jock Posted October 19, 2021 Author Share Posted October 19, 2021 @marshall Thanks for the 32 bytes of the image, but actually I meant the first 32 megabytes eheheh . No problem however, I managed to extract the original dtb from the tvbox image. I took a look into and yes, there is the need to adjust the secondary led and something seems to be a bit different for wifi too. I will craft a special overlay so you can test the whole thing. 0 Quote Link to comment Share on other sites More sharing options...
jock Posted October 19, 2021 Author Share Posted October 19, 2021 @paradigman I don't what to say, that's just a problem that has been rarely seen before and I don't have such a problematic board to investigate further. I see that your board has "R29-MXQ" marking, but we have only seen "R28-MXQ" which is indeed different. I may suggest you, if you didn't already, to try an image with the mainline kernel. 0 Quote Link to comment Share on other sites More sharing options...
marshall Posted October 19, 2021 Share Posted October 19, 2021 2 hours ago, jock said: @marshall Thanks for the 32 bytes of the image, but actually I meant the first 32 megabytes eheheh . No problem however, I managed to extract the original dtb from the tvbox image. I took a look into and yes, there is the need to adjust the secondary led and something seems to be a bit different for wifi too. I will craft a special overlay so you can test the whole thing. Oh I missed the **mega**bytes part... oof That's awesome though, I'll wait for the modifications and will test it on the board Thanks! 0 Quote Link to comment Share on other sites More sharing options...
jock Posted October 19, 2021 Author Share Posted October 19, 2021 @marshall Ok, this is the first attempt. Led configuration looks similar to led-conf5 you were using, but I don't understand why you had problems. Anyway I fixed a minor thing and switched leds. The DTB says there is a red and a blue led: working is now the red led and configured by default to stay always on; auxiliary is the blue led and is configured to react on access to internal flash. dtb may tell the wrong labels by the way. About wifi, I followed the original dtb, swapped some ACTIVE_LOW <-> ACTIVE_HIGH gpios and brought more power to them, reduced bus frequency to 37.5Mhz. I hope that now it works! Put the file in /boot/dtb/overlays directory and change led-conf5 with led-conf6 in /boot/armbianEnv.txt rk322x-led-conf6.dtbo 0 Quote Link to comment Share on other sites More sharing options...
marshall Posted October 19, 2021 Share Posted October 19, 2021 (edited) 1 hour ago, jock said: @marshall Ok, this is the first attempt. Led configuration looks similar to led-conf5 you were using, but I don't understand why you had problems. Anyway I fixed a minor thing and switched leds. The DTB says there is a red and a blue led: working is now the red led and configured by default to stay always on; auxiliary is the blue led and is configured to react on access to internal flash. dtb may tell the wrong labels by the way. About wifi, I followed the original dtb, swapped some ACTIVE_LOW <-> ACTIVE_HIGH gpios and brought more power to them, reduced bus frequency to 37.5Mhz. I hope that now it works! Put the file in /boot/dtb/overlays directory and change led-conf5 with led-conf6 in /boot/armbianEnv.txt rk322x-led-conf6.dtbo 2.59 kB · 0 downloads Did as mentioned, here are the tests / results: - Bootup has the LEDs flashing red and blue each second - Ran ifconfig, no WLAN interface showing up - rk322x-config, undetected WiFi module - echo default-on > /sys/class/leds/working/trigger - turns on the red LED, and turns off blue LED... --- Expected: RED: on BLUE: ?? - echo heartbeat > /sys/class/leds/auxiliary/trigger - no effect, the blue LED is still off, red LED still powered on... --- Expected: RED: on BLUE: "beating" 2x per second - echo kbd-numlock > /sys/class/leds/working/trigger - numlock now toggles red / blue LED... red if numlock is active, blue if not... --- Expected: RED: on/off based on numlock BLUE: "beating" 2x per second - echo timer > /sys/class/leds/auxiliary/trigger - no effect, the blue is still the direct inverse of the red LED's state... off if the red LED is on, on if the red LED is off... --- Expected: RED: on/off based on numlock BLUE: toggling on/off state every second Made sure multiple times that /boot/armbianEnv.txt contained led-conf6 instead of led-conf5, current overlays: emmc wlan-esp8089 led-conf6 Image: https://imgur.com/a/SiJnFQs Edited October 19, 2021 by marshall had to clarify that the WLAN interface was the one that isn't showing, not all interfaces 0 Quote Link to comment Share on other sites More sharing options...
jock Posted October 19, 2021 Author Share Posted October 19, 2021 @marshall Mmmh, thanks for the detailed summary! It looks to me that the two leds are just attached one the opposite to the other; are you sure in Android the two leds are controlled separately and not just switched as like happens in Armbian? Actually I found two led definitions, but also two other definitions for chips controlling front 7-segment panels: all were enabled in the dtb, but clearly the manufacturers put everything in the pot no matter what the hardware is. Mostly the wifi matters more, this other dtb restores the power gpio to the previous value, but keeps the lower bus speed and higher power gpios, so try this other and let me know... rk322x-led-conf6.dtbo 0 Quote Link to comment Share on other sites More sharing options...
