03:12airlied[d]: skip updated my kelper macbook from fc37 to fc43, once I fixed things, it seems fine just as data point
13:36gfxstrand[d]: airlied[d]: Please tell me that's new enough it's x86
14:31mohamexiety[d]: it is, should be 2012 or thereabouts
14:31mohamexiety[d]: so.. sandy bridge or ivy bridge intel
15:05Siggi: apparently mmiotrace is the hammer I was looking for.
15:15orowith2os[d]: Do you need a chisel to go with it? :P
15:48Siggi: no, I'll just use my thumbs - I have a full complement of 10 left-hand thumbs :)
17:13Siggi: OMG, this tooling is awesome! envytools/rnn/demmio augments the trace with register names. I still have no idea what any of this means, but this tooling is no less awesome for that!
17:22karolherbst: I'm surprised it still works
17:32Siggi: doesn't look like it's seen a lot of love lately, but the K2100M dates back to 2011-2012. It looks like the mmiotrace is parsed correctly, at least, though I haven't debugged
17:32Siggi: now that I have you on the horn, I was scrubbing through the nouveau sources and reading some docs
17:33Siggi: it seems the driver writes the PLL registers directly, and there's a mention somewhere that if the VRAM(?) or some clocking is too slow, this might lead to artifacting
17:34Siggi: so I guess I might try to poke at those to begin with?
17:36Siggi: I could see how right-hand side stripes and occasional full screen (line) shearing would result from some crtc underruns due to bandwidth problems
17:37Siggi: assuming there's some kind of horizontal blanking, the underrun would accumulate towards the RHS of the LCD?
17:38Siggi: (though IDK much about LCD screen refresh timing constraints - presumably the panel has some amount of buffering?)
17:38karolherbst: well the VRAM PLLs shouldn't be touched by nouveau unless you explicitly reclock
17:39Siggi: would it be reasonable to think that the card comes up with bad PLL settings, perhaps?
17:39Siggi: I guess I can figure out how to force reclocking
17:40karolherbst: yeah. that might be happening
17:41karolherbst: I know there are a handful of keplers with a broken clock config that lead to instability unless reclocked
17:41Siggi: K, thanks, that's something to look at at least
17:41karolherbst: there is a pstate file in debugfs for your GPU
17:43Siggi: K. I'm still figuring out how to hold things. Only managed to unload the nvidia driver this morning
17:43Siggi: so I guess next up is to poke at nouveau and look for differences
17:47Siggi: Mmm, looks like the nvidia driver is writing all the PLL config right out of the gate
17:50Siggi: (Mkay, gotta learn how to hold it - it's reading all the PLL config and writing ... some)
17:55Siggi: karolherbs: I'm looking at https://nouveau.freedesktop.org/KernelModuleParameters.html, not seeing anything to force reclocking
17:56Siggi: but when I look at the godforsaken NVIDIA control panel, it has a notion of some kind of performance state
17:57Siggi: I see clocks bump from (IIRC) 135MHz to something like 810MHz
17:57Siggi: does the card manage this autonomously?
18:24mhenning[d]: The card won't manage it automatically on kepler, it's something that the kernel does on that card
18:24mhenning[d]: For reclocking kepler under nouveau, https://github.com/polhdez/nouveau-reclocking-guide might be a good guide
18:38_lyude[d]: Anyone know any places outside of red hat that might be hiring nova people at the moment? I kind of assume the answer is no unfortunately, but figured I might as well ask 😒
18:49airlied[d]: gfxstrand[d]: Yes, the first retina, it's ivybridge, I think I ewasted any power ones!
19:31gfxstrand[d]: I remember. My youth pastor in highschool was Mac obsessed and he wouldn't stop talking about it. 😂
19:32karolherbst[d]: my first computer was an Atari and then one of those iMac G3s
19:32karolherbst[d]: I actually lived through the transition lol
19:32karolherbst[d]: my last mac was some intel macbook .. wait actually I also have a M2...
19:33gfxstrand[d]: When I was in university, we had a computer lab in the basement of the English building that was all iLamps.
19:34karolherbst[d]: ohh, the edu variants
19:34karolherbst[d]: I guess
19:34gfxstrand[d]: Those things were such shit. They were cool but shit.
19:35karolherbst[d]: 😄