ChanServ changed the topic of #asahi-gpu to: Asahi Linux: porting Linux to Apple Silicon macs | GPU / 3D graphics stack black-box RE and development (NO binary reversing) | Keep things on topic | GitHub: https://alx.sh/g | Wiki: https://alx.sh/w | Logs: https://alx.sh/l/asahi-gpu
capta1nt0ad has quit [Quit: Konversation terminated!]
capta1nt0ad has joined #asahi-gpu
capta1nt0ad has quit [Remote host closed the connection]
capta1nt0ad has joined #asahi-gpu
capta1nt0ad has quit []
capta1nt0ad has joined #asahi-gpu
capta1nt0ad has quit [Remote host closed the connection]
gff_ has quit [Read error: Connection reset by peer]
gff_ has joined #asahi-gpu
gff_ has quit [Read error: Connection reset by peer]
gff_ has joined #asahi-gpu
fmstrat has quit [Server closed connection]
fmstrat has joined #asahi-gpu
_whitelogger has joined #asahi-gpu
TheLink has quit [Server closed connection]
TheLink has joined #asahi-gpu
cy8aer has quit [Server closed connection]
cy8aer has joined #asahi-gpu
Yamakaja has quit [Server closed connection]
Yamakaja has joined #asahi-gpu
Method_ has quit [Server closed connection]
Method has joined #asahi-gpu
ChaosPrincess has quit [Server closed connection]
ChaosPrincess has joined #asahi-gpu
jix__ has quit [Server closed connection]
jix__ has joined #asahi-gpu
phire has quit [Server closed connection]
phire has joined #asahi-gpu
chadmed has quit [Server closed connection]
chadmed has joined #asahi-gpu
sa1 has quit [Server closed connection]
sa1 has joined #asahi-gpu
robher has quit [Server closed connection]
robher has joined #asahi-gpu
Tramtrist has quit [Server closed connection]
Tramtrist has joined #asahi-gpu
balrog has quit [Server closed connection]
mattgirv has quit [Server closed connection]
mattgirv has joined #asahi-gpu
balrog has joined #asahi-gpu
FLHerne has quit [Server closed connection]
FLHerne has joined #asahi-gpu
cylm has joined #asahi-gpu
kit_ty_kate has quit [Server closed connection]
jonmasters_ has quit [Server closed connection]
jonmasters_ has joined #asahi-gpu
kit_ty_kate has joined #asahi-gpu
_alice has quit [Server closed connection]
cylm_ has joined #asahi-gpu
_alice has joined #asahi-gpu
jbowen has quit [Server closed connection]
jbowen has joined #asahi-gpu
Venemo has quit [Server closed connection]
Venemo has joined #asahi-gpu
cylm has quit [Ping timeout: 480 seconds]
austriancoder has quit [Server closed connection]
austriancoder has joined #asahi-gpu
taowa__ has quit [Server closed connection]
taowa__ has joined #asahi-gpu
XeR has quit [Server closed connection]
XeR has joined #asahi-gpu
daniels has quit [Server closed connection]
daniels has joined #asahi-gpu
Avaflow has quit [Server closed connection]
Avaflow has joined #asahi-gpu
n1c has quit [Server closed connection]
n1c has joined #asahi-gpu
tired has quit [Server closed connection]
tired has joined #asahi-gpu
VinDuv has quit [Server closed connection]
VinDuv has joined #asahi-gpu
Swiftloke has quit [Server closed connection]
Swiftloke has joined #asahi-gpu
JTL has quit [Server closed connection]
JTL has joined #asahi-gpu
tertu has quit [Server closed connection]
yuyichao has quit [Server closed connection]
yuyichao has joined #asahi-gpu
tertu has joined #asahi-gpu
gtk2 has quit [Server closed connection]
gtk2 has joined #asahi-gpu
princesszoey has quit [Server closed connection]
steffen[m] has quit [Server closed connection]
steffen[m] has joined #asahi-gpu
princesszoey has joined #asahi-gpu
nuup has quit [Server closed connection]
nuup has joined #asahi-gpu
akemin_dayo has quit [Server closed connection]
akemin_dayo has joined #asahi-gpu
tumblingweed has quit [Server closed connection]
tumblingweed has joined #asahi-gpu
turo has quit [Server closed connection]
turo has joined #asahi-gpu
Graypup_ has quit [Server closed connection]
Graypup_ has joined #asahi-gpu
Retr0id has quit [Server closed connection]
Retr0id has joined #asahi-gpu
anuejn has quit [Server closed connection]
SSJ_GZ has joined #asahi-gpu
djorz has quit [Server closed connection]
maximbaz has quit [Server closed connection]
maximbaz has joined #asahi-gpu
LinuxM1 has joined #asahi-gpu
<LinuxM1>
I want to congratulate with Lina and Alyssa for the great work they are doing on the GPU side. I have a MacBookAir M2 and am looking forward to using GPU acceleration on Asahi Linux. I assume it will still be months before the driver is released for everyone, but I'll wait. Thank you
_jannau_ has quit [Server closed connection]
_jannau_ has joined #asahi-gpu
ah- has quit [Server closed connection]
ah- has joined #asahi-gpu
<lina>
I'm taking a look at that xorg.conf issue... I think there's a bug or something broken somewhere, because I *think* modesetting is supposed to probe all drm devices and find one with connectors.
