#
#asahi-dev · 2025-09-22 21:42
jannausigh: j474s as `compatible = [atc-phy,t6020, atc-phy,t8112]` in the ADT
#asahi-dev · 2025-09-22 17:53
svenessentially re-introduce that rtkit_backend_ops struct or however it was called and move the implementation for that structure into platform and/or soc
#asahi-dev · 2025-09-22 17:50
svenyou might have to move the rtkit backend to drivers/soc/apple as well
#asahi-dev · 2025-09-22 17:30
ruiconok thanks
#asahi-dev · 2025-09-22 17:29
chaos_princessprobably also there. also take inspiration from cros-ec, those are very similar in purpose
#asahi-dev · 2025-09-22 17:27
ruiconok, but where should i put the header file?
#asahi-dev · 2025-09-22 17:25
chaos_princessplatform probably, the same place where cros-ec lives
#asahi-dev · 2025-09-22 17:25
ruicon(on both pre-t2 and t2 macs)
#asahi-dev · 2025-09-22 17:25
ruiconthe driver has been written and tested
#asahi-dev · 2025-09-22 17:25
ruiconall intel macs
#asahi-dev · 2025-09-22 17:24
chaos_princesst2 or pre-t2 ancient smc?
#asahi-dev · 2025-09-22 17:23
ruiconsven: My SMC backend is currently under MFD, Do you have any suggestions on where it should go?
#asahi-dev · 2025-09-22 16:36
jannauI think the reasoning is not more than "this is not a battery powered device so make the gpu a little faster"
#asahi-dev · 2025-09-22 16:33
chaos_princessi don't know the reasoning for 3, but it works, so?
#asahi-dev · 2025-09-22 16:33
chaos_princesslet's add a machine compatible override
#asahi-dev · 2025-09-22 16:05
jannauI think for j274 as well, DT has 3, it's missing in the ADT and the code uses 1 then
#asahi-dev · 2025-09-22 15:58
jannauhow do we want to deal with the mismatching base pstate? use "1" from the the adt or add a machine compatible override in initdata.rs?
#asahi-dev · 2025-09-22 15:52
jannauconfirmed, after devicing cs/afr volts by 1000 everything matches for m2 pro mac mini
#asahi-dev · 2025-09-22 15:32
chaos_princessyep, seems like forgetting to convert from microvolts to millivolts in cs and afr
#asahi-dev · 2025-09-22 15:19
chaos_princessoh, that would make sense
#asahi-dev · 2025-09-22 15:19
jannauoff by factor 1000
#asahi-dev · 2025-09-22 15:03
jannau"8a 02 00 00" vs. "10 eb 09 00" so quite a bit different
#asahi-dev · 2025-09-22 14:39
chaos_princessi do not see anything obvious unfortunately, and afaict cs/afr only exist on m2+
#asahi-dev · 2025-09-22 13:32
jannaunext mismatch is "Hwdata B first mismatch: 4308", .aux_ps
#asahi-dev · 2025-09-22 13:31
jannauthat was because we override gpu-perf-base-pstate to 3 for the m2 pro mac mini
#asahi-dev · 2025-09-22 13:21
jannauit's rel_boost_freqs
#asahi-dev · 2025-09-22 13:20
chaos_princessso not in rel_max_powers?
#asahi-dev · 2025-09-22 13:18
jannauerr, if off course off because it increemnts the offset first
#asahi-dev · 2025-09-22 13:14
jannauthe mismatch is in .rel_max_powers
#asahi-dev · 2025-09-22 13:14
chaos_princessneat, will remember
#asahi-dev · 2025-09-22 13:14
jannau`awk 'BEGIN {offset = 0}; /field/ {offset += $4; print offset, $3, $4}'` on the -Zprint-type-sizes output for the struct works
#asahi-dev · 2025-09-22 13:12
jannaum2 pro
#asahi-dev · 2025-09-22 13:10
chaos_princessg14s is m2-something, right?
#asahi-dev · 2025-09-22 13:09
chaos_princessyes, it is that dumb
#asahi-dev · 2025-09-22 13:09
chaos_princesscopy to another file, then print offset_of for the field that looks reasonable and binary search from there
#asahi-dev · 2025-09-22 13:09
jannauI guess it's parsing -Zprint-type-sizes output
#asahi-dev · 2025-09-22 12:48
jannaupahole doesn't work for the rust structs in m1n1
#asahi-dev · 2025-09-22 12:47
jannauI'm looking at m1n1 GPU init and have "Hwdata B first mismatch: 4056" for g14s
#asahi-dev · 2025-09-22 12:46
jannauchaos_princess: how do you determine the fields in the GPU initdata based on the offset?
#asahi-dev · 2025-09-21 20:21
jannaufound a couple of more places but changed those on linux side
#asahi-dev · 2025-09-21 20:03
kettenisI can make those same changes on my end
#asahi-dev · 2025-09-21 19:59
jannauI think it's only https://github.com/AsahiLinux/linux/commit/4d047f2d271c0e413a0a19984c7dfec29153a0bd but I'm not sure
#asahi-dev · 2025-09-21 19:57
jannaukettenis: yes, fix is s/0/1/ at 3 places. I made the the same changes in the linux code. I also did a few cosmetic changes until I remembered that those increase the diff needlessly
#asahi-dev · 2025-09-21 19:46
svenwe're close now
#asahi-dev · 2025-09-21 19:46
sventhat's what i assumed :D
#asahi-dev · 2025-09-21 19:45
kettenissven: correct; I was waiting for you to fix all the bugs ;)
#asahi-dev · 2025-09-21 19:34
kettenishopefully I can find some time later this week to do that
#asahi-dev · 2025-09-21 19:33
kettenisfix is easy and I haven't sent the patches upstream yet
#asahi-dev · 2025-09-21 19:24
kettenisjannau: oops, no, removing that delay in those loops was not intentional
#asahi-dev · 2025-09-21 18:04
svenand openbsd doesn't even support atcphy yet afaik

plus ›