#asahi-dev · 2026-07-03 15:22
chaos_princess:(#asahi-dev · 2026-07-03 15:21
sventhe bisect result will probably be "old atcphy+dwc3 works, new one doesn"t"#asahi-dev · 2026-07-03 15:21
svenchaos_princess: very because of the whole dwc3 glue driver and atcphy changes#asahi-dev · 2026-07-03 15:21
jannauack. I have patches which avoid some of the differences, without changes#asahi-dev · 2026-07-03 15:20
chaos_princesshow annoying would a bisect be?#asahi-dev · 2026-07-03 15:20
chaos_princessthere was a working version, right?#asahi-dev · 2026-07-03 15:19
svenprobably some timing bullshit#asahi-dev · 2026-07-03 15:19
sventhe mmio traces look similar enough and yet it just refuses to work in linux#asahi-dev · 2026-07-03 15:19
sveni'm really at a loss what's going on there#asahi-dev · 2026-07-03 15:18
jannauif I only had a device with working thunderbolt. *scnr*#asahi-dev · 2026-07-03 15:17
sven(can totally recommend https://github.com/EEVengers/ThunderScope fwiw)#asahi-dev · 2026-07-03 15:17
mischa85chaos_princess: i have that exact same question#asahi-dev · 2026-07-03 15:17
jannauyuka: 14 core m3 max is special beacuse it has only 3/4 of the memeory controllers compared to the full m3 max#asahi-dev · 2026-07-03 15:16
svenjannau: yeah, something like that is my suspicion as well. i have an oscilloscope and maybe it's about time i get some m4 or m5 machine as well :-)#asahi-dev · 2026-07-03 15:16
chaos_princesswell, it won't be, so whats the problem#asahi-dev · 2026-07-03 15:16
mischa85chaos_princess: i wasn't trying to get any code accepted, that's the whole point#asahi-dev · 2026-07-03 15:15
jannauI suspect slight clock de-sync issues between device generations. m3 ultra's uart still syncs roughtly till 375000 bps but breaks above. I plan to look at it with glasgow's auto bps uart widget at it#asahi-dev · 2026-07-03 15:13
chaos_princessas you have said, it is llm-generated, which we do not accept#asahi-dev · 2026-07-03 15:08
mischa85produce some clean-room documentation. did i already say i went in with a hazmat suit and am tainted? i have strictly talked about it within #asahi-re context. tldr; it's llm slop, not ready for upstream in any way, shape or form.#asahi-dev · 2026-07-03 15:08
mischa85nice, some slander. let me be clear, i did and will not try to bring any of this upstream. this is my own personal repo, and i even tagged one commit as "RE-derived, DO NOT UPSTREAM". this is strictly RE territory. i'll tag the other commits as well with capitals, but at this point i'm also okay with putting the repo on private to prevent anyone from seeing whatever happens on there, and eventually #asahi-dev · 2026-07-03 14:50
yukabut have different midr part numbers#asahi-dev · 2026-07-03 14:49
yukaM5 Pro and M5 Max are both t6050#asahi-dev · 2026-07-03 14:49
yukat6000 vs t6001, t6020 vs t6021, t6040 vs t6041#asahi-dev · 2026-07-03 14:49
chaos_princesscorrect#asahi-dev · 2026-07-03 14:49
yukabut on M1/M2/M4, Pro and Max were still different SoC ids#asahi-dev · 2026-07-03 14:48
yukagood to know#asahi-dev · 2026-07-03 14:48
yukaokayy#asahi-dev · 2026-07-03 14:48
chaos_princessyep#asahi-dev · 2026-07-03 14:48
yukaso they use the same device tree, and then m1n1 detects the disabled cores?#asahi-dev · 2026-07-03 14:48
enverbalalici think m4 max 14 and 16 are both t6041 as well#asahi-dev · 2026-07-03 14:47
chaos_princessin fact, t6034 is the weird one#asahi-dev · 2026-07-03 14:47
chaos_princessyes#asahi-dev · 2026-07-03 14:47
yukaand were they both j314s too?#asahi-dev · 2026-07-03 14:47
chaos_princessboth 10 and 8 are t6000#asahi-dev · 2026-07-03 14:47
chaos_princessm1 and m2 pros#asahi-dev · 2026-07-03 14:46
yukaAs counter-example, M3 Max (14-core) is t6034 while M3 Max (16-core) is t6031#asahi-dev · 2026-07-03 14:46
yukaWas there any machine with different variants (cpu core count) but using the same SoC designation?#asahi-dev · 2026-07-03 14:41
svenmight be a good idea to look at those machines with an oscilloscope#asahi-dev · 2026-07-03 14:33
enverbalalictime to write that dc earlycon then#asahi-dev · 2026-07-03 14:33
enverbalalicyeah, an m2 on the other end#asahi-dev · 2026-07-03 14:32
yuka* on the other end#asahi-dev · 2026-07-03 14:32
yukathey are semi broken, don't work with high baudrates, or even at all depending on which other macbook or serial chip you have on the other devices#asahi-dev · 2026-07-03 14:31
yukaand these quirks are the reason I believe we should no longer consider debuguart the main serial console for those devices#asahi-dev · 2026-07-03 14:31
yukawith M2 <-> M4 Pro#asahi-dev · 2026-07-03 14:31
yukaflokli had the same#asahi-dev · 2026-07-03 14:31
yukaM5 Pro <-> M1 Pro also wasn't able to get any trace of serial#asahi-dev · 2026-07-03 14:30
yukanot all debuguarts work with all debuguarts#asahi-dev · 2026-07-03 14:30
yukaenverbalalic yes, you were using an M2 Macbook on the other end I believe?#asahi-dev · 2026-07-03 14:28
integralpilotoh, interesting