Posts by jernej

    Better don't compare it to Rockchip, Amlogic or Allwinner - they needed 4+ years and are partly not even there yet

    I tend to disagree (except about AML part) :) AW and RK support deinterlacing while RPi does not. RK is also almost on pair regarding HDR. In short, each platform has features that are not necessarily implemented on others. But I agree that RPi development progresses much faster.

    offbeat Thanks for detailed analysis. Enabling performance governor on H6 is not a good idea because H6 generates a lot of heat even on lower frequencies (many are not happy with that). You can of course enable it locally.

    There is a reason why Allwinner port is marked as "stable beta" in LE10 release announcements. There is plenty of corner cases which needs to be fixed, but I'm actually quiet happy that it works as well as it does. Decoding using request API in ffmpeg is known to be suboptimal. We didn't rewrite it just yet, because kernel will in near future receive some uAPI updates which would allow even more efficient decoding. As mentioned before, GPU frequency switching doesn't work on H6 for some unknown reason, which nobody can figure out. However, maybe default frequency can be raised same way it is done for A64 (via DT)? This would be acceptable until dynamic switching is figured out.

    Note that I don't know for any peripheral which would rely on CPU clock directly. You can always check how clocks are related and configured in /sys/kernel/debug/clk/clk_summary. VPU, DRAM and GPU have their own, dedicated PLLs, which makes them independent from rest of the system.

    P.S. Testing it was a pain, the way ffmpeg is packaged in LibreElec is sooo confusing, binary programs aren't included in base build at all and ffmpeg-tools addon uses completely different build options, without any hardware decoding patches...

    I actually use separate distro (Arch for ARM) when developing ffmpeg/Cedrus or other, mostly display related features. I apply all relevant LE patches to kernel and then netboot it. This allows much quicker turnaround time. However, I test Kodi only with LE.

    I have one more question. I see there are settings for buttons repeat time in CEC adaptor. Should they be set to the corresponding brand's TV remote model, or not? I am just trying to figure out why I have problems with CEC on my old Pioneer Kuro plasma TV. I've got multiple CEC100 errors when CEC adaptor (in Kodi) and HDMI control (in TV) are enabled (see the attachment).

    It's first time I see this kind of error. But then again, CEC behaves differently on different devices. I have experience only with my LG B8 TV, where CEC behaves normal. Just experiment...

    offbeat Reason why OSD helps is probably because EGL provides vsync... I don't know. Maybe windowing/gbm: Add display clock by popcornmix · Pull Request #19601 · xbmc/xbmc · GitHub will help. IDK.

    First of all, there is a BIG difference between my scenario and his. I'm not talking about iptv so connection bandwidth isn't an issue for sure. I'm talking about DVB-T broadcasted contents.

    That difference is not really important. Check test image I posted and see if dmesg is flooded with same type of lines levitsky86 reported. If it is, stream gets corrupted and decoder for some reason can't cope with such corrupted stream.

    Do I need to open the box to physically see the components or can it be done without disassembling? I have ssh access in both.

    Opening the box is the easiest way. dmesg output from Android would probably tell what storage is used.

    Fortunately, I also have OrangePi 3 :) I tested with fresh installation and overlays work. Try to flash latest nightly image to spare card and see if it changes anything.

    For reference, my extlinux.conf:

    Code
    LABEL LibreELEC
      LINUX /KERNEL
      FDT /sun50i-h6-orangepi-3.dtb
      FDTOVERLAYS /overlays/sun50i-h6-i2c0.dtbo /overlays/sun50i-h6-i2c2.dtbo
      APPEND boot=UUID=2204-3027 disk=UUID=c1742531-5dac-434c-8390-f1decb06f821 quiet console=ttyS0,115200 console=tty1

    NOTE: Your UUIDs will be different.

    If that doesn't work, then I have no idea.

    Can you please specify in what sense isn't correct?

    I have no clue, but it obviously doesn't work.

    I wrote two overlays for I2C0 and I2C2 like so http://ix.io/2WSc (and same for I2C2). Then I included them to extlinux.conf like so FDTOVERLAYS /overlays/sun50i-h6-i2c0.dtbo /overlays/sun50i-h6-i2c2.dtbo. Now I have:

    Code
    LibreELEC:~ # i2cdetect -l
    i2c-3   i2c             DesignWare HDMI                         I2C adapter
    i2c-1   i2c             mv64xxx_i2c adapter                     I2C adapter
    i2c-2   i2c             mv64xxx_i2c adapter                     I2C adapter
    i2c-0   i2c             mv64xxx_i2c adapter                     I2C adapter

    Note - this is on PineH64 model B, but I don't see any reason why it wouldn't work on OPi3.

    Binary version of overlays: Index of /overlays/

    well, there you have it - overlay you're using isn't correct - it shouldn't freeze during boot and you should get additional i2c bus listed. What is wrong with it - I don't know. From what I can see, seems good.

    There is no other i2c port on the GPIO header

    All three are there, just pins are named differently. I2C1 is on SPI1-MOSI and SPI1-MISO, I2C2 is on UART3-RX and UART3-TX and they lack pull ups, so yeah, you most probably use I2C0.

    BTW, 0x36 is PMIC address. You are probing wrong bus. Make sure you have now 3 buses and probe all except HDMI.

    Please explain wether I can continue to install nightly updates on that image you made with some output, or should I better format and do a clean install?

    The only difference between my test image and nightly from that day are just two lines of debug output. You can treat them same. Clean install is recommended every now and then, when bootloader is updated.

    Monoscopic mode not working? Although I select it, it has no effect.

    I'm not even sure what it does. Something with 3D mode? If so, that's completely untested and unlikely it will get better anytime soon. I have no 3D display nor do I plan to obtain it.

    In other words it's very like as the hardware is not capable to hold that stress load during such moments...

    Nope, that's not it. In short, driver is fed by invalid data, which cause instability. Why invalid data is sent is another story, but anyway, driver should react properly on such frames. Fixing that is not straightforward due to some HW quirks. Be patient, it's my intention to solve it, but I can't give you any timeline.