03:44 <wens> it seems I didn't get paulk's patches again

04:26 <junari> wens, apritzel: I can run desktop via vnc with command weston --renderer=gl --backend=vnc --disable-transport-layer-security --xwayland --width=1280 --height=720

04:27 <junari> glmark2 works and opengl progs too. I can see that gpu 100% loaded in nvtop

04:42 <wens> that's nice

04:43 <wens> junari: you need to also do a patch to update the mali bindings and add a new compatible string for the a523

04:49 <junari> This part remains the most confusing to me

05:10 <wens> how so?

05:46 <junari> It's not clear to me how to generate them

06:14 <wens> edit the yaml file and add an entry? copy/paste and modify?

06:44 <junari> In some cases it's just copypaste, yes. In case a523 compatible string should be here?

06:44 <junari> https://elixir.bootlin.com/linux/v6.15.5/source/Documentation/devicetree/bindings/gpu/arm,mali-bifrost.yaml#L41

06:48 <wens> I think so

08:47 <warpme> apritzel: really appreciate care on me ;-) Some hw i'm receiving from users as support of me with getting new hw useable... I would say recent community work on a523 makes this soc usable for some basic usecases Infact all use-cases where hdmi is not needed are fully relevant for a523 (as usb, eth, smp and cpu dvfs are working - headless server usecases are well handled). So my users of miniarch (which is full archlinux are already

08:47 <warpme> happy :) 

08:52 <warpme> btw: wifi also works well....

08:58 <warpme> now we have gpu - so only hdmi is missing to enable not small set of new usecases where small tiny/slow desktop is enough....

09:48 <wens> wifi mostly depends on what chip the board uses, unless of course they used pcie

10:11 <warpme> wens: indeed. e.g. cube uses aic8800 so we can use only vendor code :-( I backported it to mainline and - despite terrible style of code - it works above average wifi on various sbc (on 5G i'm getting 200-230Mbps user throughput so with ~40% protocols overhead it gives ~500mbps on air. not bad i think)....

10:34 <gamiee> aic8800 code is not perfect, but surprisingly it works quite well

10:44 <apritzel> warpme: I do get that it's "just HDMI missing", but this just is not a small thing. Look at the H616: we got basic support 3 years ago, and graphics output is still mostly missing in mainline

10:46 <apritzel> warpme: just wanted to dampen expectations: it's one thing to get basic headless functionality, since we know this stuff and it's typically documented

10:46 <apritzel> but all things graphics are much less documented, and often need to be painfully reverse engineered, and then also made to fit into the mainline frameworks

10:48 <apritzel> so if you want graphics on a new Allwinner SoC, you either wait a few years or need to do it yourself - or buy Rockchip ;-)

10:49 <warpme> it is possible that a523 switch to inno phy makes this enablement much easier/faster as this phy is used by varia other socs while h616 phy was unique (used only in one soc familly). of course key here are brains not code. if i would be aw or radxa - i would invest support into most difficult to mainline hw like i.e. hdmi. rk done like this and results are really nice....

10:50 <apritzel> the HMDI PHY is just one thing, there is also the display engine (DE) in play here

10:50 <apritzel> so it's complex and tends to gets more involved with new SoCs

10:52 <apritzel> the H616 graphics code is for the new DE only so far (we still don't have mainline-ready HDMI code), and the DE code took many months to get into a proper shape

21:06 <loki666> found this ... https://gitlab.com/tina5.0_aiot/product/docs/-/tree/be1db4639682f80d27a7c01693eaf05991435574/A527/Hardware%E7%A1%AC%E4%BB%B6%E7%B1%BB%E6%96%87%E6%A1%A3/%E7%A1%AC%E4%BB%B6%E8%AE%BE%E8%AE%A1/PCB%E5%8F%82%E8%80%83/%E7%A1%AC%E4%BB%B6DDR%E6%A8%A1%E6%9D%BF%E3%80%81%E5%8F%A0%E5%B1%82%E6%96%87%E4%BB%B6

21:06 <loki666> not sure if there is anything usefull

21:06 <loki666> seems to be stuff aveout a527 ddr

21:28 <loki666> https://x.com/RadxaComputer/status/1942069827910426706

21:28 <loki666> some datasheets for A733 / A527 / T527

22:49 <apritzel> loki666: unfortunately those "DDR" documents seem only to contain physical design recommendations, and no DRAM controller related information

22:50 <apritzel> loki666: and we learned about those manuals a few days ago already, but thanks for the heads up anyway