#
…

2026-10-28

00:11hdbngr has quit [Ping timeout: 480 seconds]
00:12hdbngr has joined #asahi-dev
00:20hdbngr has quit [Ping timeout: 480 seconds]
00:28hdbngr has joined #asahi-dev
00:36hdbngr has quit [Ping timeout: 480 seconds]
00:40fatcullen has joined #asahi-dev
00:44hdbngr has joined #asahi-dev
00:52hdbngr has quit [Ping timeout: 480 seconds]
00:56millefy147 has joined #asahi-dev
00:56millefy14 has quit [Ping timeout: 480 seconds]
00:56millefy147 is now known as millefy14
00:58hardy94 has quit [Remote host closed the connection]
01:01hdbngr has joined #asahi-dev
01:09hdbngr has quit [Ping timeout: 480 seconds]
01:19hdbngr has joined #asahi-dev
01:28hdbngr has quit [Ping timeout: 480 seconds]
01:36hdbngr has joined #asahi-dev
01:45hdbngr has quit [Ping timeout: 480 seconds]
01:47ohxorud has quit [Read error: Connection reset by peer]
01:54hdbngr has joined #asahi-dev
02:01tjemke770 has joined #asahi-dev
02:02hdbngr has quit [Ping timeout: 480 seconds]
02:08tjemke77 has quit [Ping timeout: 480 seconds]
02:08tjemke770 is now known as tjemke77
02:10hdbngr has joined #asahi-dev
02:11linkmauve has quit [Remote host closed the connection]
02:11linkmauve has joined #asahi-dev
02:19hdbngr has quit [Ping timeout: 480 seconds]
02:23linkmauve has quit [Remote host closed the connection]
02:27hdbngr has joined #asahi-dev
02:31Lyude has quit [Quit: Bouncer restarting]
02:31linkmauve has joined #asahi-dev
02:35hdbngr has quit [Ping timeout: 480 seconds]
02:43hdbngr has joined #asahi-dev
02:45qyliss has quit []
02:59hdbngr has quit [Ping timeout: 480 seconds]
03:00yurix has quit [Remote host closed the connection]
03:00yurix has joined #asahi-dev
03:00fatcullen has quit [Quit: This computer has gone to sleep]
03:06fatcullen has joined #asahi-dev
03:06hdbngr has joined #asahi-dev
03:15hdbngr has quit [Ping timeout: 480 seconds]
03:21betoissues has quit [Quit: bye bye.]
03:23hdbngr has joined #asahi-dev
03:32hdbngr has quit [Ping timeout: 480 seconds]
03:35betoissues has joined #asahi-dev
03:40hdbngr has joined #asahi-dev
03:48hdbngr has quit [Ping timeout: 480 seconds]
03:52hdbngr has joined #asahi-dev
03:56tobhe_ has joined #asahi-dev
04:01hdbngr has quit [Ping timeout: 480 seconds]
04:02tjemke776 has joined #asahi-dev
04:02tobhe has quit [Ping timeout: 480 seconds]
04:08Lyude has joined #asahi-dev
04:08tjemke77 has quit [Ping timeout: 480 seconds]
04:08tjemke776 is now known as tjemke77
04:08hdbngr has joined #asahi-dev
04:17hdbngr has quit [Ping timeout: 480 seconds]
04:26hdbngr has joined #asahi-dev
04:35hdbngr has quit [Ping timeout: 480 seconds]
04:44hdbngr has joined #asahi-dev
04:47qyliss has joined #asahi-dev
04:53hdbngr has quit [Ping timeout: 480 seconds]
05:01hdbngr has joined #asahi-dev
05:02nora has joined #asahi-dev
05:03tbodt has quit []
05:04nora_ has quit [Ping timeout: 480 seconds]
05:05tbodt has joined #asahi-dev
05:09hdbngr has quit [Ping timeout: 480 seconds]
05:12Lyude has quit [Quit: Bouncer restarting]
05:15fatcullen has quit [Quit: This computer has gone to sleep]
05:18hdbngr has joined #asahi-dev
05:21fatcullen has joined #asahi-dev
05:24fatcullen has quit []
05:27Lyude has joined #asahi-dev
05:33hdbngr has quit [Ping timeout: 480 seconds]
05:42hdbngr has joined #asahi-dev
05:51hdbngr has quit [Ping timeout: 480 seconds]
05:53hdbngr has joined #asahi-dev
06:01Pokemod97736208471281833001325 has quit []
06:07evancarroll[m] has joined #asahi-dev
06:08hdbngr has quit [Ping timeout: 480 seconds]
06:10evancarroll[m]I have some questions about AMX. What is the obstacle to getting that turned on? I'm a contributor to the tract project, and I see they support AMX instructions which are hand-rollled. https://github.com/sonos/tract/blob/main/linalg/arm64/apple_amx/apple_amx_mmm_f16_64x32.S.j2
06:10evancarroll[m]If Asahi set the register to enable AMX (even if only with a kernel boot param), and we submitted a patch for lazy context-switching (preserving the coprocessor's state around context switching only when the amx instruction where hit), what else is required?
