Jump to content

jernej

Members
  • Posts

    1144
  • Joined

  • Last visited

Everything posted by jernej

  1. You shouldn't unload I2C driver because power regulator (PMIC) is controlled over it. With I2C driver unloading you effectively disable power regulation.
  2. A bit of HW details. H3 and most of AW SoCs for that matter have RTC integrated in which can be clocked from internal or external clock. Internal clock is sadly inaccurate, so it's borderline useable for keeping time, but some boards, like NanoPi NEO still omit external clock (price consideration, I guess). So, main issue about RTC to keep time is if it has separate power supply which is still enabled when main power supply is disconnected. On schematic, this is VCC_RTC. As you can see on NanoPi NEO schematic, RTC and main power supply are connected together, so RTC will lost power and with that time every time main power supply is disconnected.
  3. Mali is only for rendering, video decoding is done on separate HW core. However, its driver is not finished, VAAPI lib for it is also not finished yet and most of all, firefox supports HW video decoding (e.g. VAAPI) only on wayland. So, nothing that you could use right now. If you want decent video player on OPi Plus2E you can use LibreELEC image (Kodi) which has all work in progress included, but that is probably not something you had in mind (multimedia center app with no browser but youtube addon can be installed).
  4. jernej

    Mainline VPU

    I don't. As you said, I'm solely focused on decoding. Maybe @Kwiboo knows?
  5. jernej

    Mainline VPU

    That is very much work in progress, so it doesn't work anywhere yet, except basic MPEG4 on Cedrus. I'm not sure if I'll ever finish MPEG4 because it has a ton of variants and even Cortex A7 is capable decoding it. On the other hand, VC-1 has only 3 profiles and it's much more demanding for CPU decoding.
  6. jernej

    Mainline VPU

    You have to make sure that libva-v4l2-request uses same *-ctrl.h files that are in kernel sources under include/media/. For starters, just copy them over from kernel to lib source folder. Then you might need to adjust lib sources to new field names. Unless these files are completely the same, you won't have any success.
  7. A20 has only 2 cores but driver supports up to 4. Perfectly normal.
  8. jernej

    Mainline VPU

    FYI, VAAPI lib is just userspace translation layer between two apis. No kernel hacking needed, apart from applying LE patches.
  9. jernej

    Mainline VPU

    That is the plan. However that can't happen until codecs interfaces in kernel are promoted to stable. That will probably take at least a few more kernel cycles. Nobody wants to screw it up user facing API so everyone is cautious. Actually, not all features are yet supported in kernel, that's also the reason you need kernel patches with experimental improvements. Work on initial vaapi library stalled, but others might take over... Nothing useful at the moment unless you understand both APIs well and you're able to fix it.
  10. jernej

    Mainline VPU

    This can be avoided by using -hwaccel_output_format drm_prime argument. Works good with null output.
  11. Not true. I maintain Allwinner port of LibreELEC and I must say that I'm pleased with H6 multimedia performance. You can try images here. Note that GPU doesn't provide any video decoding. There is VPU for that, but browsers on Linux lag a bit with integration of HW accelerated decoding. On top of that, VPU drivers are wip, so properly put whole video stack together is a bit challenging at the moment. That's why pre-built LibreELEC images exist in the first place.
  12. Simple, LibreELEC heavily patches kernel and ffmpeg for better experience. Patches are slowly getting in kernel but ffmpeg patches will need far more time to be accepted (kernel must first declare this API stable which won't happen anytime soon). You also need Kodi master (e.g. Kodi 19 pre-alpha), because it has a lot of important improvements. Last but not least, only Kodi GBM (e.g no desktop environment) version really works well (zero copy playback) and soon wayland. X11 has no chances to work well. TL;DR: LibreELEC uses features which are in development and not present in any stable version yet. BTW, also 4K HEVC videos work well.
  13. @Myy small advice - make links to https://github.com/LibreELEC/LibreELEC.tv because I often delete branches which were already merged. linux56 branch will be removed soon.
  14. No, they are disabled because nobody took time to properly enable them in board DTS until now (it will be in 5.7). No, sun4i-drm is for all Allwinner SoCs, it supports Display Engine 1.0, 2.0 and 3.0. It's just named by Linux tradition where it takes name of first (oldest) supported platform.
  15. It's completely different SoC which means that your boot attempt probably doesn't pass RAM initialization, if it even gets there. So no, you'll not see anything on the screen. I wouldn't count on UART either, since there is no guarantee it's on the same base address as on H6. Also UART related clocks may be different.
  16. @zanfi just to clear any confusion - there are multiple Tanix TX6 variants: - Tanix TX6 - Allwinner H6 - Tanix TX6 mini - Allwinner H6 - Tanix TX6s - Allwinner H616 As you can see, Tanix TX6s won't work with any image you can find, because it's using new SoC which is not supported anywhere. There are no public SDK and datasheet releases (that I know) for it, so you can't find any images besides Android from box manufacturer.
  17. If you have Pinebook in mind and the one from Olimex, they both use anx6345 IIRC. Driver for that chip was merged in kernel some time ago, but it doesn't consider "panel-supply" property. I guess that's what's wrong. @Icenowy?
  18. True, but port A and B are routed internally to AC200 co-packaged IC and not useful for outside world. Information about port A and B is not even in manual.
  19. Take a look into GPIO controller register description. You will find interrupt settings only for port F, G, H, L and M.
  20. You should send patch for that to U-Boot so next person which tries to use USB on OrangePi Win don't need to do that work again.
  21. I know, but now and then there are U-Boot changes which influence Linux. For example, I figured out that enabled display driver in U-Boot occasionally prevents Linux driver to read EDID. I could as well add unconditional HDMI reset in kernel, but then display would blink. Eventually, we will also add support for DT overlays in extlinux.conf which needs patched U-Boot, at least at first, until changes get merged upstream. Because of such things I like to keep U-Boot under control in official LE images. I have nothing against community builds which change this behaviour and enable multi-boot systems, but this is can of worms if not done correctly. I don't want to support that as official LE feature at this point.
  22. Allwinner SoCs can't boot from USB natively. So you must have SD card or eMMC to boot from anyway. But then it doesn't really matter if you load kernel and base filesystem from SD, eMMC or USB, does it? Once you booted Linux, you can use USB sticks as additional storage device.
  23. Well, there is another issue. U-Boot detected spurious key press on some keyboards, which were plugged in during boot. This caused abort of boot and clueless user thought that something is broken (there was nothing on screen). Such reports happened several times already. Besides you, nobody complained that USB support in U-Boot is missing. I think most users don't care about U-Boot, they just want that LE is booted either from SD or eMMC.
  24. I disabled USB in U-Boot because it considerably extends boot time (up to 5 seconds on some boards): https://github.com/LibreELEC/LibreELEC.tv/blob/master/projects/Allwinner/bootloader/config
×
×
  • Create New...

Important Information

Terms of Use - Privacy Policy - Guidelines