• 1
  • 50
  • 51
  • 52(current)
  • 53
  • 54
  • 111
OpenELEC Testbuilds for RaspberryPi (Kodi 15.0) Part 1
(2015-02-25, 17:17)slack3r Wrote: I have found that there is a big difference in GUI rendering between build #0104 and #0106.

Is this the case when "limit GUI updates when playing videos" is disabled in video/acceleration settings?

#104 brought in "dirty rectangles for video" which broke "limit GUI updates" until #125.
With "limit GUI updates" disabled, you should be able to compare more fairly.
Another thing I've noticed with DVDPlayer is that when you access the OSD for the first time (= it's loaded in memory), you get one or two dropped frames. This is confirmed by the codec info overlay. The same happens when you then access the audio OSD settings dialog, video OSD, bookmark dialog etc (every dialog accessible from the OSD). Once these windows are loaded in memory no single frames is dropped, even for a different video (https://github.com/xbmc/xbmc/blob/master...se.cpp#L76).
I'm not able to reproduce every time, but pretty frequently (85%?). Of course you need to restart Kodi to see this. I'm wondering if all these could be loaded on startup (it's a media center...).
(2015-02-25, 18:03)menakite Wrote: Another thing I've noticed with DVDPlayer is that when you access the OSD for the first time (= it's loaded in memory), you get one or two dropped frames. This is confirmed by the codec info overlay. The same happens when you then access the audio OSD settings dialog, video OSD, bookmark dialog etc (every dialog accessible from the OSD). Once these windows are loaded in memory no single frames is dropped, even for a different video (https://github.com/xbmc/xbmc/blob/master...se.cpp#L76).
I'm not able to reproduce every time, but pretty frequently (85%?). Of course you need to restart Kodi to see this. I'm wondering if all these could be loaded on startup (it's a media center...).

Yes, agreed that could be useful. Suggest a patch that loads the OSD on startup and we'll try it in the test builds (and it may be sensible for pushing upstream).
About mysql compression.

I use mysql over vpn, and have 2-3 ms to the mysql server, Both server and client is pretty fast.

Three runquerys from kodi without compression. 469, 439, 429 and 433.
Three with compression, 159, 153, 149

If I use mysql client and do select * from movieview ; I get 0.29 without and 0.09 with compression.
(2015-02-25, 17:49)popcornmix Wrote:
(2015-02-25, 17:17)slack3r Wrote: I have found that there is a big difference in GUI rendering between build #0104 and #0106.

Is this the case when "limit GUI updates when playing videos" is disabled in video/acceleration settings?

#104 brought in "dirty rectangles for video" which broke "limit GUI updates" until #125.
With "limit GUI updates" disabled, you should be able to compare more fairly.

No, Is not: GUI updates are limited to 10fps. But disabling it I get the same: build #0106 runs at 30fps, #0104 at 60fps.
(2015-02-25, 18:06)popcornmix Wrote: Yes, agreed that could be useful. Suggest a patch that loads the OSD on startup and we'll try it in the test builds (and it may be sensible for pushing upstream).
Check https://github.com/anaconda/xbmc/commit/...505a37d6cf - it's completely untested as of now.
I have changed the loading type of every dialog I could find in https://github.com/xbmc/xbmc/blob/master...deoOSD.xml - I don't have PVR but I guess the 2 PVR dialogs behave the same.

Also, I meant skipped frames, not dropped.
(2015-02-25, 19:33)menakite Wrote: Check https://github.com/anaconda/xbmc/commit/...505a37d6cf - it's completely untested as of now.

Looks good. Get this on startup:
Code:
18:27:50 4647.194336 T:1980153856    INFO: Loading skin file: VideoOSD.xml, load type: LOAD_ON_GUI_INIT
18:27:50 4647.257324 T:1980153856   DEBUG: Load VideoOSD.xml: 62.40ms

and when bringing up OSD

Code:
18:30:27 4804.357422 T:1980153856   DEBUG: ------ Window Init (VideoOSD.xml) ------
18:30:27 4804.357910 T:1980153856   DEBUG: Window VideoOSD.xml was already loaded
(2015-02-25, 03:10)Milhouse Wrote: New OpenELEC I****** build #0224: RPi / RPi2
(Supercedes previous build)


Build Details:

[*]newclock4:
  • New commits in this build:
  • Commits no longer in build:
    • native resolution (615486f1)
    • allow to disable native resolution switching through GUI (4cadac04)

Does this mean there is no native resolution (no upscaling by raspberry but let the receiver with upscaling processor do a better job) in this build anymore? Is it upstream or did it go away?
(2015-02-25, 22:02)tdn135 Wrote: Does this mean there is no native resolution (no upscaling by raspberry but let the receiver with upscaling processor do a better job) in this build anymore? Is it upstream or did it go away?