<lina>
Ahh, no... it only does that for secondary cards.
<lina>
Aha! I found the trick ^^
<lina>
We can ship a xorg.conf OutputClass snippet that will do the right thing without having to hardcode DRI device nodes
<lina>
OK, KDE is very broken on Xorg and I think it's the vblank thing...
<lina>
Stuff just hangs for long periods of time
LinuxM1 has quit [Ping timeout: 480 seconds]
<jannau>
yes, it's the vblank issue. it "works" when switching the brightness quickly between 2 values
<lina>
The DRM documentation says the DRM core handles faking vblank events for hardware without them
<lina>
Maybe that logic works better than what dcp does?
<jannau>
I haven't found that code yet or how I tell drm that the driver doesn't support vblank irq
<lina>
just don't initialize it
<jannau>
also I think we have to do the vblank events after a swap to get fb swaps to work properly
<lina>
I think the helpers do that?
<jannau>
the default helper waits for vblank with some large timeout (iirc 300ms)
<jannau>
to use generated vblank events we would need at least control them to disable/enabled during modesets
<jannau>
if we don't use vblanks after page swaps we need something else to prevent two concurrent swaps
<lina>
concurrent swaps don't work? they don't queue?
<lina>
you could just use a waitqueue for that though
<lina>
like a completion
<jannau>
I'd rather have the the workaround affect users of the legacy kms api than the atomic kms api
<lina>
I'm not sure we even need a workaround? I think hardware without vblank is already a thing
<jannau>
we might need to to use timestamps for queued swap. without them they are "blocking" and dcp doesn't like it to start the next swap before the previous one completed
<jannau>
also I've never seen overlapping swap sequences in macos traces, it's swap_start, swap, swap_complete
<jannau>
not sure there is other complex hw supporting tear free swaps without vblank interrupts
<jannau>
advise I got in #dri-devel last weekend was to synthesize vblank events but I'm not sure if the weekend is the best time to ask this
<jannau>
drivers/gpu/drm/xen seems to do something which could match what dcp needs
<lina>
I'm tracing through the helpers now
<lina>
xen uses the simple display stuff which is too simple for DCP
<lina>
But it might be a good reference
<jannau>
I think not calling drm_vblank_init() and setting drm_crtc_state.no_vblank to true during modeset and swap might do what we need
<lina>
Working on that ^^
<jannau>
don't you have a blog post to write? I can look after it later today
<jannau>
thanks, I'll see if it's possible to remove more vblank hacks or switch back to the default drm_mode_config_helper_funcs.atomic_commit_tail
<lina>
jannau: One thing that doesn't work is that Xorg's idea of vsync is broken (e.g. glxgears claims it's vsynced but isn't) but that might be expected on drivers without explicit vsync IRQ support
<lina>
Wayland does work though
<lina>
Both with and without that patch
<lina>
But I think that's acceptable, it mostly affects some animations and is a good reason for people to eventually switch to Wayland
<lina>
This also never worked with simpledrm so it's not a regression
<jannau>
what means vsync is broken? Does it run significantly faster/slower than the display refresh rate?
<lina>
E.g. the plasma splash screen spinner is too fast, glxgears claims to be vsynced but gets 600fps anyway
<lina>
I see XORG issuing DRM_IOCTL_CRTC_{GET,QUEUE}_SEQUENCE all the time which return -EOPNOTSUPP, I get the feeling that's related
cylm_ has quit [Quit: WeeChat 3.6]
<lina>
Those are only supported if vblank IRQs are
<jannau>
ah, yes, the drm code for DRM_IOCTL_CRTC_{GET,QUEUE}_SEQUENCE made me believe that there's no support for hw without vblank irq in drm
LinuxM1 has joined #asahi-gpu
lawrence has joined #asahi-gpu
chadmed_ has quit [Remote host closed the connection]
chadmed_ has joined #asahi-gpu
chadmed_ has quit [Remote host closed the connection]
compassion has quit [Quit: lounge quit]
compassion has joined #asahi-gpu
joske has joined #asahi-gpu
<joske>
lina: thx for looking into Xorg problem. Your fix doesn't work on my system unfortunately. Still getting [ 153.858] (EE) Cannot run in framebuffer mode. Please specify busIDs for all framebuffer devices
<joske>
GNOME on wayland still works, but lightdm doesn't start, and startx also gives same error
<joske>
I'm running lastest from gpu/rust-wip and your latest oq mesa branch
<jannau>
the check for the fake mode has to happen in mode_fixup and not in atomic_check. at least when checking the mode against the modes known by dcp
<jannau>
atomic_check gets a all zero mode on disable
<lina>
Ah! Yeah, I was wondering if it was something like that
joske has quit [Remote host closed the connection]
joske has joined #asahi-gpu
<joske>
lina: OOM on X11 when playing youtube video. After about 10 seconds, desktop gone and back in gdm. Looking in journalctl I see this: https://paste.debian.net/1262036/
<lina>
joske: I can't reproduce that...
