#asahi-dev · 2023-06-06 00:53
marcanjannau: let's aim for 13.5#asahi-dev · 2023-06-05 23:16
marcanalyssa: ... am I supposed to answer that question? :p#asahi-dev · 2023-06-05 22:35
marcanso probably easy to support#asahi-dev · 2023-06-05 22:35
marcanthat said, the device tree regs look exactly like normal apcie except a single port (which makes sense since it goes to the bridge chip)#asahi-dev · 2023-06-05 22:35
marcanand they just share the same PHYs#asahi-dev · 2023-06-05 22:35
marcanso maybe the digital side is a completely separate controller from "traditional" apcie#asahi-dev · 2023-06-05 22:34
marcanthe MMIO addresses for apcie-ge are totally different#asahi-dev · 2023-06-05 22:34
marcanohh but this is interesting#asahi-dev · 2023-06-05 22:23
marcanthis at least explains why the changes to the PCIe controller I guess, if they were planning to use the same old APCIe for slots #asahi-dev · 2023-06-05 22:23
marcanI should test that on the machines I have...#asahi-dev · 2023-06-05 22:11
marcan(because that's how Apple rolls)#asahi-dev · 2023-06-05 22:11
marcana whole ass 64-bit ARM too#asahi-dev · 2023-06-05 22:11
marcanbut each NAND chip has its own sub-controller#asahi-dev · 2023-06-05 22:11
marcanthe SSD controller is in the SoC#asahi-dev · 2023-06-05 22:10
marcanhttps://social.treehouse.systems/@marcan/110493886026958843 <- best guess as to the topology#asahi-dev · 2023-06-05 22:09
marcanthey have their own controllers on the NANDs#asahi-dev · 2023-06-05 22:09
marcanit's custom#asahi-dev · 2023-06-05 22:09
marcanapple NANDs are PCIe#asahi-dev · 2023-06-05 22:09
marcanapple doesn't use that#asahi-dev · 2023-06-05 22:09
marcannv-ddr?#asahi-dev · 2023-06-05 21:53
marcanyeah, that tracks#asahi-dev · 2023-06-05 21:51
marcanlol.#asahi-dev · 2023-06-05 21:48
marcanlike they planned this for a real, separate Mac Pro chip with actually more PCIe, canceled that, and "made it work" with M2 Ultra#asahi-dev · 2023-06-05 21:47
marcanI'm starting to think it's a big cheat with PCIe switches and it's all funneled down to 16 lanes on the SoC side#asahi-dev · 2023-06-05 21:47
marcanthe Mac Pro PCIe lane count baffles me#asahi-dev · 2023-06-05 08:28
marcannewer mesa properly neutered that driver entirely so now it's just dead code#asahi-dev · 2023-06-05 08:28
marcanthat's what the lovely manjaro people enabled for us which broke the ALARM mesa#asahi-dev · 2023-06-05 08:27
marcan< mps> I expected UAPI difference warning but didn't seen it <- if you build non-asahi mesa with asahi enabled there is no UAPI, that code is broken and never supposed to run at all#asahi-dev · 2023-06-05 08:27
marcanwell yes, that's kinda the point :p#asahi-dev · 2023-06-05 08:27
marcancompared to what?#asahi-dev · 2023-06-05 08:27
marcanthan with -4?#asahi-dev · 2023-06-05 08:22
marcan(particularly older versions, that got fixed recently)#asahi-dev · 2023-06-05 08:22
marcancrashing X sounds like what happens when you build regular mesa with the asahi driver enabled#asahi-dev · 2023-06-05 01:28
marcanI haven't tested on all machines yet, but if I can't repro it anywhere I think you'll have to debug that one yourself. could be a musl thing.#asahi-dev · 2023-06-05 01:28
marcanmps: mesa doesn't care about machine differences and if you had kernel GPU problems I would expect faults in dmesg, so that sounds like a random userspace problem#asahi-dev · 2023-06-04 13:40
marcanlook in /proc/device-tree/chosen#asahi-dev · 2023-06-04 11:18
marcaneric_engestrom: I think the crash thing should be fixed with the m1n1 update I just pushed#asahi-dev · 2023-06-04 06:34
marcanI'll let it run while I port this into m1n1#asahi-dev · 2023-06-04 06:34
marcanI did some magic pokes and it works now#asahi-dev · 2023-06-04 06:34
marcanok, the j413 issue is thermal I think. we were missing some PMGR init to enable throttling.#asahi-dev · 2023-06-04 03:46
marcanpovik: if you can make the alias change I can merge that ^#asahi-dev · 2023-06-03 06:14
marcanthe package versions always match the tags in some fairly obvious pattern/transformation#asahi-dev · 2023-06-03 06:13
marcansid-linux: I forget to push repos all the damn time, but the tags are always there because the build scripts rely on that, so you can always just look for the latest tag in the relevant repos#asahi-dev · 2023-06-03 06:12
marcanjannau: there may secretly be more than fixes ;)#asahi-dev · 2023-06-03 06:11
marcanpushed the PKGBUILDs repo#asahi-dev · 2023-06-03 02:48
marcan(no fix for the j413 thing yet, just the other stuff)#asahi-dev · 2023-06-03 02:48
marcanupdated kernel/mesa pushed to asahi-dev#asahi-dev · 2023-06-02 14:34
marcanit's the client lib#asahi-dev · 2023-06-02 14:34
marcanchadmed: this is libpulse which is what everything uses with pipewire#asahi-dev · 2023-06-02 12:26
marcanif it's a CI issue I'm sure you can wrap the CI scripts in an fd cleaner ;)