ArkhanLK Posted October 19, 2021 Share Posted October 19, 2021 On 5/13/2021 at 3:26 AM, jock said: Here it is! It's still weird you have building issues with Focal too, in a step (write uboot) that is mostly just bash scripting... Hey @jock, I just lost my minimal image due to a dead sdCard. Woul you happen to have it still available...? Sorry for troubling you with it! If not, I will try to build one myself (but I'm affraid I'd hit that wall once more...) 0 Quote Link to comment Share on other sites More sharing options...
marshall Posted October 20, 2021 Share Posted October 20, 2021 9 hours ago, jock said: It looks to me that the two leds are just attached one the opposite to the other; are you sure in Android the two leds are controlled separately and not just switched as like happens in Armbian? Not entirely sure as I can't really test it on Android... Not sure how... I'm speculating that it might be the same way as I am doing it right now, only with a Terminal Emulator app, but I haven't tried... On that dtb, the LEDs works exactly the same as the one before... About the WiFi, after trying that dtb, it still doesn't get detected on either ifconfig or rk322x-config Thanks for your time! 0 Quote Link to comment Share on other sites More sharing options...
jock Posted October 20, 2021 Author Share Posted October 20, 2021 17 hours ago, ArkhanLK said: Hey @jock, I just lost my minimal image due to a dead sdCard. Woul you happen to have it still available...? Sorry for troubling you with it! If not, I will try to build one myself (but I'm affraid I'd hit that wall once more...) Building right now, will be ready soon edit: ready and available here: https://users.armbian.com/jock/rk322x/ 1 Quote Link to comment Share on other sites More sharing options...
jock Posted October 20, 2021 Author Share Posted October 20, 2021 9 hours ago, marshall said: Not entirely sure as I can't really test it on Android... Not sure how... I'm speculating that it might be the same way as I am doing it right now, only with a Terminal Emulator app, but I haven't tried... On that dtb, the LEDs works exactly the same as the one before... Well I was referring to a much less elegant approach, like using the remote and see if the leds are both on at the same time or if they just alternate like it is happening on armbian. 9 hours ago, marshall said: About the WiFi, after trying that dtb, it still doesn't get detected on either ifconfig or rk322x-config Ok, dmesg is always helpful here. Is happening exactly the same as before or got worse? Unfortunately I don't have a board with esp8089 so can't test by myself the behaviour of this chipset, it turned out to be a bit picky 0 Quote Link to comment Share on other sites More sharing options...
marshall Posted October 20, 2021 Share Posted October 20, 2021 1 hour ago, jock said: Well I was referring to a much less elegant approach, like using the remote and see if the leds are both on at the same time or if they just alternate like it is happening on armbian. Ok, dmesg is always helpful here. Is happening exactly the same as before or got worse? Unfortunately I don't have a board with esp8089 so can't test by myself the behaviour of this chipset, it turned out to be a bit picky Oh, sorry I thought they weren't that necessary... First coldboot of the day: https://imgur.com/a/TbHgD72 (Same as before I think, it got detected, yet it can't connect or scan for APs) After a reboot: https://imgur.com/a/fv2pAmV (with dmesg, and sha1sum to make sure that the file is the same) Checksum on my side before copying into a flash drive: 0 Quote Link to comment Share on other sites More sharing options...
jock Posted October 23, 2021 Author Share Posted October 23, 2021 @marshall It would be more helpful if you could post dmesg | grep mmc, but still I can't realize where the issue could be... 0 Quote Link to comment Share on other sites More sharing options...