<joske>
let me try again
<lina>
You didn't mention your desktop environment or browser but I guessed them from the log
joske has quit [Remote host closed the connection]
<lina>
Please be specific...
dordoka has joined #asahi-gpu
joske has joined #asahi-gpu
<joske>
lina: reproduced in about 30 seconds
<lina>
I've been running chromium on MATE for minutes now playing YouTube videos and RAM usage is not growing
<joske>
just started playing a random youtube video in chromium in theatre mode (not fullscreen)
<lina>
also wait the oq branch why?
<joske>
did not check RAM usage, but after 30 seconds it becomes choppy and bam there goes X
<lina>
That is not a supported branch
<lina>
Please don't run random branches...
<joske>
isn't that needed on latest GPU driver?
<joske>
because of the UAPI?
<lina>
Yes and this is WIP software so sometimes branches will be out of sync...
<sven>
there’s a supported branch? :p
<lina>
There's at least asahi/main
<joske>
yes, sure, just reporting what I see broken ;-).
<lina>
But please don't run random branches...
<joske>
does asahi/main work on latest gpu/rust-wip?
<lina>
Not yet
<joske>
that's why I'm using oq ;-)
<joske>
no worries, all fine in GNOME
<lina>
This isn't released software...
<sven>
the issue is that you’re wasting everyone‘s time by running random branches and then reporting issues
<joske>
I only report issues I think would be useful for you, if you don't want me to report any issues, ok then
dordoka has quit [Quit: Page closed]
<lina>
Reporting issues in random branches is never useful...
<joske>
ok, point taken, will not report any issues I see in anything other than released packages
<Tramtrist>
Id probably make the whole thing private.. lots of noise
<joske>
sorry, that's the last thing I wanted
<lina>
Now I see why alyssa keeps her WIP branches private... she sends me a branch to implement occlusion queries just so I can add the UAPI and random people start running it T_T
zalyx has quit [Quit: later alligator]
<Tramtrist>
はい
<Tramtrist>
Its for the best
<sven>
I’ve just made my wip branches so unusable by everyone except probably jannau :D
zalyx has joined #asahi-gpu
<joske>
I think it's great that it's in the dvelopment is in the open, and that I get to try it. I won't say anything anymore
joske has quit [Quit: Leaving]
LinuxM1 has quit [Quit: Leaving]
ashi has joined #asahi-gpu
<ashi>
maybe you can reduce noise by renaming this channel to #asahi-dev-gpu or something, many people (including me) were not aware that it is complete unwanted to give any noob feedback
<jannau>
there's /topic which says "GPU / 3D graphics stack black-box RE and development (NO binary reversing) | Keep things on topic"
<ashi>
true
<dottedmag>
Maybe add a semi-transparent overlay that says "do not report bugs" to all WiP branches related to GPU? Just kidding.
<lina>
jannau: I figured out the backlight restore problem on M2 Air (and probably others) ^^
<lina>
I think it's a firmware bug... it turns out it desyncs its idea of backlight brightness on poweroff/on
<lina>
So we not only need to force a set on poweron, but also turn it off on poweroff, otherwise it just ignores the set as a no-op
<lina>
That took a while to work out... I didn't understand why it was ignoring my requests...
<lina>
Pushed to my usual branch ^^
<lina>
Along with some cleanup patches and another thing which is probably unrelated but sounded like a good idea?
<jannau>
lina: yes, I've come to the same conclusion, seems to be bug on t8103 and t8112. I've never seen seen that on t6001
<jannau>
lina: if you chain the the requests like that there no need to use out of band for dcp_set_power_state(). the true as second parameter
<jannau>
I've also noticed that on the imac macos 12.4 shuts dcp downs completely (including a MMIO CPU_RUN=0) when it turns off the display
<jannau>
I can't remember seeing that on devices with external displays or on the j314
<jannau>
lina: still broken on the imac. I suspect we need to wait after the iomfb has finished it's callbacks after the setPowerState
<jannau>
wtf, it fixed itself after a quite several seconds without setting the brightness
<jannau>
reproducible but much more than a few seconds. more like 2-3 minutes. it complains after the poweron about idac value out of range, which has to be the 0x0 fromm the poweroff
cylm has joined #asahi-gpu
<jannau>
brightness is restored after 2 muntes and 45 seconds without any AP -> DCP call
<jannau>
s/muntes/minutes/
yamii has quit [Quit: WeeChat 3.6]
yamii has joined #asahi-gpu
<jannau>
lina: pushed with small fixes to asahi-6.1-rc6-5_dcp-fixes (removed the oob setPowerState and dev_info)
<jannau>
your fix wasn't enough on an imac (maybe m1 laptops as well). I have a fix for clients which do regular swaps but I'm not sure what to for x11
<jannau>
I'll check what macos does on other devices with integrated displays when it powers the display off
<jannau>
it looks like the brightness is restored ~3 minutes after power-off. i.e. the fix works if the display was off for at least 3 mintes
<jannau>
if it was off less than 3 minutes it comes up dark and brightness is restored ~3 minutes after poweroff