ebach has joined #armbian-amlogic
<Armbian-Discord> <a​deepv> Hi. In its current form, this patch will not be accepted into the kernel. We had a little discussion with Jerome Brunet about it.
<Armbian-Discord> <a​deepv> Here's his opinion:
<Armbian-Discord> <a​deepv> "My guess is the core phase might need to be adapted depending on
<Armbian-Discord> <a​deepv> * Mode: It seems the DDR modes are done by making the controller run faster then diving the output clock by 2. This divider might mess up the phase shift needed.
<Armbian-Discord> <a​deepv> * Speed: Maybe there is delay contraint between the input clock and the core and we need at adapting the phase shift depending on the rate ?
<Armbian-Discord> <a​deepv> I think what could be tried - with a LOOONG RFT giving a chance people to report regressions - is:
<Armbian-Discord> <a​deepv> * Defaulting the Core phase to 270: Despite AML HW engineer recommendation, this is what the vendor code does. Maybe this will help with board that seems to require 270 to start.
<Armbian-Discord> <a​deepv> I know if stays for all modes, it will cause problems
<Armbian-Discord> <a​deepv> 1) Set 180 when switching to DDR modes
<Armbian-Discord> <a​deepv> 2) Wether switching to 180 for high speed SDR modes (UHS, HS200) is required, or not, is a bit unclear. If the problem is the internal divider, it should not be required. If the problem is the delay, maybe it is."
<Armbian-Discord> <a​deepv> I don't have a chance to dive into this problem right now, but I'll try to get into it later this year.
<Armbian-Discord> <a​deepv> Ooh, I did not finish reading, you already discussed it 🙂
<Armbian-Discord> <r​pardini> Yes, thanks Vyacheslav -- we're going with your DT approach, for practical reasons. Turns out some non-AXG also need 270 to work in practice. It's been tested working with BpiM5 (sm1) while keeping Radxa Zero (also sm1) working.
<Armbian-Discord> <r​pardini> I will later add a README.md or such to the patch folder with some of the rationale/decisions behind our patching.
<Armbian-Discord> <a​deepv> in my tests some emmc modules works with 180 and 270 phase, some only 270. Odroid C4 in this list too (sm1 family)
<Armbian-Discord> <r​pardini> Ok.... If it turns out that C4 needs it I will personally patch it's DT too. If it turns out that "most of SM1" needs it, I will invert: patch the dtsi for SM1 (like AXG), and de-patch the Radxa. Either way it's gonna work...
<Armbian-Discord> <r​pardini> One last thing: do you think the clock phase stuff might affect SDIO wifi things? I don't understand enough of the basics to say
<Armbian-Discord> <a​deepv> armbian has more devices in the lab than I have 🙂 that's why I asked to test on everything. I see that home-assistant in their haos for odroid c4 apply a phase value of 270
<Armbian-Discord> <r​pardini> yes. Also LibreELEC applies it for everything
<Armbian-Discord> <r​pardini> (and it does not work...)
<Armbian-Discord> <a​deepv> I haven't researched this point enough, but on my devices there were no problems.
<Armbian-Discord> <a​deepv> libreelec use amlogic kernel 4.9 as far as I know
<Armbian-Discord> <s​erkanp> yes coreelec and libreelec uses kernel 4.9, amlogic company supported till 4.9 then they abondoned kernel updates..
<Armbian-Discord> <a​deepv> not exactly, amlogic started the s4 series and supports it in its kernel 5.4
<Armbian-Discord> <s​erkanp> sm1 and s905 series limited to 4.9
<Armbian-Discord> <a​deepv> But they didn't move support for all "old" processors to 5.4 😞
<Armbian-Discord> <r​pardini> Nah, it's mostly Christian's series. See https://github.com/LibreELEC/LibreELEC.tv/tree/master/projects/Amlogic/devices/AMLGX/patches/linux
<Armbian-Discord> <r​pardini> Do not confuse CoreELEC with LibreELEC...
<Armbian-Discord> <a​deepv> this is the same patch that is used everywhere and that was also in the armbyan before my version
<Armbian-Discord> <r​pardini> yes...
<Armbian-Discord> <r​pardini> we've a secret non-alliance of LibreELEC / Armbian / HassOS 😉
<Armbian-Discord> <r​pardini> (meaning: I steal patches from all of them)
ebach has quit [Quit: ebach]