#asahi-dev · 2026-06-10 08:22
StanfordAppliedCyber[m]good news: with the exception of possibly audio i think they all have graceful non exclave fallback paths#asahi-dev · 2026-06-10 08:21
StanfordAppliedCyber[m]dcp, audio, some mtp stuff#asahi-dev · 2026-06-10 08:21
StanfordAppliedCyber[m]LOL i had a message sent on that exact thing but forgot to hit enter#asahi-dev · 2026-06-10 08:15
yukawhich things are put into exclaves#asahi-dev · 2026-06-10 08:15
yukaStanfordAppliedCyber: can you elaborate?#asahi-dev · 2026-06-10 08:12
StanfordAppliedCyber[m]also, the amount of stuff put into exclaves is fucking insane#asahi-dev · 2026-06-10 08:12
StanfordAppliedCyber[m]checking the two "can we use 26.5 as our a18 pro firmware version" blockers now#asahi-dev · 2026-06-09 21:30
flokli:+1:#asahi-dev · 2026-06-09 21:29
StanfordAppliedCyber[m]whenever i'm back to my mac i'll see if it loses register state#asahi-dev · 2026-06-09 21:29
StanfordAppliedCyber[m]unfortunate#asahi-dev · 2026-06-09 21:23
flokli26.5.1#asahi-dev · 2026-06-09 21:22
StanfordAppliedCyber[m]what firmware version? 26.5?#asahi-dev · 2026-06-09 21:21
flokliStanfordAppliedCyber[m]: this is a boot which oopses on a secondary cpu as it's loosing state presumably: https://paste.linuxlounge.net/OE#asahi-dev · 2026-06-09 20:06
StanfordAppliedCyber[m]we'll see, maybe i'm simply hopelessly naive#asahi-dev · 2026-06-09 20:04
svenjust getting the gpu work downstream is gonna take a lot of time#asahi-dev · 2026-06-09 20:04
StanfordAppliedCyber[m]i never said i'd upstream anything but the hypervisor#asahi-dev · 2026-06-09 20:04
StanfordAppliedCyber[m]:)#asahi-dev · 2026-06-09 20:03
StanfordAppliedCyber[m]presumably that's the metal IR that gets compiled to the actual shader isa once we know exactly what platform it's running on?#asahi-dev · 2026-06-09 20:03
svenso just a quick 2-3 year project ;p#asahi-dev · 2026-06-09 20:03
StanfordAppliedCyber[m]for a bit of context on me my goal is to upstream the hypervisor, get the machine to a daily driver usable state without the GPU and DCP, then try to add the GPU, then likely head to another project that is not linux related#asahi-dev · 2026-06-09 20:02
chaos_princesswhen using gpus, you don't ship compiled code, instead use some kind of IR. there is an undocumented something called AIR that some macos games ship#asahi-dev · 2026-06-09 20:02
StanfordAppliedCyber[m]yeah that's my current understanding, thanks!#asahi-dev · 2026-06-09 20:02
chaos_princessso, m1, m3, m5#asahi-dev · 2026-06-09 20:02
chaos_princessisa changes roughly every other chip generation#asahi-dev · 2026-06-09 20:01
StanfordAppliedCyber[m]presumably userspace, especially for the gpu drivers, is very stable however right? ie, the actual shader isa and stuff won't change#asahi-dev · 2026-06-09 20:01
svenand even if that abi was stable we can’t update firmware (or specifically iboot where they have to fix it) from linux. that might even need a trip to 1TR#asahi-dev · 2026-06-09 20:00
svenyeah :(#asahi-dev · 2026-06-09 19:56
StanfordAppliedCyber[m]insane, unstable, firmware abi :)#asahi-dev · 2026-06-09 19:56
StanfordAppliedCyber[m]fair, somehow i blanked on the whole "INSANE FIRMWARE ABI" thing#asahi-dev · 2026-06-09 19:55
svenand since we can’t easily update firmware we don’t want that as a downstream only hack either #asahi-dev · 2026-06-09 19:55
svengood luck getting that upstream #asahi-dev · 2026-06-09 19:54
StanfordAppliedCyber[m]ACK, is the wfi loses register state thing a blocker for specifically asahi linux? ie there's an easy hack, save all registers, restore all registers#asahi-dev · 2026-06-09 19:53
svenwould just be good to file that feedback report thing soon because getting that fixed is probably gonna take a long while again #asahi-dev · 2026-06-09 19:53
svenno worries #asahi-dev · 2026-06-09 19:52
StanfordAppliedCyber[m]busy rn so cant check immediately#asahi-dev · 2026-06-09 19:51
StanfordAppliedCyber[m]ack, i'll check both sometime today and report back on the neo#asahi-dev · 2026-06-09 19:51
svenand probably file a feedback report thing about wfi if it still loses state #asahi-dev · 2026-06-09 19:51
StanfordAppliedCyber[m]i know it was not accessible on 26.4.1 but i actually never checked 26.5 because the code i wrote for 26.4.1 just kept working on 26.5#asahi-dev · 2026-06-09 19:51
svenwe’ll have to wait for a version where those things are fixed #asahi-dev · 2026-06-09 19:50
StanfordAppliedCyber[m]sven: let me check that now#asahi-dev · 2026-06-09 19:50
StanfordAppliedCyber[m]sven: loses register state last i checked but i never checked on 26.5#asahi-dev · 2026-06-09 19:47
svenand does wfi work correctly there or does it lose register state?#asahi-dev · 2026-06-09 19:46
svenonly if that aic register is accessible there #asahi-dev · 2026-06-09 19:46
StanfordAppliedCyber[m]for the m4, well, probably the same tbh because i'm the first one to look at it and i'm using 26.5 --> fair chance the version we use is 26.5#asahi-dev · 2026-06-09 19:45
StanfordAppliedCyber[m]flokli: just update to 26.5 on the neo#asahi-dev · 2026-06-09 17:02
Ardu0us_Yeah probably at this point. 7.1 is slated to be released mid to late June, and 7.2 in late August.#asahi-dev · 2026-06-09 16:58
Tachyonray[m]you’ll probably skip over 7.0#asahi-dev · 2026-06-09 16:57
Ardu0us_I'm unsure Tachyonray[m]. Judging by the release cadence found here, I'd guess 7.x is coming relatively soon. https://copr.fedorainfracloud.org/coprs/g/asahi/kernel/builds/#asahi-dev · 2026-06-09 16:47
Tachyonray[m]or is stable always based on the lts kernel release?#asahi-dev · 2026-06-09 16:46
Tachyonray[m]i’d expect the asahi repo’s to package 7.0.11 as well