panic:15
    V3D: pri: 0 panic: 0
   H264: pri: 8 panic: 0
   JPEG: pri: 0 panic: 0
   TRAN: pri: 2 panic: 3
    ISP: pri: 4 panic: 0
    USB: pri:13 panic: 0
ARM_UC: pri:10 panic: 0
ARM_L2: pri: 5 panic: 0
ARM_PER: pri: 8 panic: 0

I think the first test is to make HVS low priority and see if that helps.
Code:
vcgencmd arbiter set hvs 0 0

You might also want to set the ARM to be high priority
Code:
vcgencmd arbiter set arm_uc 15 15
vcgencmd arbiter set arm_l2 15 15
vcgencmd arbiter set arm_per 15 15

Or low priority
Code:
vcgencmd arbiter set arm_uc 0 0
vcgencmd arbiter set arm_l2 0 0
vcgencmd arbiter set arm_per 0 0

USB can't be changed when running, so you can't change that.
V3D should have an effect, but we have experimented with that before.

this setting works best for me
http://sprunge.us/fJVa

but still some rare artifacts

hvs 0 0 and or arm 0 0 are even worse

green artifacts are still there when watching 720p channels
could this be a usb problem and no deinterlace problem?
today I tried setting usb to 1.1 with dwc_otg.speed=1 but my tv card doesn't support usb 1.1 so I can't test if it would improve anything.
New OpenELEC K* build #1223: RPi / RPi2
(Supercedes previous build)

Code:
# uname -a
Linux rpi512 4.4.0-rc6 #1 Wed Dec 23 21:11:57 GMT 2015 armv6l GNU/Linux

# vcgencmd version
Dec  8 2015 14:44:44
Copyright (c) 2012 Broadcom
version e591b5eb05e2cdb1b5ae25512b27d33127d7bee9 (clean) (release)

# lsb_release
OpenELEC (Milhouse) - Version: devel-20151223211105-#1223-g7081584 [Build #1223]

# vcdbg log msg 2>&1 | grep DTOK
001696.877: Kernel trailer DTOK property says yes

# Kernel device tree status: Enabled

Based on tip of OpenELEC master (70815845, changelog) and tip of XBMC master (cc7040ab, changelog) with the following modifications: Build Highlights:
  1. PVR and dcadec changes
Build Details:
  1. OpenELEC:
    • Roll nvidia-legacy driver into Generic (PR:4486, 6 commits, 12 files changed)
  2. XBMC:
    • Uniformize code: empty() over size() compare, pragma... (PR:8429, 3 commits, 168 files changed)
    • [osx] Ignore Makefile after platform shuffle (PR:8650, 1 commit, 1 file changed)
    • [PVR] wip: handle disabled timers/readonly timers better in the GUI (PR:8653, 2 commits, 15 files changed)
    • Revert "[PVR] wip: handle disabled timers/readonly timers better in the GUI" (PR:8656, 1 commit, 15 files changed)
    • Fix Playlist.Getitems returning empty artist array (PR:8654, 1 commit, 2 files changed)
  3. dcadec:
    • Use memset/memcpy for ADPCM history. (7ad6629f)
    • Shift and clip XLL output in single pass. (4ebfc929)
    • Use saturation instruction for clipping on ARM. (ba2989a9)
    • Factor out XLL downmix into separate function. (e2ad1b95)
    • Rename structure field. (2dc17356)
    • Fix doubling sampling frequency during sync loss. (5653a436)
  4. pvr.hts:
    • Fix UpdateTimer to allow enable/disable of timers created by timerecs and autorecs (PR:152, 2 commits, 2 files changed)
  5. newclock5:
    • Commits no longer in build:
      • squash me (04094ae9)
      • squash: allow sync adjustment with passthrough and pll adjust (f69bf6c1)
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.
Has lastest librtmp.so.1 from dec 2015?
Thanks, great work.
(2015-12-23, 17:08)doldi Wrote: Here is a little Video, recorded from screen with my Microsoft WebCam about Buffering and pixeld video (12MB):

Problemvideo

Can you provide a debug log showing the corruption.
There is one reason where corruption can occur when the video stream is considered stalled and reset - I'd like to know if that is the cause.
(2015-12-24, 06:13)goibe Wrote: Has lastest librtmp.so.1 from dec 2015?

No but I've just updated it. Currently untested (except a compile test) so please report if it work okay in tonight's build.
hi popcornmix

Well, have uploaded

Tzz, the problem is not so strong @#1223, if Debug Log is aktivated.
But KodiCPU is over 100% if video begins.

Merry Christmas!
Harald
@doldi can you try again with sync method set to "resample audio" rather than PLL adjust.
  • 1
  • 18
  • 19
  • 20(current)
  • 21
  • 22
  • 146

Logout Mark Read Team Forum Stats Members Help
OpenELEC Testbuilds for RaspberryPi (Kodi 17.0)6