06:10evancarroll[m]I mean to say, we don't actually need the kernel to support AMX's ISA. We just need it turned on and the hand rolled instructions in tract would just work?
06:10evancarroll[m]* AMX instructions on OSX which are
06:12JamesCalligeros[m]We will never enable AMX. It is a proprietary ISA extension that has no business in upstream Linux. Newer machines support SVE and SME, which are the replacements for AMX.
06:15evancarroll[m]Arm itself is a propeietary ISA... Shouldn't the only quesiton be whether or not it's documented and in use? https://github.com/corsix/amx
06:16svenpeter[m]upstream has already rejected the TSO bit, AMX is much more invasive
06:16hdbngr has joined #asahi-dev
06:20jcbm has quit [Remote host closed the connection]
06:21jcbm has joined #asahi-dev
06:22Pokemod97736208471281833001325 has joined #asahi-dev
06:25hdbngr has quit [Ping timeout: 480 seconds]
06:33hdbngr has joined #asahi-dev
06:35ddxtanx has quit [Quit: Konversation terminated!]
06:38mattia013 has joined #asahi-dev
06:38JamesCalligeros[m]Its a proprietary isa with an architecture manual and standard extensions that are expected to be used. Apple are free to do whatever they want in their walled garden per their architecture licence, we however are not.
06:42hdbngr has quit [Ping timeout: 480 seconds]
06:44sven has quit [Quit: The Lounge - https://thelounge.chat]
06:45sven has joined #asahi-dev
06:46umo is now known as Guest4135
06:46umo has joined #asahi-dev
06:49hdbngr has joined #asahi-dev
06:53Guest4135 has quit [Ping timeout: 480 seconds]
07:03hdbngr has quit [Ping timeout: 480 seconds]
07:04lacamar_ has joined #asahi-dev
07:10lacamar has quit [Ping timeout: 480 seconds]
07:13hdbngr has joined #asahi-dev
07:21hdbngr has quit [Ping timeout: 480 seconds]
07:23bgb has quit [Remote host closed the connection]
07:29hdbngr has joined #asahi-dev
07:29yukajannau: nice, now the logging also works in the experiment
07:29bgb has joined #asahi-dev
07:33mattia013 has quit [Remote host closed the connection]
07:37hdbngr has quit [Ping timeout: 480 seconds]
07:41yukaoeeeehhh
07:41yukaI don't know how I missed this before
07:41yukamacOS: SetShmem [...] FLAG=0x4
07:42yukaexperiment: SetShmem [...] FLAG=0x0
07:42yukanow trying in linux
07:42chadmedauuughhhhh
07:43yukaexperiment looks like this now: https://termbin.com/823j
07:43umo has quit [Remote host closed the connection]
07:43yukathis looks better right? ;)
07:43umo has joined #asahi-dev
07:43svenjust a little bit ;)
07:44hdbngr has joined #asahi-dev
07:45JamesCalligeros[m]Why cant they just
07:46yukanow need to find the place where this message is actually sent in the kernel
07:48yukathe kernel already sets flag to 4
07:49yukaSo then the kernel is failing in another way
07:49yukawill trace it again
07:50yukayes: linux: SetShmem [...] FLAG=0x4
07:51yukathat's why I didn't see it before, there was no difference!
07:53yukastill, I now have a working-ish experiment?
