#asahi-dev · 2024-03-09 21:15
jannegI guess the next step should be to fix the hypervisor (was it working for dberlin on t603x?) and then to find out how "iop,ascwrap-v6" differs from "iop,ascwrap-v4"#asahi-dev · 2024-03-09 21:12
jannegat least it boots with `maxcpus=1` until it tries to mount the root fs#asahi-dev · 2024-03-09 21:04
jannegthe boot cpu (4, first perf core) and only that has "{clock,memory,bus,peripheral,fixed,timebase}-frequency = 24000000;" in it's ADT node#asahi-dev · 2024-03-09 20:59
jannegI've remembered that when I read the link#asahi-dev · 2024-03-09 20:59
j`eyI reckon maz or null might have some insight here#asahi-dev · 2024-03-09 20:58
j`eynot sure, just saying it as a data point#asahi-dev · 2024-03-09 20:58
jannegbut is M3 (partially) 8.6?#asahi-dev · 2024-03-09 20:56
j`ey8.6 says it must be 1GHz#asahi-dev · 2024-03-09 20:51
janneghmm, https://developer.arm.com/documentation/ka005374/latest/#asahi-dev · 2024-03-09 20:48
jannegin m1n1#asahi-dev · 2024-03-09 20:48
jannegthat is after "correcting" CNTFRQ_EL0 to 24 MHz on the efficiency cores#asahi-dev · 2024-03-09 20:46
jannegsigh, apparently iboot configures CNTFRQ_EL0 correctly: https://paste.debian.net/1310072/#asahi-dev · 2024-03-09 13:25
chadmedj`ey: i mean our situation vis a vis firmware/kernel/userspace boundary is kinda similar to post-turing cards :p #asahi-dev · 2024-03-09 13:24
chadmedjanneg: supposedly not on nvidia. just emerged @world this evening while i was out but it applied my patch so ill try with a vanilla kwin tomorrow morning some time #asahi-dev · 2024-03-09 11:22
janneganne#asahi-dev · 2024-03-09 09:22
jannegchadmed: did it improve with 6.0.1? there was a bug letting it commit the pageflip later than intended. That has improved things here but not fully fixed it. dcp is still acting in an unexpected way#asahi-dev · 2024-03-09 08:41
j`eychadmed: so were as bad as nvidia /s#asahi-dev · 2024-03-09 04:55
chadmedjanneg: gentoo-kde folks have reproed the kwin pageflip bug on nvidia hardware too just fyi #asahi-dev · 2024-03-08 21:31
janneglooks like linux might not expect booting from cpu != 0 either#asahi-dev · 2024-03-08 21:30
jannegcpu 0: 0, 1-3: 1000000000 and 4-7: 24000000#asahi-dev · 2024-03-08 21:06
j`eyyou mean it has that value in m1n1 or?#asahi-dev · 2024-03-08 21:04
jannegvalue seems to be written by iboot on purpose though: 1000000000#asahi-dev · 2024-03-08 21:02
j`eythat value is writable.. maybe iboot was setting it before, but now m1n1 has to#asahi-dev · 2024-03-08 21:01
janneglinux isn't particularly happy: "CPU features: SANITY CHECK: Unexpected variation in SYS_CNTFRQ_EL0. Boot CPU: 0x000000016e3600, CPU1: 0x0000003b9aca00"#asahi-dev · 2024-03-08 20:56
jannegnot much to show yet except replicating/verifying dberlin's work#asahi-dev · 2024-03-08 20:41
j`eynice to see you looking at m3!#asahi-dev · 2024-03-08 20:40
janneg.revision/.version are of course unmodified#asahi-dev · 2024-03-08 20:38
jannegsigh even#asahi-dev · 2024-03-08 20:32
jannegsign, for some reason the boot_args on m3 are at least 416 bytes larger. mem_size_actual seems to be at 0x478 instead of 0x2d8#asahi-dev · 2024-03-04 14:15
j`eya bit surprising#asahi-dev · 2024-03-04 14:14
jannegm3 macbook airs but no m3 mac mini. I guess/hope that will be released with the mac studio#asahi-dev · 2024-03-03 12:13
chadmedseems like the badness is effectively on our end though #asahi-dev · 2024-03-03 12:13
chadmedtrue but id rather not have bespoke "front porch" values across my laptops and desktops :p#asahi-dev · 2024-03-03 12:08
jannegthat would be extreme for 165Hz, try 3 or 4 ms#asahi-dev · 2024-03-03 12:05
chadmedis the patch to bump it to 8ms adequate? #asahi-dev · 2024-03-03 12:04
jannegit allows 1.5 ms + vblank time. I would have expected that that is enough but apparently it is not#asahi-dev · 2024-03-03 12:01
jannegchadmed: kwin commits the pageflips too late for the current dcp behavior resulting in missing every second vblank and rendering at half the rate#asahi-dev · 2024-03-03 12:00
jannegleio: I haven't looked at the PR yet#asahi-dev · 2024-03-03 11:11
leiomaybe I should go try supertuxkart or something#asahi-dev · 2024-03-03 11:11
leiobut I haven't seen any particular problems in GNOME outside glxgears/es2gears_wayland that I wouldn't even have tested if not for it having been pointed out here#asahi-dev · 2024-03-03 11:11
leiohttps://gitlab.gnome.org/GNOME/glib/-/merge_requests/3949 didn't help for https://gitlab.gnome.org/GNOME/mutter/-/issues/3247 fwiw#asahi-dev · 2024-03-03 11:07
chadmedinteresting that it doesnt on the laptop panels though, it just stutters like crazy #asahi-dev · 2024-03-03 11:07
chadmedjust didnt notice because half the refresh rate is over 80 fps :p#asahi-dev · 2024-03-03 11:06
chadmedhm okay yeah kwin is definitely syncing at half the refresh rate on the studio #asahi-dev · 2024-03-02 13:44
chadmedi think ive mentioned before that a similar failure mode can be observed when rapidly changing the brightness#asahi-dev · 2024-03-02 13:43
chadmedjanneg: given that patch i wonder if it's related to our iomfb swap being really slow?#asahi-dev · 2024-03-02 13:42
leiothe numbers go especially down the drain if I jiggle the mouse around with no hardware cursor plane in use#asahi-dev · 2024-03-02 13:42
leiolooks like I can see https://gitlab.gnome.org/GNOME/mutter/-/issues/3247 too, just haven't really noticed on anything in practice, which might have more to do with not having steam than it not being a problem#asahi-dev · 2024-03-02 13:38
leio(the developer is on asahi)