marshall Posted October 24, 2021 Share Posted October 24, 2021 22 hours ago, jock said: @marshall It would be more helpful if you could post dmesg | grep mmc, but still I can't realize where the issue could be... Thanks for the reply, Here's the output: [ 2.912601] dwmmc_rockchip 30000000.dwmmc: IDMAC supports 32-bit address mode. [ 2.912760] dwmmc_rockchip 30000000.dwmmc: Using internal DMA controller. [ 2.912778] dwmmc_rockchip 30000000.dwmmc: Version ID is 270a [ 2.912851] dwmmc_rockchip 30000000.dwmmc: DW MMC controller at irq 45,32 bit host data width,256 deep fifo [ 2.912910] dwmmc_rockchip 30000000.dwmmc: No vmmc regulator found [ 2.912920] dwmmc_rockchip 30000000.dwmmc: No vqmmc regulator found [ 2.912942] dwmmc_rockchip 30000000.dwmmc: GPIO lookup for consumer cd [ 2.912951] dwmmc_rockchip 30000000.dwmmc: using device tree for GPIO lookup [ 2.912982] of_get_named_gpiod_flags: parsed 'cd-gpios' property of node '/dwmmc@30000000[0]' - status (0) [ 2.913033] dwmmc_rockchip 30000000.dwmmc: Got CD GPIO [ 2.913044] dwmmc_rockchip 30000000.dwmmc: GPIO lookup for consumer wp [ 2.913052] dwmmc_rockchip 30000000.dwmmc: using device tree for GPIO lookup [ 2.913062] of_get_named_gpiod_flags: can't parse 'wp-gpios' property of node '/dwmmc@30000000[0]' [ 2.913070] of_get_named_gpiod_flags: can't parse 'wp-gpio' property of node '/dwmmc@30000000[0]' [ 2.913079] dwmmc_rockchip 30000000.dwmmc: using lookup tables for GPIO lookup [ 2.913088] dwmmc_rockchip 30000000.dwmmc: lookup for GPIO wp failed [ 2.925921] mmc_host mmc0: Bus speed (slot 0) = 400000Hz (slot req 400000Hz, actual 400000HZ div = 0) [ 2.938821] dwmmc_rockchip 30000000.dwmmc: 1 slots initialized [ 2.939319] dwmmc_rockchip 30010000.dwmmc: num-slots property not found, assuming 1 slot is available [ 2.939415] dwmmc_rockchip 30010000.dwmmc: IDMAC supports 32-bit address mode. [ 2.939514] dwmmc_rockchip 30010000.dwmmc: Using internal DMA controller. [ 2.939533] dwmmc_rockchip 30010000.dwmmc: Version ID is 270a [ 2.939595] dwmmc_rockchip 30010000.dwmmc: DW MMC controller at irq 46,32 bit host data width,256 deep fifo [ 2.939654] dwmmc_rockchip 30010000.dwmmc: No vmmc regulator found [ 2.939663] dwmmc_rockchip 30010000.dwmmc: No vqmmc regulator found [ 2.939683] dwmmc_rockchip 30010000.dwmmc: GPIO lookup for consumer wp [ 2.939693] dwmmc_rockchip 30010000.dwmmc: using device tree for GPIO lookup [ 2.939704] of_get_named_gpiod_flags: can't parse 'wp-gpios' property of node '/dwmmc@30010000[0]' [ 2.939714] of_get_named_gpiod_flags: can't parse 'wp-gpio' property of node '/dwmmc@30010000[0]' [ 2.939722] dwmmc_rockchip 30010000.dwmmc: using lookup tables for GPIO lookup [ 2.939732] dwmmc_rockchip 30010000.dwmmc: lookup for GPIO wp failed [ 2.939814] dwmmc_rockchip 30010000.dwmmc: allocated mmc-pwrseq [ 2.952648] mmc_host mmc1: Bus speed (slot 0) = 976562Hz (slot req 400000Hz, actual 244140HZ div = 2) [ 2.963464] dwmmc_rockchip 30010000.dwmmc: 1 slots initialized [ 2.973996] mmc0: Problem switching card into high-speed mode! [ 2.974106] mmc_host mmc0: Bus speed (slot 0) = 25000000Hz (slot req 25000000Hz, actual 25000000HZ div = 0) [ 2.974229] mmc0: new SDHC card at address 0001 [ 2.976074] mmcblk0: mmc0:0001 ASTC 14.8 GiB [ 2.978743] mmcblk0: p1 [ 3.005160] mmc_host mmc1: Bus speed (slot 0) = 976562Hz (slot req 300000Hz, actual 244140HZ div = 2) [ 3.055260] mmc_host mmc1: Bus speed (slot 0) = 976562Hz (slot req 200000Hz, actual 162760HZ div = 3) [ 3.111249] mmc_host mmc1: Bus speed (slot 0) = 976562Hz (slot req 100000Hz, actual 97656HZ div = 5) [ 6.393533] EXT4-fs (mmcblk0p1): mounted filesystem with writeback data mode. Opts: (null) [ 9.302119] EXT4-fs (mmcblk0p1): re-mounted. Opts: commit=600,errors=remount-ro 0 Quote Link to comment Share on other sites More sharing options...