07:53yukaSo I can compare experiment and linux
07:53hdbngr has quit [Ping timeout: 480 seconds]
07:55hdbngr has joined #asahi-dev
08:04hdbngr has quit [Ping timeout: 480 seconds]
08:04yukaI also needed iova_range=(0x007800000, 0x40000000), where did that come from?
08:04yuka(for the experiment)
08:06jannauthe iova allocator from dart.py. the default overlaps with already allocated iova
08:06jannauI might have been misguided by crashlog buffer though
08:09yukaHope the setshmem flag difference helps if you want to try again
08:10jannauyuka: dcp works in linux with pd_ignore_unused and "apple,min-state = <4>;" sprinkled over more disp power-states
08:10yukaon t603[124]?
08:10jannauon t6031
08:11chadmedso are we just not turning something on properly
08:12jannauno, we turning something off too much. pd_ignore_unused is not required
08:12hdbngr has joined #asahi-dev
08:13yukagot it working as well
08:13yukajust moved the min-state from _fe to _cpu
08:14jannauack. that's the minimal change that makes it work here as well
08:15yukaAnd I also changed the dependencies of dart / dcp to _fe since that's now the "top-level" one
08:15yukalet's see if dcpext also works
08:18yukadcpext crashes the kernel
08:18yukaor well
08:18yuka> Unable to handle kernel NULL pointer dereference at virtual address 0000000000000098
08:19chadmedcan you paste the full dump
08:19jannauI see occasional null pointer derefs under the HV because the piodma subdevice fails to initialize
08:19jannauunfortunately hard to reproduce
08:19yukaah yes this is under hv
08:19yukalet me try without
08:20chadmedoh ok not what i thought it might be
08:20chadmedprobably hv shenanigans if it cant even boot the system ep
08:21jannauyuka: "apple-dcp 2cac00000.dcp: error 0000000000000000: Failed to created PIODMA iommu child device" either try again or there's something wron in the dt
08:21yukaunder not-hv
08:22hdbngr has quit [Ping timeout: 480 seconds]
08:27MichaelReeves[m]jannau: just saw your ping of me yesterday, sorry I haven't had a chance to finish preparing the 7.3 wifi branch yet, tomorrow is the last day of exams so I should be able to do it then
08:29Bhyve has joined #asahi-dev
08:31chadmed has quit [Read error: Connection reset by peer]
08:31chadmed_ has joined #asahi-dev
08:31hdbngr has joined #asahi-dev
08:32chadmed_ is now known as chadmed
08:33jannauMichaelReeves[m]: no problem. I need to look more closely at your series. it might break bcm4387 (M1 Pro / Max devices)
08:36MichaelReeves[m]the way I did it, it ideally shouldn't, but I might have messed something up by accident as I can't test it on bcm4387. sashiko raised a potential bug but im not sure if it was hallucinating, I guess it needs to be tested on real hardware
08:38jannauthis is from testing on hw, wifi did not work but I haven't looked at it at all
08:44hdbngr has quit [Ping timeout: 480 seconds]
08:47MichaelReeves[m]ok, I'll have to look more into the identified "bugs" it says for older firmware and in the branch I'll fix any if they are real. if you have time, could you also share dmesg?
08:51hdbngr has joined #asahi-dev
09:00yukaI think that's failing because some dart nodes are set to status disabled
09:00yukadcpext is also disabled but it's reenabled by m1n1
09:03bgb has quit [Remote host closed the connection]
09:08bgb has joined #asahi-dev
09:09yukaI'm confused how these are enabled on t602x
09:11jannausofus: maybe we should rename libva_v4l2_request to asahi-va-driver? that way we can install it as asahi_drv_video.so and libva will use it automatically (based on the DRM driver name). we also do not clash with the existing libva_v4l2_request package (on fedora for example)
09:11jannauinspired by https://github.com/radxa-pkg/qcom-va-driver which is libva_v4l2 under different name
09:12jannauwe could of course just install an asahi_drv_video.so symlink
09:14chaos_princessif we intend to upstream our changes to libva_v4l2_request, we should probably go the symlink route
09:15jannauyuka: I think the darts are not disabled. the dcpext are enabled via aliases
09:16jannauyuka: for usable dcpext your diff is missing to add dispext* to 'display's "iommus"
09:19chaos_princessin that case, yea, let
09:19chaos_princess's just rename
09:24yukahmm, not sure what I'm doing wrong then
09:25yukastill getting NULL pointer dereference
09:29yukajannau: {dcp,disp}ext*_dart all have status = "disabled"; in arch/arm64/boot/dts/apple/t602x-dieX.dtsi
09:30yukaAha I guess m1n1 only enables the dcpext itself if there's an alias for it
09:30yukaand in the same file where the alias for dcpext0 is added, there's also an override to enable the darts
09:30yukaso we want to actually not add aliases for the unused dcpexts
09:31jannauyes, aliases just for the used dcpext
09:32yukado you have a preference on _cpu vs _fe for power-domains?
