Danct12 has quit [Remote host closed the connection]
tnovotny has joined #linux-sunxi
dsimic is now known as Guest3814
dsimic has joined #linux-sunxi
Guest3814 has quit [Ping timeout: 480 seconds]
warpme has quit []
bauen1 has joined #linux-sunxi
kuba2k2 has quit [Read error: Connection reset by peer]
tnovotny_ has joined #linux-sunxi
tnovotny has quit [Read error: Connection reset by peer]
Daanct12 has quit [Quit: WeeChat 4.0.5]
warpme has joined #linux-sunxi
<warpme>
junari: fyi: great work for h618 lpddr4 support. i backported your code and it work grat on my zero3
kuba2k2 has joined #linux-sunxi
kuba2k2 has quit [Ping timeout: 480 seconds]
junari__ has joined #linux-sunxi
junari_ has quit [Ping timeout: 480 seconds]
JohnDoe_71Rus has quit []
<warpme>
fyi: lpddr4 based opi zero2w also works nicely with lpddr4 code :-)
Danct12 has joined #linux-sunxi
<apritzel>
warpme: what are the parameters you use? Because junari__'s didn't work for me, I still see 128MB and then a hang. Haven't dug deeper yet, or need to double check that the PMIC is set up correctly
<apritzel>
warpme: also: as you seem to have an Zero2W: care to send a DT? And do you have the add-on board as well?
<warpme>
apritzel : i'm using dram para from junari - but i need to incr. dcdc3 to 1150mV. With 1100mV i have kernel hang freeze in middle of boot :-(
<warpme>
btw: i'm still on 2021.07 codebase as i added Eth support for ac200/ac300 + axp305/axp313/axp1530 support. Soon i have plan to play with kernel in-kernel h616 ephy support (but not having strong motivation as there is no yet kernel ac300 support afaik)
<warpme>
re zero2w add-on board: i'm awaiting for shipment. currently i'm using uwe5622 based wifi.
<warpme>
junari: fyi: your code works nicelly for lpdd4 - but compilation fails (for me) for lpddr3 variants. (probably some cosmetics in h616_ddr3_1333.c is needed?)
<apritzel>
warpme: ah, 1150mV was a good hint, that now seems to work
<apritzel>
(it actually writes 1140mV, since it's the 20mV range already)
<apritzel>
and actually even 1140mV is not enough for me, it hangs then after leaving TF-A, but 1160mV goes further
<warpme>
apritzel : indeed - i recall 1140mV was a first stable for me so I added 100mV for safety :-)
<apritzel>
or wait, I think the manual is wrong about the DCDC3 encoding:
<junari>
apritzel: warpme: you can look in my patch for axp313 there are correct value for dcdc3
<junari>
it was many time ago, when we understand that datasheet is wrong
<warpme>
junari - i compared mine /drivers/power/axp313.c with yours (from patchset) and only difference are some printfs....
<apritzel>
junari: Firefox dislikes the certificate from yandex.ru, and the company firewall outright denies the whole domain anyway :-(
<apritzel>
so I will have to wait till tonight to have a look
JohnDoe_71Rus has joined #linux-sunxi
<apritzel>
so does anyone have an H616/H618 board with LPDDR4, but less than 4GB of DRAM? OrangePi shipped different images for each DRAM size, so likely the DRAM parameters are different
<junari>
apritzel: company fyrewall - maybe, but my firefox say that site secure
<junari>
warpme: can you please post my patch somewhere for apritzel?
<warpme>
sure give me sec....
<apritzel>
junari: it's not a biggie, since I can override it, but I get: Error code: SEC_ERROR_UNKNOWN_ISSUER
<junari>
hmm, maybe redirection or something else
<warpme>
apritzel : may you try hxxp://warped.inet2.org/opi3-zero-ddr4-uboot-junari-patches.zip ?
<junari>
oh this free internet
<junari>
My first opizero3 had a defective HDMI soldering, maybe something wrong with your
<warpme>
junari: do you have stable operation on zero3 with your code. I had 2 hangs in last 10min....let me try zero2w
<apritzel>
warpme: thanks, I got it
<warpme>
apritzel : be my guest!
<junari>
warpme: yes, but I test with openwrt
<junari>
without problem
wasutton- has joined #linux-sunxi
<warpme>
my tests are with hd live tv continuous playback - so probably a bit more stresfull (dma from video decoder to gpu with iommu)
<warpme>
now i'm testing on zero2w...
<buZz>
tv playback is stressfull?
<warpme>
hmm - zero2w also freezes with distorted picture. oh....
wasutton3 has quit [Ping timeout: 480 seconds]
<warpme>
buZz : well - after all it is contineius transfers of data between Eth - decoder - gpu - hdmi in dma mode with iommu
<buZz>
at pretty low rates, isnt it?