chaka Posted October 26, 2021 Share Posted October 26, 2021 Hello! Help me please. Changing the TV box on the rk3228a chip. I want to start armbian from the SD card, download the stable version of the firmware and then do everything according to the instructions for writing to the SD card, but I can't start, the blue LED is on and that's it. The multitool starts, I can pour back my native firmware, that is, the box is not a brick. And after downloading the armbian image to the SD card, I can not open it on the computer and see the files inside the SD card, I plug it into another usb port, a disk appears, but it does not open and requires formatting. Tell me what could be the problem. I will try to upload a photo of my box. Boxing name Rombica smart box A1 0 Quote Link to comment Share on other sites More sharing options...
jock Posted October 26, 2021 Author Share Posted October 26, 2021 2 hours ago, chaka said: Hello! Help me please. Changing the TV box on the rk3228a chip. I want to start armbian from the SD card, download the stable version of the firmware and then do everything according to the instructions for writing to the SD card, but I can't start, the blue LED is on and that's it. The multitool starts, I can pour back my native firmware, that is, the box is not a brick. And after downloading the armbian image to the SD card, I can not open it on the computer and see the files inside the SD card, I plug it into another usb port, a disk appears, but it does not open and requires formatting. Tell me what could be the problem. I will try to upload a photo of my box. Boxing name Rombica smart box A1 Decompress the image before writing to sdcard? Or use a proper burning software that does it for you? 0 Quote Link to comment Share on other sites More sharing options...
jock Posted October 26, 2021 Author Share Posted October 26, 2021 @marshall sorry for taking so long in answering. At the moment I'm a bit busy and a bit tired too. Your issue seems to require more attention than expected, still could not figure out what is wrong. From the log you posted I see something very odd: I don't see the eMMC card (mmc2), the sdio wifi chip (mmc1) is not detected at all and the sdcard (mmc0) has problems getting into high speed mode. This is the hardest combo ever! 1 Quote Link to comment Share on other sites More sharing options...
chaka Posted October 28, 2021 Share Posted October 28, 2021 26.10.2021 в 20:12, jock сказал: Распаковать образ перед записью на sdcard? Или использовать подходящую программу для записи файлов, которая сделает это за вас? Hello again!I forgot to say that on the boxing board there is a sticker with the inscription MINI Q01, most likely this is the name of the boxing model. I was tortured with this TV box, but I don't want to retreat :-). What I just didn’t do, I changed the SD cards, the hdmi cable, thinking that the problem was in them, wrote down different images, inserted different dtb into the boot folder, downloaded different rk322a_loaders, but everything is useless. I still can't run armbian on my box. I read a lot in this forum, saw somewhere a message about Armbian_20.05.0-trunk_Rk322x-box_buster_current_5.5.16_minimal but could not find it to download. On my set-top box, we only hire such a memory chip written by 5NA92. It may be that the image does not start because of of this chip? https://www.clubedohardware.com.br/uploads/monthly_2021_03/mxlocalJumperq_rk3229_emcp_v3.1.jpg.373501677f0929eefdce8f001149b619.jpg 0 Quote Link to comment Share on other sites More sharing options...
chaka Posted October 28, 2021 Share Posted October 28, 2021 Sorry, there are typos in my texts, as I use a translator 0 Quote Link to comment Share on other sites More sharing options...
chaka Posted October 29, 2021 Share Posted October 29, 2021 Ha Ha HA. Congratulations, I did run it on my TV box. It's strange, but after I erased the flash not with a multitool, but with an android tool, it started. But now another problem has appeared, the wifi does not work. writes an unknown device. I opened and looked at the marking of the chip and the marking of S9012P. W 4 часа назад, chaka сказал: Hello again!I forgot to say that on the boxing board there is a sticker with the inscription MINI Q01, most likely this is the name of the boxing model. I was tortured with this TV box, but I don't want to retreat :-). What I just didn’t do, I changed the SD cards, the hdmi cable, thinking that the problem was in them, wrote down different images, inserted different dtb into the boot folder, downloaded different rk322a_loaders, but everything is useless. I still can't run armbian on my box. I read a lot in this forum, saw somewhere a message about Armbian_20.05.0-trunk_Rk322x-box_buster_current_5.5.16_minimal but could not find it to download. On my set-top box, we only hire such a memory chip written by 5NA92. It may be that the image does not start because of of this chip? https://www.clubedohardware.com.br/uploads/monthly_2021_03/mxlocalJumperq_rk3229_emcp_v3.1.jpg.373501677f0929eefdce8f001149b619.jpg hoever faced such a problem, how to solve it? 0 Quote Link to comment Share on other sites More sharing options...
