#
✕ nick:jannau
#asahi-alt · 2023-03-06 13:53
jannauarnd: rebase the branch ontop of v6.3-rc1 and skip all commits with conflicts looks reasonable. I think there are slightly different commits in the tree compared to what went upstream
#asahi-alt · 2023-03-02 08:16
jannaur0ni: the patch is in the archlinuxarm package so it's no longer required in the asahi repo. see https://archlinuxarm.org/packages/aarch64/libunwind/files/PKGBUILD
#asahi-alt · 2023-02-26 14:29
jannaunot sure, quite possible that dcp does something unexpected on display power off/on. I'll try if I can find what triggers displaying the display settings in xfce4
#asahi-alt · 2023-02-26 14:27
jannauarnd: I'll try to reproduce here with xfce4 and look at the code
#asahi-alt · 2023-02-26 14:12
jannauyes, merged on 2023-01-22
#asahi-alt · 2023-02-26 14:11
jannauI think that didn't include dcp runtime pm
#asahi-alt · 2023-02-26 14:08
jannauhas you v6.1 branch the same bits/200-dcp or was it based on bits/200-dcp from around the v6.1 release?
#asahi-alt · 2023-02-26 14:07
jannauarnd: looks like runtime pm suspend and resume are running concurrently. clearly a bug in dcp. simplest workaround would be to merge bits/200-dcp only up to 125aeecbc011d
#asahi-alt · 2023-02-26 13:44
jannauhow did you get to your v6.2 based kernel? merge all the bits branches into v6.2
#asahi-alt · 2023-02-26 13:43
jannauok, sound broken. Can you get me the dcp kernel log from before switching the monitor off? I'm still on on v6.2-rc3 (asahi-wip)
#asahi-alt · 2023-02-26 13:39
jannaudoes it only happen if the monitor is off longer or can you reproduce it with switching the display off and on
#asahi-alt · 2023-02-26 13:38
jannauarnd: does switching to a different vt (and back) fixes it? I assume you have already tried to turn the monitor off and on again?
#asahi-alt · 2023-02-25 13:12
jannauhmm, you might be right. dcpext is ufp-endpoint 0. on m2 dcp is ufp-endpoint (on j473) and dcpext is ufp-endpint 1 (on all m2 machines)
#asahi-alt · 2023-02-25 13:04
jannaumight work there as well but might depend on something apple didn't get to implement in the t8103 dcp fw or doesn't work on t8103
#asahi-alt · 2023-02-25 12:28
jannauit might just work (at least on the m2 notebooks)
#asahi-alt · 2023-02-25 11:34
jannauonly difference is that dcp can be routed to the usb-c ports on the m2 mac mini
#asahi-alt · 2023-02-25 11:33
jannauthe m2 mac mini has thunderbolt 4 and the macbook air m2 has thunderbolt 3 / usb 4
#asahi-alt · 2023-02-10 19:43
jannautry `echo -n 'Y' > /sys/module/macsmc_power/parameters/log_power`, with the dev kernel that logs power usage to the kernel log
#asahi-alt · 2023-02-10 19:42
jannauit's in line with my experience. it's not constantly warm here but it is from time to time
#asahi-alt · 2023-02-10 19:37
jannauthere is nothing wrong with the kernel or distro. m1/m2 devices are with the current state very energy efficient. m1 pro/max devices not so much
#asahi-alt · 2023-02-10 14:42
jannauit seems to be a firmware bug. if the dcp is shutdown less than 3 minutes, dcp doesn't restore the backlight brightness. We update the brightness on wakeup but that doesn't work either
#asahi-alt · 2023-02-10 14:33
jannauif the issue is that the display brightness is minimal after dpms wakeup
#asahi-alt · 2023-02-10 14:32
jannaumps: probably a known issue, workarounds are using wayland or let the display at least 3 minutes in dpms off
#asahi-alt · 2023-02-10 13:17
jannauI think the microwatts values for the operating states are missing
#asahi-alt · 2023-02-09 17:13
jannauis considerably higher than on the m1 and m2
#asahi-alt · 2023-02-09 17:12
jannauenergy use on the m1 pro/max notebooks
#asahi-alt · 2023-01-26 13:58
jannauhmm, the issue the nstaller was then probably that the image size was not a multiple of 4k
#asahi-alt · 2023-01-26 13:50
jannauduring image generation is the target_blocksize used when creating the file system? either via `losetup --sector-size` or if btrfs tools support image files directly as option there
#asahi-alt · 2023-01-26 13:47
jannauthe nvme has a sector size of 4k so changing target_blocksize to fix the btrfs issue is not a solution
#asahi-alt · 2023-01-26 13:41
jannauhave you tried changing it back? 4k is afaik correct. not sure why that would cause troubles at writing the disk image though
#asahi-alt · 2023-01-26 13:38
jannauivabus: looks like EINVAL from write(), unfortunately not really helpful. should be doing 128k writes. is that reproducible? has anything changed in the way you generate the image?
#asahi-alt · 2023-01-25 12:05
jannauI doubt apple cares about 32-bit arm (anymore)
#asahi-alt · 2023-01-25 10:10
jannau... , x86 applications on ARM, ..."
#asahi-alt · 2023-01-25 10:10
jannaucy8aer: I think you misread the changelog. "This is an important milestone on the road to supporting various features such as
#asahi-alt · 2023-01-14 12:37
jannaukinetic scroll
#asahi-alt · 2022-12-31 11:45
jannaurust for linux has PR for rustc 1.66 https://github.com/Rust-for-Linux/linux/pull/947
#asahi-alt · 2022-12-31 10:09
jannausee Documentation/rust/quick-start.rst for one wayto use supported versions of rustc and bindgen
#asahi-alt · 2022-12-31 10:08
jannauRy_Darcy: your bindgen is too new. if you fix that coompilation will still fail since rustc 1.66 is too new as well
#asahi-alt · 2022-12-30 17:04
jannauarnd: Xorg and/or mesa is not supposed to crash if the GPU driver is missing (DRM_ASAHI) and it doesn't here. (with a slightly older mesa asahi snapshot though, 20221219-1)
#asahi-alt · 2022-12-29 13:05
jannaunot sure how rare it is with older SoCs
#asahi-alt · 2022-12-29 13:03
jannauncopa: yes, every HW block which might require physically continuous memory is behind an iommu
#asahi-alt · 2022-12-27 19:19
jannauI ment from a VT
#asahi-alt · 2022-12-27 19:14
jannautpw_rules: it should work but is not useful since it is vsynced
#asahi-alt · 2022-12-27 19:03
jannautpw_rules: can you try with weston for testing or glmark2-es2-drm from a tty
#asahi-alt · 2022-12-27 09:29
jannaudo you see the those low scores under X11? I'm only testing with drm/(x)wayland. can you try an off-screen benchmark? I get with `glmark2-es2-wayland -s 1920x1080 --off-screen` a score off 2171 on 24 GPU core M1 Max
#asahi-alt · 2022-12-24 09:50
jannauseems to 'enx' + mac addr
#asahi-alt · 2022-12-24 09:43
jannaujanrinze: the driver is not in a state to compare it with other hardware. also a benchmark which produces > 1000 fps is not necessary a good measure for graphics performnce
#asahi-alt · 2022-12-23 23:22
jannauthat sounds like CPU rendering, expected score is 1-2k. does the score gets worse if you increase the resolution?
#asahi-alt · 2022-12-23 22:44
jannauI wouldn't start looking at chromium but with the mesa test programs and if those work continue with glmark2
#asahi-alt · 2022-12-23 22:29
jannau`eglinfo | grep 'driver name'` should report apple and asahi

plus ›