09:33 <DarkNeutrino> apritzel so i checked the OPP driver. One thing thats bit concerning is that we have 4 variants. But 1 is the lower bin. Which seems to be 0x4. Second one is the higher bin with lower mV for the freq. That one is 0x3. Up to 1200Mhz only those bins are used but for the 3 frequencies above (1296, 1344, 1512) And on those we have the 0x1 and 0x2 bins. 1344 has 0x4. And when i read the BSP driver for it, its completely insane. A s

09:34 <DarkNeutrino> version to one of the 4

09:35 <DarkNeutrino> The h6 BSP driver in comparasion is walk in the park LOL

09:38 <apritzel> DarkNeutrino: well, but isn't that just a matter of carefully *once* creating the DT nodes in sun50i-h616-cpu-opp.dtsi?

09:46 <DarkNeutrino> Not really. The driver has to be modified. And from what i see quite heavily to account for this. https://github.com/bigtreetech/CB1-Kernel/blob/kernel-4.9/kernel/linux-4.9/drivers/cpufreq/sun50i-cpufreq-nvmem.c here is the BSP driver im looking at and https://github.com/bigtreetech/CB1-Kernel/blob/kernel-4.9/kernel/linux-4.9/arch/arm64/boot/dts/sunxi/sun50iw9p1.dtsi#L132 here is the DT. Some bins just directly dont support so

09:47 <DarkNeutrino> 0x1 which enables the highest frequency if its some combination or what that it applies also the lower frequencies

09:48 <DarkNeutrino> Or if it just supports the highest one. Its weird to say the least.

09:50 <apritzel> PLEASE don't look at BSP code too closely, that only gives you a headache ;-)

09:51 <apritzel> I could be wrong, but if you look at the mainline H6 CPU OPP code, that just figures out the bin ID, then leaves all the rest to existing (generic) code

09:52 <apritzel> so the voltages and frequencies for *every* bin version is just listed in sun50i-h6-cpu-opp.dtsi

09:53 <DarkNeutrino> Correct. H6 OPP just grabs the ID and applies the freq and voltage according to bin. There is the issue tho. H6 lists all frequencies for all bins. Just different voltage

09:53 <DarkNeutrino> H616 doesnt do that and some frequencies are on some bins just outright disabled.

09:56 <apritzel> is that really a problem? what happens if you don't list opp-microvolt-speed<x> property for bin <x> in a frequency node?

09:58 <apritzel> or you could just put the voltage of the next OPP in, so that say both 1.08 GHz and 1.32 GHz run at the same voltage, namely that one for the higher frequency

09:58 <apritzel> that's not really efficient, but should be safe

10:03 <DarkNeutrino> Leaving the property off should be fine. But then which frequencies does 0x1 use if its not on max ? Actually let me just flash the CB1 system image onto a spare board and test.

10:42 <DarkNeutrino> Right the kernel lists no files in cpufreq folder. Just great LOL

22:37 <machinehum> apritzel: Sorry, yeah fully open source hardware

22:37 <machinehum> Layout and schematic, all the software used to design it is also open source

22:39 <machinehum> apritzel: The idea is I design these SOMs, and then generate leads and do consulting for people

22:39 <machinehum> Potentially they want to just have an integrated board

22:40 <machinehum> Potentially they want to design a board in 3hrs and slam a SOM on it to solve a $100k problem

22:41 <machinehum> Details and docs are slim for the A523

22:48 <machinehum> But thanks

23:10 <apritzel> sure, the A523 is brand new, but at least Yuzuki has had hands on it

23:11 <apritzel> machinehum: you still need to know what the idea/target of this SoM should be, people will want to know why not just use an RPi CM4