jock Posted October 29, 2021 Author Share Posted October 29, 2021 13 hours ago, chaka said: Ha Ha HA. Congratulations, I did run it on my TV box. It's strange, but after I erased the flash not with a multitool, but with an android tool, it started. But now another problem has appeared, the wifi does not work. writes an unknown device. I opened and looked at the marking of the chip and the marking of S9012P. W hoever faced such a problem, how to solve it? Congratulations, but no logs, no photos, no dtbs, no original firmware => can't help 0 Quote Link to comment Share on other sites More sharing options...
dorin Posted October 30, 2021 Share Posted October 30, 2021 On 5/29/2020 at 5:57 PM, jock said: mpv: mpv does not require any particular treatment. Can be used directly from X11 in fullscreen mode or from a virtual terminal without X11. Install it if it is not already installed via usual apt command line. Run it with these switches to enable full video acceleration: mpv --hwdec=rkmpp -vo=gpu --gpu-api=opengl --gpu-context=drm --opengl-es=yes <video_file> Thanks, I tried this and it works. Now I am trying to play a couple of video streams simultaneously and display them in a grid. This is possible to achieve with omxplayer (https://github.com/popcornmix/omxplayer) but omxplayer works exclusively on raspberry pi GPU. But I cannot find how to do this with mpv, I had a look on the mpv documentation and the window geometry flags are ignored in gpu mode, and gpu rendering flags are a bit esoteric for me: https://mpv.io/manual/master/#gpu-renderer-options It's actually possible to start multiple mpv processes but since the window is always fullscreen, the image flickers between the sources. Any idea? 0 Quote Link to comment Share on other sites More sharing options...
jock Posted October 30, 2021 Author Share Posted October 30, 2021 1 hour ago, dorin said: Thanks, I tried this and it works. Now I am trying to play a couple of video streams simultaneously and display them in a grid. This is possible to achieve with omxplayer (https://github.com/popcornmix/omxplayer) but omxplayer works exclusively on raspberry pi GPU. But I cannot find how to do this with mpv, I had a look on the mpv documentation and the window geometry flags are ignored in gpu mode, and gpu rendering flags are a bit esoteric for me: https://mpv.io/manual/master/#gpu-renderer-options It's actually possible to start multiple mpv processes but since the window is always fullscreen, the image flickers between the sources. Any idea? Uhm, never actually tried to play two videos together on any rockchip box; I don't know the capabilities of rkmpp and what are the available switches and options to handle such situation. With mainline kernel it could be more viable, because it uses drmprime which is the standard and tidy way to render an hardware decoded stream, and uses the regular DRM capabilities. You should be able to get hardware video decoding on mainline kernel with instruction and mpv compiled available in this github post: https://github.com/armbian/build/pull/3152#issuecomment-922757262 The interesting arguments of mpv for you are --drm-draw-plane and --drm-drmprime-video-plane (see documentation here). The former allows you to select which is the plane mpv should use to render the GUI (seek bar, elapsed time, title, volume indications, info, etc...), the latter is the plane used to render the video. A plane is a hardware resource: the SoC provides one, two or more planes. The SoC, also, composes the planes in hardware to render the final output. As an example, when you run X server, the desktop is displayed on primary plane (which is rendered first, so it appears "below"), and videos are displayed on overlay plane (which is rendered next, so it appears "above"). rk322x unfortunately has just three hardware planes: two of them are full-featured planes, the remaining is for hardware cursor. The plane for hardware cursor is always rendered as last plane and is very limited (up to 128x128 pixels). About the two full-featured planes, one is allocated by the kernel as primary, and the other as overlay. Now, what you can try to do is run mpv rendering the first video on primary plane and the second video on the overlay plane. You should find a way to disable the GUI though, because otherwise one mpv instance will get control of both planes, leaving no available planes for the other. Another task is to find a way to tell mpv it has to resize the plane to wanted dimensions: since primary plane is below and overlay is above, if you run both videos fullscreen, the video allocated to primary plane will never be visible because it is always below. Other rockchip SoC are more capable in this sense: rk3328 has three full-featured planes. Bigger SoCs (like rk3288 or rk3399) have even more. Raspberry Pis are peculiar from this point of view. They have an hardware compositor which is capable of handling an arbitrary number of planes. You can create and destroy hardware planes at runtime and even change the z-order of them. The hardware compositor will render all of them until it is hardware limits are reached; after then, the HDMI output will start flickering and the monitor loses synchronization. Usually with raspberry Pi (even the oldest Pi1) you can at least run with two full-hd hardware planes without having issues. Good luck and keep us informed with the progress 0 Quote Link to comment Share on other sites More sharing options...