It's been dropped as being buggy and not looking likely to be fixed up soon. Info on current state here.
New OpenELEC I****** build #0225: RPi / RPi2
(Supercedes previous build)

Code:
# uname -a
Linux rpi512 3.19.0 #1 Wed Feb 25 21:02:16 GMT 2015 armv6l GNU/Linux

# vcgencmd version
Feb 24 2015 12:50:21
Copyright (c) 2012 Broadcom
version 784e0c6a418a9f1e07c0531468162e968acdeefd (clean) (release)

# lsb_release
OpenELEC (Milhouse) - Version: devel-20150225210128-#0225-g24ed310 [Build #0225]

# vcdbg log msg 2>&1 | grep DTOK
001528.881: Kernel trailer DTOK property says yes

# Kernel device tree status: Enabled

Based on tip of OpenELEC master (24ed3102, changelog) and tip of XBMC master (7db3ab08, changelog) with the following modifications: Build Highlights:
  1. Load OSD dialogs on startup (avoid skipped frames)
Build Details:
  1. XBMC:
    • [smbdirectory] mark dot files and folders hidden (fixes #15680) (PR:6529, 1 commit, 1 file changed)
    • [PVR] fix deleted recordings view (PR:6525, 1 commit, 4 files changed)
    • [gui] fix activating of context menu after skin reload while context menu was active (PR:6540, 1 commit, 1 file changed)
    • [3D] don't back out from 3D handling on resolution changes (PR:6235, 1 commit, 1 file changed)
    • [xcode] sort PVRActionListener below existing folders in /pvr (PR:6544, 1 commit, 1 file changed)
  2. newclock4:
    • New commits in this build:
      • Load OSD dialogs on startup. (f5e2c71b)
      • dvdplayer: sync audio after having issued GENERAL_RESYNC (03838299)
      • - force re-create of ffmpeg resampler after error (6398bbad)
      • dvdplayer: flush audio errors on large sync error (42ae5531)
    • Commits no longer in build:
      • [jenkins/addons] - fix error detection during addons build (c7d51417)
      • FIXUP: fix python PIL after #5218 (d2516a23)
      • [win32] - fix DirectSound for streams with very low sample rates (2a82a7dd)
      • [pvr] Fix compiler warning in CGUIWindowPVRRecordings (e7bc5e16)
      • [configure] make sure the gcc-4.9 SSE workaround is only enabled on x86 (14859cf7)
      • [linux] prepend -Wall so it doesn't override flags from environment (e9d8e865)
      • [videoinfoscanner] reset library bools only once - fixes #15805 (a7fbcf8d)
      • dvdplayer: reset error integral when flushing sync errors (ea702816)
      • dvdplayer: fix missing audio info (0351e726)
      • cmake: fix debug builds by passing CMAKE_BUILD_TYPE to the addon's build (620f585b)
      • Fix AnnouncementManager::Announce(). Announcers may be removed while iterating announcers vector. (2b8f2c0e)
      • dvdplayer: consider audio stalled as long as message queue is empty (659e038b)
      • dvdplayer: sync av players after unsynced seeks which may occur when rw/ff (0045f392)
      • addons: no need to retrieve general.addonforeignfilter in a block that is only called if its value is true (27340bb2)
      • [win32] addons: fix filtering foreign language addons (fbb05f71)
      • simplify and document CLangInfo::GetLanguageLocale() (70e835b3)
      • [seek] fix crash if no seek steps are found for the selected direction (50107914)
      • json-rpc: fix duplicate "lastplayed" in "Video.Fields.TVShow" (thanks Tolriq) (2ee10fcb)
      • [ActiveAE] - ActiveAESink: Change several LOGNOTICE to LOGDEBUG to reduce kodi.log "spamming". (70cfe10f)
      • don't back out from 3D handling on resolution changes if we're switching to a different 3D mode (11c79665)
      • Revert "don't back out from 3D handling on resolution changes if we're switching to a different 3D mode" (de7e365c)
      • [graphics] Don't set stereo mode based on resolution (4bade805)
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.
(2015-02-25, 18:35)marantz Wrote: About mysql compression.

I use mysql over vpn, and have 2-3 ms to the mysql server, Both server and client is pretty fast.

Three runquerys from kodi without compression. 469, 439, 429 and 433.
Three with compression, 159, 153, 149

If I use mysql client and do select * from movieview ; I get 0.29 without and 0.09 with compression.

Thanks.

Note that starting with #0224 MySQL compression is disabled by default, but can be re-enabled by adding "<compression>on</compression>" to your <videodatabase>/<musicdatabase> tags in advancedsettings.xml.

MySQL compression looks like it could be be a significant benefit on WiFi connections, but is probably going to have a negative impact over a decent wired connection.
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.
Hi guys. First post, but long time admirer of Milhouse builds... Big Grin

Plenty of

Code:
unexpected IRQ trap at vector 00

in dmesg outputs of RPi builds #0225 and #0224.

No such errors in dmesg output of RPi build #0223 or in RPi2 dmesg outputs of any build mentioned here.
(2015-02-26, 16:09)FauDrei Wrote:
Code:
unexpected IRQ trap at vector 00

in dmesg outputs of RPi builds #0225 and #0224.

Yes, I spotted that. Should be fixed in next build:
https://github.com/raspberrypi/linux/com...3a88af0406
(2015-02-25, 20:32)popcornmix Wrote: Looks good.
Thanks for the quick feedback. Tested this too, no skipped frames the first time these dialogs are accessed, except the bookmarks one (always 1 frame). I'll investigate this - will ping you in case the commit in newclock4 needs an update.
(2015-02-26, 16:50)popcornmix Wrote: Yes, I spotted that. Should be fixed in next build:
https://github.com/raspberrypi/linux/com...3a88af0406

That was quick. Wink Thanks.
  • 1
  • 50
  • 51
  • 52(current)
  • 53
  • 54
  • 111

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