• 1
  • 90
  • 91
  • 92(current)
  • 93
  • 94
  • 111
OpenELEC Testbuilds for RaspberryPi (Kodi 15.0) Part 1
@Milhouse for info

new ARM NEON optimized HEVC epel : https://ffmpeg.org/pipermail/ffmpeg-deve...70781.html

If you want to test with your build
(2015-03-27, 21:54)fab67 Wrote: @Milhouse for info

new ARM NEON optimized HEVC epel : https://ffmpeg.org/pipermail/ffmpeg-deve...70781.html

We should already have that:
https://github.com/popcornmix/xbmc/commi...8c3a858c5e
Oh yes right, I thought that was a new one
(2015-03-27, 22:58)fab67 Wrote: Oh yes right, I thought that was a new one

I think it's new to ffmpeg. I grabbed it from openhevc where the patches seem to appear first.
Keep an eye out for any missing from newclock4 though (or any that have been updated).
New OpenELEC Isengard build #0327: RPi / RPi2
(Supercedes previous build)

Code:
# uname -a
Linux rpi512 3.19.3 #1 Fri Mar 27 21:56:45 GMT 2015 armv6l GNU/Linux

# vcgencmd version
Mar 27 2015 18:48:55
Copyright (c) 2012 Broadcom
version b2109b1c1c8ae8717df455cfaf312a6c1801f45a (clean) (release)

# lsb_release
OpenELEC (Milhouse) - Version: devel-20150327214009-#0327-g2886fa0 [Build #0327]

# vcdbg log msg 2>&1 | grep DTOK
001651.654: Kernel trailer DTOK property says yes

# Kernel device tree status: Enabled

Based on tip of OpenELEC master (2886fa0b, changelog) and tip of XBMC master (3d02a51b, changelog) with the following modifications: Build Highlights:
  1. New firmware, with fixes for reverse seek and VC-1 stutter (thanks @popcornmix/@deborahc)
  2. New 3.19.3 kernel
Build Details:
  1. Firmware (Mar 27):
    • firmware: video codec: fix regressions on VC-1 decode See: link
    • firmware: video_codec: fix stall in categoriser when flushing outside seq. See: link
    • firmware: di_adv: Tidy up naming of prev/cur/next images
  2. OpenELEC:
    • projects/WeTek_Play: Remove support for CVBS output, because it causes problems for HDMI (PR:4047, 1 commit, 2 files changed)
    • linux: update to linux-3.19.3 (b5738cd0)
    • gdb: update to gdb-7.9 (04dc11ba)
    • libinput: update to libinput-0.13.0 (b6b7bea6)
    • systemd: disable IFLA_BRPORT_UNICAST_FLOOD (ecf362b5)
  3. XBMC:
    • Fix compile warnings + problem in URL.cpp (PR:6791, 2 commits, 19 files changed)
    • [README.ubuntu] Enables PPA sources with -s key (PR:6821, 1 commit, 1 file changed)
  4. newclock4:
    • New commits in this build:
      • playback an error sound if window switch is refused (ea7cd244)
      • don't switch to another window if there are active modal dialogs (00c3e441)
  5. kernel 3.19.y:
    • New commits in this build:
      • dmaengine: bcm2708: Remove chancnt affectations (347884f6)
      • dmaengine: bcm2708: do device registration in the board file (42e6ec45)
      • dmaengine: bcm2708: don't restrict DT support to ARCH_BCM2835 (bec2ee1b)
      • BCM270x_DT: add bcm2835-dma entry (4e1b441c)
      • bcm270x: add mmc-bcm2835 clock (1a309e51)
      • mmc: bcm2835-mmc: enable DT support for all architectures (d9443f3c)
      • BCM270x_DT: add bcm2835-mmc entry (efa7d263)
      • mmc: bcm2835-mmc: fix probe error handling (e39ad97a)
      • dmaengine: bcm2708: overwrite dreq only if it is not set (0b8c424a)
      • config: Add OABI_COMPAT to fix gdb illegal instructions (4ef70ab8)
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.
Good job! VC1 1080p goes smoothly. Also gone the problem with progressive, as described here http://forum.kodi.tv/showthread.php?tid=...pid1956784 . Dark City remux now played without stuttering and flickering. Interlace still gives artifacts.
Playback/rewind much better for me as well. Was surprised to hear that deinterlacing is bob for 1080 despite the setting being mmal-advanced which I saw in a previous post applies to a lower resolution.

With the pi2 hardware (video chip being the same as original pi), will anything other than bob be possible for hd resolutions?
(2015-03-28, 16:18)zaphod24 Wrote: With the pi2 hardware (video chip being the same as original pi), will anything other than bob be possible for hd resolutions?

The advanced algorithm reads 3 fields of input and writes two frames of output for each video frame. That is just too much work for 1080p.
The bob currently does a linear interpolate for the duplicated pixels. We could probably handle a cubic interpolate which would improve sharpness a touch.

If anyone can suggest a deinterlace algorithm that only requires a single field, or perhaps a pair of fields from a single frame, then that may be feasible.
And how about the upscaling? I mean lanczos, spline. Or Video Core is too weak for this?
(2015-03-28, 16:52)MONSTA Wrote: And how about the upscaling? I mean lanczos, spline. Or Video Core is too weak for this?

Pi has high quality cubic scaling (Using the Mitchell Netravali kernel) for video.
This is supported in hardware so doesn't have a cpu cost.
Really? By default? Then super. Thank you.
Good info! If cubic interpolation is included in one of the builds I'd be glad to see if that gets rid of or minimizes the flickering/artifacts I'm seeing when watching 1080i60 live tv.
I'm trying to get the DVB-T tuner working on my brother's RPI B. I've got the identical one on my RPI B and it works fine but I just get page timeout when I try and open http://192.168.1.83:9981 on his PC and the dmesg shows it's detecting different hardware for him for some reason, 'ITE 9135 Generic' for him, 'Kworld UB499-2T T09' for me

We're both using the same remote receiver as well, the main difference is that I installed a Belkin USB Hub in his case, whilst I'm running with just the two built-in USB ports. The hub is cascaded and I found some issues with using the second part of the cascade, so I blocked off those ports with the case and he can only plug anything into the two ports attached to the first part of it. Even if he could make do with the RPI's two ports for just remote receiver and USB tuner, which isn't really practical as he needs at least one more for a USB stick for storage, it will be impossible for him to plug the tuner directly into the RPi as there's not room in the case.

He was on a late February build, whilst I'm on 3 March, so I updated him to the latest #0327 but the dmesg still shows the same thing. I updated to #0327 and it still works fine here, although it did disable the PVR client and I had to re-enable that.

My #0303 dmesg: http://pastebin.com/DtUJTZGR

My #0327 dmesg: http://pastebin.com/xSP48upe

His #0327 dmesg: http://pastebin.com/1ikk3UiB

My #0303 kodi.log: http://pastebin.com/zK9x07Dg

My #0327 kodi.log: http://pastebin.com/6tfKp0YP

His #0327 kodi.log: http://pastebin.com/A0mGFgr6

His STB Freeview tuner is very much on it's last legs, so if I could get this working for him it would be great.
@doveman2: Usual thing, you'll need to identify the last working build for your brother, presumably somewhere between late February and early March.
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.
I take it auto source resolution output (i.e. 1080i) is not possible like I asked about before? This would allow my TV or AVR to deinterlace which sounds like it would be better than what can be done on the Pi?
  • 1
  • 90
  • 91
  • 92(current)
  • 93
  • 94
  • 111

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