dorin Posted October 30, 2021 Share Posted October 30, 2021 Hmm but it sounds like the rockchip cannot provide a sustainable solution this way. Even if I got a source playing and sized correctly in each plane, I would have a hard limit of 2-3 streams. My goal is to monitor cctv cameras, like the rpi can play 16 streams quite easily (example), which I guess it's possible due to the arbitrary number of hw planes you mentioned. 0 Quote Link to comment Share on other sites More sharing options...
jock Posted October 30, 2021 Author Share Posted October 30, 2021 1 hour ago, dorin said: Hmm but it sounds like the rockchip cannot provide a sustainable solution this way. Even if I got a source playing and sized correctly in each plane, I would have a hard limit of 2-3 streams. My goal is to monitor cctv cameras, like the rpi can play 16 streams quite easily (example), which I guess it's possible due to the arbitrary number of hw planes you mentioned. i think you have to find a different solution than connecting hardware video decode to hardware plane directly for such use case. Ideally it should be possible to do hardware decode into a buffer that is then available for OpenGL to blit wherever you wish on your plane, I guess that drmprime is all about this, but you need to ask people with more knowledge than me. Maybe mpv forums can help you and mpv as is does not suit your task. Such job would involve GPU and you should be capable of getting it to work on rk322x too. How well in terms of FPS rk322x would perform is something that has to be discovered later. 16 hardware planes are probably too much for any SBC and, for your particular use case, are not needed. 0 Quote Link to comment Share on other sites More sharing options...
dorin Posted October 30, 2021 Share Posted October 30, 2021 (edited) 4 hours ago, jock said: i think you have to find a different solution than connecting hardware video decode to hardware plane directly for such use case. Ideally it should be possible to do hardware decode into a buffer that is then available for OpenGL to blit wherever you wish on your plane, I guess that drmprime is all about this, but you need to ask people with more knowledge than me. Maybe mpv forums can help you and mpv as is does not suit your task. Such job would involve GPU and you should be capable of getting it to work on rk322x too. How well in terms of FPS rk322x would perform is something that has to be discovered later. 16 hardware planes are probably too much for any SBC and, for your particular use case, are not needed. I was able to get a 2x2 grid working in hw accelerated ffmpeg this way: glrun ffmpeg \ -i $CH1 -i $CH2 -i $CH3 -i $CH4 \ -filter_complex \ "[0:v][1:v]hstack=inputs=2[top]; \ [2:v][3:v]hstack=inputs=2[bottom]; \ [top][bottom]vstack=inputs=2[v]" \ -map "[v]" -f matroska - \ | glrun ffplay - (CH1-4 are the video source URI's) While it's faster than non-accelerated ffmpeg (invoked without glrun), it's still way too slow compared to mpv player. To put it into perspective, based on some rought monitoring with htop: ffplay -i $CH1: uses about 340% of CPU (almost maxes out the capacity) glrun ffplay -i $CH1: uses about 140% of CPU mpv $CH1 uses about 300% of CPU mpv --hwdec=rkmpp -vo=gpu --gpu-api=opengl --gpu-context=drm --opengl-es=yes $CH1: uses only about 15% of CPU So obviously mpv has the potential to be much faster than ffmpeg, therefore I tried: # WARNING: this doesn't work mpv --hwdec=rkmpp -vo=gpu --gpu-api=opengl --gpu-context=drm --opengl-es=yes \ $CH1 --external-file=$CH2 --external-file=$CH3 --external-file=$CH4 \ --lavfi-complex='[vid1] [vid2] hstack [t1] ; [vid3] [vid4] hstack [t2] ; [t1] [t2] vstack [vo]' But this doesn't work because the ffmpeg filter used under the hood apparently cannot be accelerated: [ffmpeg] Impossible to convert between the formats supported by the filter 'mpv_src_vid1' and the filter 'auto_scaler_0' [lavfi] failed to configure the filter graph Again, running mpv without hw accel but with --lavfi-complex works, and works slightly faster than ffmpeg but still not satisfactory. I'll have a look on the mpv forums. Edited October 30, 2021 by dorin 0 Quote Link to comment Share on other sites More sharing options...
paradigman Posted November 1, 2021 Share Posted November 1, 2021 I just want to point out that the slightly neglected (side-handled) RK3228A CPU is equipped with TV boxes sold on Aliexpress for approx. 75%. If that’s not a big deal, I’d like to ask you to pay a little more attention to this, because there’s a growing deficit between the boxes that are commonly used here successfully (or officially supported by the Armian) and the specimens that occur in real life. As I indicated earlier, there is a new series of inserts starting with "R29". It would be important to me because I would buy hundreds of such boxes for business purposes, but slowly for a year I have been unable to find one on which the armbian would run flawlessly and the price would be right. I am a software developer and would love to cut into making custom armbian images, but the documentation found here is unsuitable for learning the operation. 0 Quote Link to comment Share on other sites More sharing options...
