11:45 <brads> @narmstrong have been testing vdec on S922X with 5.7-rc3 but found I needed to increase the size of the kernel cma area to get anything substantial to decode. Increased it from size = <0x0 0x10000000>; to size = <0x0 0x80000000>;

11:46 <brads> It still fails at times to start usually with a few lines in the syslog like "alloc_contig_range: [74e00, 753ef) PFNs busy"

11:48 <brads> And sometimes stalls with heavily loaded with the same errors. Is there a better way of allocated more contiguous ram specifically for the driver / firmware?

12:08 <chewitt> brads: I've had it set to 768MB for a while (originally 896MB) and Neil/Maxime have the same hack in their testing trees

12:11 <narmstrong> probono: it collided with your system libfdt

13:09 <brads> chewitt: Thanks i'm not going crazy then :) I'm still having some intermittent troubles with some video streams and strangely if I run it via strace it works. 1080p video to around 40fps without dropped frames and its decoding 4k with a few atifacts although framerates are down to no more than around 10-15 fps in most cases.

13:11 <brads> But im not using drmprime which I probably should be, also maybe wayland libMali is conflicting with mpv

13:13 <brads> one of the kernel workers (events) goes a little crazy at times which causes some slowdown but I did notice this prior to vdec usage as well

13:28 <narmstrong> brads: you should try https://patchwork.kernel.org/patch/11514473/ and https://patchwork.kernel.org/patch/11514475/

13:37 <brads> narmstrong: oh very nice, will give it a try

15:11 <brads> narmstrong: those patches help out quite a bit with the smoothness of video. I am seeing alot of maxed out kworker/1:0+events (causing shudder) and kworker/0:3+events causing freezes.

15:15 <brads> Wondering if you have any insights to what might be causing this, occurs in all 5.x versions ive tried.

15:19 <brads> I will try to play with some cpu affinities for the irq's

17:00 <probono> hi narmstrong, thanks for your response. i don't have libfdt installed but it turns out I must *not* check out the armlogic git in a subdirectory of the custodians git, then the compilation works

17:27 <probono> I am trying to build u-boot for a p281 based retail device. Looking at https://groups.io/g/u-boot-amlogic/message/181:

17:28 <probono> "If you feel confident enough, you'll need to dump the eMMC content to save the original bootloader (in the first 4MiB) and use the p281 amlogic u-boot config to generate the FIP binaries as explained in the mainline README file of P212."

17:28 <probono> Now, I am looking at https://gitlab.denx.de/u-boot/custodians/u-boot-amlogic/-/blob/u-boot-amlogic/board/amlogic/p212/README.p212

17:28 <probono> but I don't seem to find instructions on how to use the dumped first 4MiB from the device

17:28 <probono> Can you give me a hint? Thanks

18:18 <chewitt> probono: do you have the fip sources for p281?

18:18 <chewitt> if yes, all GXL derived boards follow the same signing recipe

18:19 <chewitt> https://github.com/chewitt/LibreELEC.tv/blob/amlogic-master/projects/Amlogic/bootloader/install#L244

18:27 <probono> i only have the device in question, and i have dumped its first 4 MB

18:27 <probono> i don't have any sources from the manufacturer

18:28 <probono> just compiling custodian u-boot and chainloading it results in an immediate reboot of the device

18:32 <probono> chewitt, i understand that I need to extract something from the dumped first 4 MB from the device, and use that when I compile u-boot. am I wrong?

18:37 <probono> getting:

18:37 <probono> gxl_p281_v1#go ${loadaddr}

18:37 <probono> ## Starting application at 0x01080000 ...

18:37 <probono> "Synchronous Abort" handler, esr 0x02000000

18:37 <probono> (reboot)

20:59 <hexdump0815> probono: for me the following worked well on different gxl systems

20:59 <hexdump0815> probono: fatload mmc 0 0x01000000 u-boot.bin

21:00 <hexdump0815> probono: go 0x01000000

21:00 <hexdump0815> probono: so maybe try a different address?

21:05 <probono> thanks hexdump0815. unfortunately i get the same result.

21:05 <probono> do ou have your u-boot.bin somewhere for download?

21:05 <probono> did you try it on a x96 mini by any chance?

21:08 <hexdump0815> probono: looks like this is your lucky day :) - yes and yes - https://github.com/hexdump0815/imagebuilder/blob/master/boot/boot-amlogic_gx-aarch64/u-boot.bin

21:10 <probono> cool, thanks hexdump0815

21:14 <probono> hexdump0815, wow, that one works for me

21:24 <probono> ...until i do "usb start", at which point the device reboots :(