00:10 <apritzel> wens: thanks for the MCU clock patches, btw: I will review them in the next days (now that the CPU clock driver is off my back)

04:44 <wens> what was the need for the identifier?

04:54 <wens> IMO a string might be slightly harder to use than an integer

09:57 <apritzel> wens: well, this string is what's printed on the chip, listed in the wiki, and used by Syterkit. There is more than one A527 version, so this is used to discern them

10:00 <apritzel> wens: anyway, this string is purely informational, for the end user, which could then be used to tell them: your chip does not have the NPU

10:02 <apritzel> I think in the driver we must not use that anyway. Can't we just bail out if the NPU ID register reads zero? I guess the whole NPU MMIO frame is RAZ/WI then?

14:13 <diego71> apritzel: I've got a cubie A7

14:13 <diego71> *A7A

14:14 <diego71> I'm looking at your patch for A733 cpu, but I see no dts file for it.

14:15 <diego71> Do you have any suggestion how to try your patches?

14:58 <wens> apritzel: yeah, I guess it should bail out

15:23 <wens> apritzel: for some reason you didn't do internal clocks for the a523 main ccu?

16:15 <apritzel> diego71: I made one simple generic DT, for testing purposes, will send it later

16:17 <apritzel> wens: yes, that's D1 clock driver style, and I didn't really know what clocks we would need or not

16:47 <wens> I feel like there's no real downside to that style either?

17:31 <apritzel> Yeah, it's not a clear decision. It sounds better to not expose every clock to the DT, but ultimately that would work regardless, since the DTB features numbers only anyway, and all numbers would work

17:31 <apritzel> I think for such a small CCU like the CPU one this sounds more feasible

17:56 <wens> unless you don't expose them through the OF provider, it really doesn't matter since one can still use raw numbers