jock Posted November 1, 2021 Author Share Posted November 1, 2021 3 hours ago, paradigman said: I just want to point out that the slightly neglected (side-handled) RK3228A CPU is equipped with TV boxes sold on Aliexpress for approx. 75%. If that’s not a big deal, I’d like to ask you to pay a little more attention to this, because there’s a growing deficit between the boxes that are commonly used here successfully (or officially supported by the Armian) and the specimens that occur in real life. As I indicated earlier, there is a new series of inserts starting with "R29". It would be important to me because I would buy hundreds of such boxes for business purposes, but slowly for a year I have been unable to find one on which the armbian would run flawlessly and the price would be right. I am a software developer and would love to cut into making custom armbian images, but the documentation found here is unsuitable for learning the operation. Sorry but I have to be sincere about this: I don't like the reasoning, I found it unethical. Believe it or not, I don't receive any kind of money from anyone; Armbian team also does not receive anything from these no-name boards, instead has to pay electricity, bandwidth to host images and forums. It is just courtesy of Armbian guys if all of this is possible, they ask nothing in charge; although Armbian does not officially support any TV Box - as clearly and boldly stated in the first posts of all my threads. My role in all of this is just driven by passion and learning, not money. I decided to share my work and studies with others because I think that if other people does the same, the world will become a better and funnier place. As you see, the learning curve is pretty steep, and trying to build an out-of-the-box working solution for boards whose specifications and datasheets are kept secret is a huge time wasting job. What I expect in change is usually a "Thank you": most of the time it suffices. Here I read, and I really hope I misunderstood the post, that someone is going to make business around this. That's pretty ok to me: I do support tv boxes for passion. I also use Linux which is free and made by others for passion and work and I pay nothing to anyone; nonetheless I contribute to opensource the way I can. Now if your concern is missed support, don't ask community what can do for you and point out what community is not doing for your business; start thinking about what you can do for community that helps both community and your business. In all of this long thread I don't know how many people donated to Armbian project - I'm not part of the official team. But I'm totally sure that, except for the very generous board donations from @fabiobassa to whom all people here should as well be very grateful, I never ever received a penny or one board from anyone. And note that I don't ask boards as gratification - I already have several of them taking dust - but to study them. Nonetheless I am still here available for free support and for fun. Asking me, or anyone else, to spend money to buy cheap crap and spend time reverse engineering that crap for your business is just unethical. It's parasitizing. You need good Linux support for your tv boxes? Ask the manufacturers of those boxes, and see what they answer to you. The problem about the "right price" is that the price can't be right if you don't include software developing and maintaining costs. Chinese tv boxes manufacturers are fetching "just working" Android distro into their products. If they would want to support a full linux environment, the price would be three or four times the actual price. Want to talk about business? Send a precise request to the developer of your choice (me included) and wait for a proper price quotation. Want to do everything by yourself? You can. Armbian developers spent hours to write proper documentation, available on docs.armbian.com. Source code is publicly available on github repository. If you find it difficult navigate into, spend hundreds of hours lo learn things, as me and several others did, to raise your skills to keep up with your business. 5 Quote Link to comment Share on other sites More sharing options...
paradigman Posted November 1, 2021 Share Posted November 1, 2021 @jock: I’m not unethical, at most outspoken. I believe that 20+ years in the profession entitle me to judge certain projects and I've spent hundreds of hours with it. Here I see the armbian team heroically struggling with time and lack of capacity (and money) while keeping the most important info that the tools of reverse engineering need for themselves. "Want to do everything by yourself? You can." No, I can't. The "docs.armbian.com" is all about building images, changing some parameters, and running a script. You are also well aware that when a new pcb comes out it is too little. "Want to talk about business?" I have no objection to paying for it. In fact, it would even be good if there were different payment options posted in this forum for different support needs. Perhaps this would immediately solve the ongoing funding problems of this community. 0 Quote Link to comment Share on other sites More sharing options...