09:32yukaI think both will work since _fe is currently always-on
09:33yukaI guess on the dcpexts it might make a difference
09:33jannaulet's try to replicate what's in the ADT and deviate from it only if that does not work
09:34jannaua comment explaining differences to all other devices is helpful
09:34yukathat would be _cpu then
09:37yukaI still don't get how dcpext0_dart is enabled on the M2 Pro/Max MBP
09:37yukaIt includes the M1 Pro/Max dtsi
09:37yukaI keep forgetting that
09:37JamesCalligeros[m]i forget this sometimes too
09:43yukaAh shit we have dcpext on t6030 vs dcpext0 on t603[124]
09:43yukaso t603x-j514-j516.dtsi will be a mess
09:43yukafor hdmi
09:45jannauyuka: just rename things to them more consistent. but t6030 should have dcpext0 / dcpext1
09:45yukaAh right the nodes were just not added in that PR
09:46yukado we maybe want to just add minimal dcp for the internal display and stabilize that first, and then add the dcpext nodes later?
09:47jannaut8122 however will only have dcpext but doesn't conflict with t603x-j514*
09:47chaos_princessdcpext nodes would be useful for testing the new atcphy dp
09:49jannauyes. that my plan. I applied #608 locally for 7.3 and have some small fixes/changes
09:49yukaOK then I'll just wait for that :)
09:49chaos_princessi want to send the entire m3 atcphy series shortly after the merge window close
09:51jannauthe dcpext nodes can be added, diabled and without aliases. I think #608 has at least some dcpext nodes
09:51yukaYes, it has dcpext nodes but also aliases which should be removed until it's properly integrated
09:53jannauindeed. I did not run into issues since I enabled just dcp in m1n1
10:06bgb has quit [Remote host closed the connection]
10:06bgb has joined #asahi-dev
10:52umo has quit [Remote host closed the connection]
10:52umo has joined #asahi-dev
11:06yukais it possible that suspend is broken on 7.2 / asahi-wip?
11:07yukaI noticed it on my M1 Pro
11:07yukaOr at least waking up the display
11:13yukadcp works well on t6034
11:13yukaalso very good performance ^^
11:17kennw has quit [Ping timeout: 480 seconds]
11:17kennw has joined #asahi-dev
11:21chadmedlets gooo
11:22jannauI haven't noticed suspend issues but I test mostly under hv and with no_console_suspend
11:26Bhyve has quit []
11:46fatcullen has joined #asahi-dev
12:25jcbm has quit [Ping timeout: 480 seconds]
12:35jcbm has joined #asahi-dev
12:40kaazoo has joined #asahi-dev
12:42kaazoo has quit []
12:59fatcullen has quit [Quit: This computer has gone to sleep]
13:26chadmed has quit [Quit: Konversation terminated!]
13:26chadmed has joined #asahi-dev
13:28chadmed has quit []
13:28chadmed has joined #asahi-dev
13:49meow676767676 has quit [Quit: The Lounge - https://thelounge.chat]
14:11salva has joined #asahi-dev
14:14chadmed has quit [Quit: Konversation terminated!]
