00:03 <MoeIcenowy> apritzel: does H616 come with different bins?

00:05 <apritzel> MoeIcenowy: I don't know, but warpme_ seems to build on this assumption ...

00:06 <apritzel> I just wanted to test that setting the PLL_CPUX to 1.8GHz works, and it does

00:07 <apritzel> some OPPs are twice in the BSP DT, so I think there is something to pick the right one at runtime

10:41 <evadot> mhm, anyone familiar with the a64 thermal ?

10:41 <evadot> with a 24mhz clock and a thermal_per of 244 I get an data interrupt at ~1sec rate

10:42 <evadot> so does the thermal generate an interrupts every 4 samples and not for each one ?

10:45 <evadot> mhm, the filter was set to 8, so that doesn't explain the behavior I'm seeing ...

10:45 <evadot> setting it to 4 gives me an interrupts at a ~500ms rate

11:05 <warpme_> MoeIcenowy: apritzel: re: q about "does H616 come with different bins?" - yes. i'm assuming that as bsp read sid for bin (https://github.com/orangepi-xunlong/linux-orangepi/blob/06f40bef0d9a08c233b4b2ed6f9064c3abe3ecb6/arch/arm64/boot/dts/sunxi/sun50i-h616-orangepi-zero2.dts#L5421)

11:07 <warpme_> apritzel: re:" your q about "did you see this of_device_is_compatible() call in sun50i-cpufreq-nvmem.c?" - yes. i added compatible: https://github.com/warpme/minimyth2/blob/7898f5b61979fd63e346bd10d0dd72903fdd9dc7/script/kernel/linux-5.13/files/0571-drivers-thermal-allwinner-add-h616-ths-support.patch#L20

11:13 <evadot> meh ok, my calculation were all wrong :) I got it now

11:41 <MoeIcenowy> warpme_: strange it's the lowest 16bits of the first word of chipid

11:41 <MoeIcenowy> maybe similar to H2+/H3, bin info is in chipid?

11:50 <warpme_> MoeIcenowy: honestly: idk. but printk shows read value is 0. how you conclude that it is first word of chipid? Generally: my logic was quite simple (and mayby wrong): there is efuse and @00 addr. there is value saying what bin is...

11:51 <MoeIcenowy> warpme_: well just because it's an offset

11:51 <MoeIcenowy> and it's little endian

11:57 <warpme_> but imho this shouldn't causing kernel trap issue i have on sun5i-cpufreq-nvmem right? In my both devices read value is 0 - so we are asked to use speebin0 from opp table - or i'm wrong? (we can investigate this later - when somebody better than me will review patches)

12:57 <warpme_> hmm: interesting (re aw859a wifi):

12:57 <warpme_> https://www.irccloud.com/pastebin/uS9mlctz/

14:12 <warpme_> apritzel: jernej: just fyi: i measured cpu_vdd on opi-zero2 and for 1.3G opp it should be 1120mV but is 912mV. this explain why i can't get any opp over 1.3GHz. What will be your suggestion here?

14:46 <apritzel> warpme_: is there any easily accessible test point for the voltage? at some coil, maybe?

14:56 <apritzel> and what is your workload? Do you just switch to 1.3GHz and let it idle?

15:29 <warpme_> apritzel: yes. let me try to depict this :-)

15:30 <warpme_> workload is idle. yes - i set higher opp to 1.3G & set governor to performance

15:39 <apritzel> so I used pure cpufreq-dt yesterday, and ran hackbench for a few minutes (loadavg goes into the 100s), at 1.8GHz (verified) and 1.12V (programmed, but not verified)

15:43 <warpme_> apritzel: will it be ok: https://imgshare.io/image/opizero2-vdd-cpu-pin.pm64hp

15:44 <apritzel> warpme_: cool, thanks, looks good

15:45 <warpme_> oh - i see i choose wrong profession.....

16:33 <apritzel> warpme_: I measure 1.13V (with my lousy equipment) when in the 1,5GHz@1.12V pstate

16:35 <apritzel> warpme_: did you adjust the dcdca max voltage and also referenced dcdca as the "cpu-supply"?

17:41 <warpme_> apritzel: i think so: https://github.com/warpme/minimyth2/blob/7898f5b61979fd63e346bd10d0dd72903fdd9dc7/script/kernel/linux-5.13/files/0597-arm64-dts-allwinner-h616-enable-ths-support.patch#L295 But it looks i lost dcdca as the "cpu-supply" in my patches. After disabling sun5i-cpufreq-nvmem and adjusting cpu-supply in dt it started to work :-) (so far tested 1.5G opp but it should be ok also with higher opps i think). many

17:41 <warpme_> many thanx for great help!

17:54 <warpme_> 1.8G opp also work nice!