jock Posted November 1, 2021 Author Share Posted November 1, 2021 46 minutes ago, paradigman said: @jock: I’m not unethical, at most outspoken. I believe that 20+ years in the profession entitle me to judge certain projects and I've spent hundreds of hours with it. Here I see the armbian team heroically struggling with time and lack of capacity (and money) while keeping the most important info that the tools of reverse engineering need for themselves. This demonstrates that you know very little about Armbian. Official support is granted for boards which have official specs and whose manufacturers provide datasheets. Hopefully those manufacturers also provide fundings, sponsorship or any kind of support. There is no reverse engineering with officially supported boards: manufacturers provide specs, in form of openly available electrical diagrams and general documentation. Tv boxes don't provide any datasheet about board specs, often chinese manufacturers even provide fake specs to buyers to sell their crap. They definitely does not sponsor the project in any form, thus Armbian does not endorse any tv box, nor do I: you need something for a serious project? Buy serious hardware, not chinese cheap and unsupported crap. 53 minutes ago, paradigman said: "Want to do everything by yourself? You can." No, I can't. The "docs.armbian.com" is all about building images, changing some parameters, and running a script. You are also well aware that when a new pcb comes out it is too little. If you think Armbian purpose is to tell you how to understand boards and their chips, understand datasheets, write proper device trees, configure and compile the linux kernel, and whatever it requires to properly support a board, you're out of context. I spent time doing all of these things above, so everyone else can do that. If you don't have enough skills to do that, that's not a fault of the project. You can still build up those skills, if you have enough time and will to sacrifice for the purpose. 58 minutes ago, paradigman said: "Want to talk about business?" I have no objection to paying for it. In fact, it would even be good if there were different payment options posted in this forum for different support needs. Perhaps this would immediately solve the ongoing funding problems of this community. There is a donation and help providing page, not just money, but whatever people can do is appreciated, even writing documentation. Plus the community does not need to be funded with money, it is funded by will of people who wants to do something useful for others. Armbian is not a company that does things on people request and I don't think you can "hire" someone from inside the project for your particular needs. But again I repeat: you can still contact and pay developers - ANY developer, not necessarily an Armbian developer - for your needs. 2 Quote Link to comment Share on other sites More sharing options...
fabiobassa Posted November 1, 2021 Share Posted November 1, 2021 @paradigman Quote I’m not unethical, at most outspoken ..... Ohhhhh that is great so we can all be outspoken without doubt of being misanderstood and without dramas before I answer for MY side and MY point of view , maybe it worth to remember how this " adventure " on rk3229 has began https://forum.armbian.com/topic/12401-long-story-linux-on-rk3229-rockchip/ Now... I am not a programmer, I am much more on hardware side because own instruments ( oscilloscope among other, vesy useful to catch clocks and data pins) and all that story is made of trial and errors, intuitions, hex editors , unsleepy night. YOU can follow the same approach if you want, it is public BUT... but..but be ready to invest time and money since neither me neither @jock neither armbian guys I guess own the latest boards. Then one day I met by fortune ( because life is made of right peoples in right moments ) @jock on this forum , and this adventure started: HE into programming, Me into hardware and replacements and here we go on armbian. But isn t only jock and me.. many ideas have been taken from other " OPEN MINDED PEOPLE" such as @knaerzche @hexdump , many other that for the simpe fancy of studying have in GREAT ways contribute to this walking into NON OFFICIAL supported tvboxes. And all of them for FREE while all of us could have taken money advandage to sell home assistant or voip pbxses or even kiosks .. no, this wasn't our approach. Now i came to the point that personally found the most aggressive, unrespectful, at limit of being polite, so here I will be outspoken as well as you have been. Quote I believe that 20+ years in the profession entitle me to judge certain projects.... Hell.. NO ! You can take your 20 years in the professional and trhow it away if you want to judge the work of someone else or even the approach to work of someone else, or even more the IDEA of work of someone else You are NOONE ....ABSOLUTELY noone TO SAY THIS, this is very narcissistic approach and you can take it for yourself. Not the right place here to judge. One can be the biggest company of the world and still being the smallest man of the earth when these sentences come out .. " my 20 years in professional " .......... Yours 20 years in professional are for me NOTHING. And at this point not only for me I guess. if 20 years in professional aren't enough to let you study what even ME studied without programming bases, well is YOUR problem if 20 years in professional aren't enough to open an hex editor and try by " trial and errors" , well is YOUR problem if 20 years in professional aren't enough to understand the way this armbian works, well is YOUR problem. But guess you have no time.. business business and again business.... Believe... is not a problem of money, donation's or fama and success. Approach only looking for business and not compartecipating to discoveries are sooooo far away from the spirit of many on this forum. Best wishes for your business , if you are 20 years on the professional you also know what is sacrifice and not sleeping or sleep and waken out with a solution in mind. AS @jock already said contact directly rockchip.. may be for hundreds or even thousand pieces they will release the infos you need.. with money all is possible and i will NOT judge this approach. Just is not mine. I DONATED boxes to jock.. many .. 3228.. 3229... 3318 for ?????? FOR ??? FOR STUDY!! And after you spent money to obtain info sure we don't aspect you will share here for free. Business as always. Investments must be protected, keep secret. WE will continue to " play " as children with our little discoveries and you will continue in you 21 years professional As I told.. no dramas. Just outspoken. 0 Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.