14:15chadmed has joined #asahi-dev
14:59jannauyuka: re suspend on t6000/t6001. check that pmgr_misc has a "apple,dcs-min-ps" property. I did break that in asahi-wip for 7.2
15:27hardy94 has joined #asahi-dev
15:32msimonelli22 has quit [Read error: Connection reset by peer]
15:32msimonelli22 has joined #asahi-dev
15:37aditya has joined #asahi-dev
15:43umo has quit [Read error: Connection reset by peer]
15:46umo has joined #asahi-dev
16:03meow676767676 has joined #asahi-dev
16:12svenyay, looks like we’re (usb4 maintainer, dt-binding maintainer, qcom and us) in agreement to use graph bindings for usb4/tbt now
16:12sventime to get back to that series now I guess
16:19umo is now known as Guest4181
16:19umo has joined #asahi-dev
16:27Guest4181 has quit [Ping timeout: 480 seconds]
16:31svendoes anyone remember where arm_regs.json came from? I need to add two sysregs but figured we could just update to whatever the latest ARM ARM has anyway
16:31svencan't find a source though
16:32yukaI could swear a nested virtualization talk by maz mentioned several (semi-)open sources for this kind of info
16:34svenweb search is just so damn broken these days :(
16:34yuka> ARM publishes two sets of documents, released quarterly, under the AARCHMRS name:
16:34sventhat keyword might be useful. let's see
16:34yukathe slide also has a link
16:35yukayes that one
16:37caszue has quit [Server closed connection]
16:37caszue has joined #asahi-dev
16:38mazyuka: I have a tree with all the BSD-licensed JSON files at git://git.kernel.org/pub/scm/linux/kernel/git/maz/AARCHMRS
16:42k70 has quit [Server closed connection]
16:43k70 has joined #asahi-dev
16:45yukasven also see reg2json.py
16:45yukain m1n1/tools/
16:45svenoh, lol, nice. I’m heading out for today but will take a look tomorrow
16:49ieyasu has quit [Server closed connection]
16:49ieyasu has joined #asahi-dev
16:50jannaujust in time for releasing and publishing about 7.2.12 with usb4 support
16:50lethalbit has quit [Server closed connection]
16:50lethalbit has joined #asahi-dev
16:53ohtwoone has quit [Server closed connection]
16:53mattia013 has joined #asahi-dev
16:54ohtwoone has joined #asahi-dev
16:55sofusjannau: thanks will update the name and tag a new version ~soon. I might try upstreaming again if i see any activity
16:56Tramtrist has quit [Server closed connection]
16:57bgb has quit [Remote host closed the connection]
16:57Tramtrist has joined #asahi-dev
17:03bgb has joined #asahi-dev
17:09ephemera_ has joined #asahi-dev
17:09Yushi has quit [Remote host closed the connection]
17:10apc has quit [Server closed connection]
17:10apc has joined #asahi-dev
17:10Yushi has joined #asahi-dev
17:10jannauMacbook Pro with PROLED XDR Display
17:12yukatouchscreen 🥳👯
17:12jannau320g lighter
17:14jannauno 5G modem
17:15myria has quit [Server closed connection]
17:16myria has joined #asahi-dev
17:18jannau1000 nits full screen brightness, VRR from 10 to 120 HZ
17:18dud has quit [Remote host closed the connection]
17:20jannauand most importantly illuminated Apple logo on the back of the display
17:20chadmed has quit [Quit: Konversation terminated!]
17:21chadmed has joined #asahi-dev
17:22kit_ty_katestarting at $3k yeesh
17:24chaos_princess(please keep the lcd versions)
17:27kennw has quit [Server closed connection]
17:27kennw has joined #asahi-dev
17:33yukakit_ty_kate almost 4k€
17:36yukajannau I'm still having issues with suspend on M1 Pro, despite the latest 7.2.13-1 and it does have the pmgr_misc property
17:36yukalet me try to get loghs
17:38yukano in fact it does not wake up at all
17:39yukaor is at least not reachable via network
17:43francisssssssss[m]jannau: is it? i don't see it illuminated in any material
17:46aditya has quit [Quit: Connection closed for inactivity]
17:52jannauit is on https://www.apple.com/macbook-pro/proled-xdr-display/ under the "Everything to see here" heading
18:04jannauA mac studio with 512GB RAM costs a little more than 2 256GB ones. 2 * €12,430 vs €25,080
18:05opticronoh wow, they've dropped the notch with that one, too (now just a hole punch)
18:05francisssssssss[m]i prefer the notch icl
18:06francisssssssss[m]island looks a little ugly in that shape
18:06yukaeverything the leaks promised :P
18:06francisssssssss[m]i don't like reading leaks :P
