04:21 <eiln> there's no evidence multi-ane systems can "load balance" a single job (model)

04:21 <eiln> identical firmwares across h13, no shared cache, DMA travels encoded at compile time

04:21 <eiln> coreml compiles all h13 the same, which says everything :P

04:21 <eiln> i predict this influenced the "disabled" ane1/ane3 & lack of ane1 in m2 max decision

04:21 <eiln> so to exploit the "double" power requires running two models simultaneously

04:21 <eiln> what is the best way to bind/manage devices in this case?

04:22 <eiln> exposing a render node for each seems straightforward, don't think it's the best though

04:38 <orowith2os> Hi all, I've been working on making a Flatpak for the mesa/asahi fork until it's upstreamed into Mesa and so far it seems to be in a mostly working state. The only problem now is publishing it, which I've looked into doing so via Flathub (Beta) and OBS, but neither seem to be well suited for a temporary package like this or are too hard to use.

04:38 <orowith2os> Would Asahi Linux be open to maintaining a Flatpak repository I can use to push to?

04:39 <orowith2os> appropriate package manifest is located here: https://github.com/orowith2os/mesa-asahi-flatpak/

04:41 <orowith2os> I would be pushing aarch64, x86_64, and i386 builds of the Mesa so that Flatpak can use x86 qemu emulation with GPU acceleration if desired

04:41 <orowith2os> Client: HexChat 2.16.1 • OS: GNOME 44 (Flatpak runtime) • CPU: AMD Ryzen 5 1600 Six-Core Processor (3.65GHz) • Memory: Physical: 30.5 GiB Total (21.2 GiB Free) Swap: 7.8 GiB Total (7.8 GiB Free) • Storage: 245.9 GB / 460.9 GB (214.9 GB Free) • VGA: Advanced Micro Devices, Inc. [AMD/ATI] Radeon R7 340 @ Advanced Micro Devices, Inc. [AMD] Family 17h • Uptime: 9h 48m 6s

04:42 <orowith2os> wooooah okay, didn't mean to do that.

04:42 <orowith2os> IRC is weird.

04:48 <orowith2os> Also, I'd appreciate it if I had some native aarch64 runners too, if a Flatpak repo is set up; aarch64 builds currently take an hour and a half to build using GitHub's runners.

06:39 <jannau> they left but I don't think offering the mesa driver as flatpak is desireable while the uapi is not stable

06:41 <jannau> drm maintainers are already wary of offering patched kernels and userspace. With distro packages kernel and mesa updates can at least synchronized in the case of UAPI changes

06:43 <jannau> with the flatpak you have no control over when different distros update their kernel and no control over users not updating or not rebooting to the new kernel

06:44 <jannau> if it finds users you will have complaints over broken software for days on UAPI changes

21:18 <XT_> Hello everyone, nice to finally check this place out, Asahi has been an irreplaceable part of using my last device, and I appreciate it immensely

21:19 <XT_> I noticed that M2 Max devices are not yet supported, which is entirely understandable given how new they are, but I just wanted to put the offer out there in case I could be of any help testing anything on my hardware

21:20 <XT_> I have the 14in 38-Core GPU M2 Max with 96GB of RAM, and I don't particularly have anything irreplaceable on it, so if it could be useful, I'd be more than happy to test on my hardware