<jakllsch>
decoded video isn't low rate
<jakllsch>
it's a raster image(s) for every frame
<buZz>
over DMA though, shouldnt take much load?
<jakllsch>
CPU no, memory yes
<buZz>
is the 'with iommu' the culprit?
<jakllsch>
i don't think that'll make much difference..
<jakllsch>
using the IOMMU would add some iommu page table lookup traffic though
<warpme>
fyi: i tested zero3/zero2w whole night playback with xulong uboot. both were stable iirc
<warpme>
only diff is uboot
<junari>
It's strange that your boards need more voltage when mine runs at 1100
<buZz>
junari: maybe cable has high resistance ;)
<apritzel>
for video decoding (and also 3D rendering) you having constant *concurrent* load on the DRAM controller:
<apritzel>
one for the display engine to scan out the framebuffer (say 1080p@60Hz => ~500 MB/s)
<apritzel>
and the other for the GPU or video decoder to fill the framebuffer, with a similar data rate
<junari>
warpme: What do I need to do to check the situation on minimyth?
<apritzel>
junari: your version of axp313.c is doing the right thing, so you really get 1.1V when you say so
<buZz>
~355MB/s , isnt it?
<buZz>
i dont know but it sounds like not very high for ram?
<apritzel>
not for PC hardware, no ;-)
<buZz>
:)
<buZz>
ok
<apritzel>
but also the point is that two DMA engines constantly access potentially the same pages
<apritzel>
or say nearby pages, so the DRAM controller has to close them, and reopen the other pages, which means the timing setting are really exercised and can become critical
<apritzel>
so hangs during 3D rendering or video playback can point to imprecise DRAM parameters
<apritzel>
that is what the old lima-memtester was exercising
<apritzel>
on normal (DRAM heavy) workloads the DRAM controller stress is often dampened by the caches
<apritzel>
buZz: btw: 355 MB/s is for 24bit depth, but typically framebuffers (at least on sunxi) are using 32 bits per pixel
<warpme>
junari: i'm developing 2 distros. second is miniarch (enabler/bootstrap for ArchLinux on SBCs). If you want I can prepare image for you - so you will have full standard ArchLinux on Zero3....
<warpme>
apritzel : nice mailing :)
<buZz>
apritzel: ah, alpha too?
<apritzel>
buZz: I don't know if it's actually used for an alpha channel, but just aligning each pixel on a word makes things a lot easier
<buZz>
hmhm
<junari>
warpme: yes please. I'll try it on my board
<warpme>
sure give me few sec
bauen1 has quit [Ping timeout: 480 seconds]
mps has left #linux-sunxi [#linux-sunxi]
<junari>
warpme: about lpddr3, I think it's because of the new tpr6 parameter
tnovotny_ has quit []
<warpme>
junari: oh it was small cosmetics in /arch/arm/mach-sunxi/dram_timings/h616_ddr3 files. i fixed this and tested code on zero2. works ok :-)
junari has quit [Remote host closed the connection]
<warpme>
have cpu cooler - board may randomly turn-off by kernel ths when it reach crit trip (when die will reach 100C)
<warpme>
(interestingly: adding "debug" in kernel command line significantly reduces such critical temp power-offs for me. fyi: on /boot part there is boot-debug.scr with enabled kernel "debug". you can use it to have debug kernel verbosity at Archlinux boot)
apritzel has quit [Ping timeout: 480 seconds]
junari has joined #linux-sunxi
bauen1 has joined #linux-sunxi
warpme has quit []
vagrantc has joined #linux-sunxi
<junari>
warpme: On the first board the image is loaded, on the second - not
<junari>
and hdmi does not work
warpme has joined #linux-sunxi
<junari>
But HDMI didn’t work for me on my tvbox with a TV and monitor, so I think it’s a matter of the mainline drivers
<junari>
With the second board that not starting, I'll see what can be done. Maybe dram_para should be another
warpme has quit []
warpme has joined #linux-sunxi
ftg has joined #linux-sunxi
junari has quit [Ping timeout: 480 seconds]
bauen1 has quit [Ping timeout: 480 seconds]
apritzel has joined #linux-sunxi
warpme has quit []
JohnDoe_71Rus has quit []
tlwoerner_ has joined #linux-sunxi
warpme has joined #linux-sunxi
warpme has quit []
tlwoerner has quit [Ping timeout: 480 seconds]
apritzel has quit [Ping timeout: 480 seconds]
bauen1 has joined #linux-sunxi
<GrantM11235[m]>
Is the tcon on the f1c100s similar to any other chips?