18:10Ardu0us has quit [Quit: %Reality broken: (I)gnore, (R)etry, (A)bort%]
18:10jannauexcept for 5G connectivity
18:11Ardu0us has joined #asahi-dev
18:13svenwhich is the only thing I wanted :-(
18:14Bhyve has joined #asahi-dev
18:26jannaumaybe you should ask for macos on ipad pros (now that macos has touch support)
18:47tryhard[m]kinda weird they didn't make the laptop bendable to a tablet
19:41jannauthey rather sell you an ipad pro
19:53veloek_ has joined #asahi-dev
19:54chadmed has quit [Read error: Connection reset by peer]
19:54chadmed has joined #asahi-dev
19:54veloek has quit [Ping timeout: 480 seconds]
19:55lacamar has joined #asahi-dev
19:56chadmed has quit []
19:56chadmed has joined #asahi-dev
19:59lacamar_ has quit [Ping timeout: 480 seconds]
20:00hardy1994 has joined #asahi-dev
20:03hardy94 has quit [Ping timeout: 480 seconds]
20:04hardy94 has joined #asahi-dev
20:08hardy1994 has quit [Read error: Connection reset by peer]
20:10Bhyve has quit []
20:17nicolas17 has joined #asahi-dev
20:24veloek_ is now known as veloek
20:25PenelopeGwen[mds] is now known as Pen[mds]
20:27tryhard[m]just how touchscreen is a pro when you can't bend the laptop... who wants to hold their hands up fatiguing your shoulders just so you can play piano
20:28nicolas17literally qwerty is better than touchscreen for piano
20:28tryhard[m]waste of money for them and for the customers
20:31nicolas17apple published a second "how our new hardware will make your life harder" set of videos for app developers https://developer.apple.com/touch-on-mac/
20:34chaos_princessthat thing feels like pure downsides - oled screen (those do not belong on computers), the dynamic island is harder to hide than the notch, and a useless touch digitizer (touch does not belong on not 2in1 laptops)
20:35nicolas17higher price
20:41jannauI think the display got 22 additional lines, I hope that means below the dynamic island is still space for an unobstructed 3024x1890 16:10 display. if that's true hiding it shouldn't be an issue
20:42jannauharder to use the pixels left and right from it though
20:42chaos_princessthe "dynamic" part is the problem, you can hide the notch by just adding a black bar at the top of the wallpaper, but since macos will put various app nonsense there, it can't be hidden as well
20:43chaos_princesson linux it is easy, just carve it out
20:47chadmedcan we just not support the touchscreen as a matter of principle :P
20:48chaos_princessit is probably just z2 yet again
20:48chaos_princesswhy wouldn't it be
20:48chadmedstatus = "disabled"; /* this sucks you dont need it */
20:48travltux[mds]I don't even use the touchscreen on my work lenovo, I like a clean screen without fingerprints :)
20:50jannauah, I care very little about macOS
20:50yukaM3 Max dcp crashes when my compositor tries to turn off the screen
21:08bgb has quit [Remote host closed the connection]
21:13bgb has joined #asahi-dev
21:21jannauyuka: which compositor? can't reproduce with kwin
21:38jannauyuka: I pushed my tree (mostly #608, a little #607 but edited by me) to asahi-wip maybe you spot a difference
21:39jannaumy dcp syslog looks much alike except that there are a additional lines instead of the crash
21:44Lyude has quit [Quit: Bouncer restarting]
21:49Lyude has joined #asahi-dev
21:57Chat7dafdbHelper has joined #asahi-dev
22:01Chat7dafdbHelper has quit [Remote host closed the connection]
22:02Chat7dafdbHelper has joined #asahi-dev
22:12Chat7dafdbHelper has quit [autokilled: This host violated network policy. Contact support@oftc.net for further information and assistance. (2026-10-28 22:12:45)]
23:05hub27 has joined #asahi-dev
23:07hub2 has quit [Read error: Connection reset by peer]
23:07hub27 is now known as hub2
23:30iwuvmika has joined #asahi-dev
23:30rivendell has quit [Remote host closed the connection]
23:31rivendell has joined #asahi-dev
23:31amarioguy has joined #asahi-dev
23:43rivendell has quit [Remote host closed the connection]
23:43rivendell has joined #asahi-dev
23:46hardy1994 has joined #asahi-dev
23:47amarioguy has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
23:49amarioguy has joined #asahi-dev
23:51hardy94 has quit [Ping timeout: 480 seconds]
23:51hardy1994 is now known as hardy94
…