#asahi-dev · 2024-06-24 22:19
jannaunicolas17: also failed on actual hw (m2 pro mini)#asahi-dev · 2024-06-24 22:18
chadmed_hopefully once (if?) the driver is merged we can add it to the yaks and others can help out here#asahi-dev · 2024-06-24 22:18
chadmed_jannau: not a problem at all. i will rip out the smc key stuff in the MR and replace it with just the keys i know exist on the m2 max studio for now (later) and we can sort out other machines as we figure them out#asahi-dev · 2024-06-24 21:22
jannau2GB so far at ~0.5 MB/s#asahi-dev · 2024-06-24 21:02
nicolas17ok someone confirmed the 15.0b2 ipsw doesn't even install in a VM (I forgot about that possibility -_-)#asahi-dev · 2024-06-24 20:59
jannaudisconnecting power while the wrong ipsw is dfu restored seems be a bad idea. It spends a long amount of time in "Your mac is currently downloading for the connected mac", I suspect it resumed the dfu restore#asahi-dev · 2024-06-24 19:28
nicolas17it looks like it would fail because https://wkms-public.apple.com/fcs-keys/vqTQNjxKOR8CCw4-SSqrJCmVu7x1zktNdOXoNm7RIXQ= is missing#asahi-dev · 2024-06-24 19:26
nicolas17anyone brave enough to try installing macOS 15.0b2 from the ipsw?#asahi-dev · 2024-06-24 19:02
jannaucounter frequency is still 24 MHz so this is probably preparation to switch to ! GHz#asahi-dev · 2024-06-24 18:59
jannauBIT(0) in the sysreg s3_1_c15_c1_5 switches counter scaling on M3#asahi-dev · 2024-06-24 18:32
nicolas17so I guess the kernel is changing the frequency back and forth on context switch depending on the process's SDK version#asahi-dev · 2024-06-24 18:32
nicolas17"For compatibility purposes, this change will only be visible when using the SDK associated with this release or later"#asahi-dev · 2024-06-24 18:31
nicolas17release notes of today's betas: "On Apple Silicon based devices with M3 or later, and A16 Bionic or later, the values returned by reading the CNTFRQ_EL0 and CNTVCT_EL0 registers have been updated to 1 GHz, instead of the prior value of 24 MHz."#asahi-dev · 2024-06-24 13:55
jannauI haven't yet looked at the DCP plane/blanding changes. the SMC keys took much more time than anticipated#asahi-dev · 2024-06-24 13:38
chadmedanyone had a chance to look at https://github.com/AsahiLinux/linux/pull/304? all good if not. im taking a week off work and going to melbourne next week so ill have plenty of time to sort out this and anything outsanding for the hwmon driver too#asahi-dev · 2024-06-23 22:48
jannau"Te0[12]" might be efficiency core temp sensors. Only issue with that is that the ultra has only Te2[01] which have bogus values#asahi-dev · 2024-06-23 21:02
nicolas17but later under quirks "reading any of the keys afterwards returns 0, except after a write to one of them, which allows you to read data once more (but just once)"#asahi-dev · 2024-06-23 21:02
nicolas17heh, wiki says "So far, I haven't been brave enough to try 2 [read with payload] or 3 [write]."#asahi-dev · 2024-06-23 21:01
jannauperf cores see a 7+ degree jump#asahi-dev · 2024-06-23 21:01
jannauI assume I get a better signal if I run both cores#asahi-dev · 2024-06-23 20:59
jannaubut none of the CPU temp sensors "Tp0." or "ftP." show a significant increase running `stress-ng -c 2 --taskset 0,1`#asahi-dev · 2024-06-23 20:57
jannauI asssume the ftG[0-7] keys are GPU cluster sensors. I haven't checked if they make sense#asahi-dev · 2024-06-23 20:56
chadmedthe CPU ones are all there properly too but i didnt positively confirm that they are "correct"#asahi-dev · 2024-06-23 20:55
chadmedGPU#asahi-dev · 2024-06-23 20:55
jannauGPU or CPU?#asahi-dev · 2024-06-23 20:54
chadmedi think tonight after work i will just split out the common etc and we will call it there, theres clearly very little actually common to all socs#asahi-dev · 2024-06-23 20:54
chadmedjannau: both of the gpu temp sensors work properly on m2 max. in fact, most of the keys work properly there just not anywhere else :p#asahi-dev · 2024-06-23 20:45
nicolas17jannau: you should update https://github.com/AsahiLinux/docs/wiki/HW:SMC with your WIP findings ^^#asahi-dev · 2024-06-23 18:39
jannaustill no temp sensors for efficiency cores identified#asahi-dev · 2024-06-23 15:46
jannauhardcoded list inside the devicetree#asahi-dev · 2024-06-23 15:41
j`eyhardcoded lists inside the kernel is sad#asahi-dev · 2024-06-23 15:27
jannauGPU sensors are at ~7.8 degrees, yet I sit next to it in t-shirt/shorts, I wonder if those only read something useful when the GPU is powered on#asahi-dev · 2024-06-23 15:25
jannauI wonder if the "ftP*" keys are better CPU core temp keys. at least there are exactly $NUM_COREs ones#asahi-dev · 2024-06-23 15:19
jannaumost of the common keys are not present on the M1 ultra#asahi-dev · 2024-06-23 15:18
jannausigh, looks like we need to create a database of SMC keys#asahi-dev · 2024-06-22 23:17
chadmedjannau: yeah my ICC profile still works, and somehow so does builtin even though we have no EDID :p #asahi-dev · 2024-06-22 17:15
jannauvirtual desktop switch still fixes it#asahi-dev · 2024-06-22 17:05
jannauon fedora I now get completely black screen. a kde devoler seems to see the same behavior on intel#asahi-dev · 2024-06-22 16:56
jannauchadmed: did you try color management with kwin 6.1 / mesa 20240527#asahi-dev · 2024-06-22 13:30
jannaum4 die shot confirms dual dcpext (I think) and 4 atc#asahi-dev · 2024-06-22 09:53
jannauchaos_princess: I can't#asahi-dev · 2024-06-22 09:53
jannaumarcan: ^^^#asahi-dev · 2024-06-22 09:51
chaos_princessjannau: if you have time, would you mind merging https://github.com/AsahiLinux/m1n1/pull/384#asahi-dev · 2024-06-21 07:12
jannausomething I missed: M4 appears to have 4 thunderbolt controllers based on the pmgr ADT#asahi-dev · 2024-06-19 16:00
kettenisa few extra bytes are not going to hurt on this hardware#asahi-dev · 2024-06-19 16:00
kettenissorry, but such "optimizations" are pretty pointless#asahi-dev · 2024-06-19 13:21
zzywysmfeel free to steal#asahi-dev · 2024-06-19 13:20
zzywysmhttps://github.com/zzywysm/u-boot/commit/1ee0bb6e30de81b8589e74b6b7f6baa23808e706#asahi-dev · 2024-06-19 13:20
zzywysmkettenis: i spent some time exploring u-boot's apple_m1_defconfig inside menuconfig yesterday, and found a couple of (what i believe are) no-brainer improvements