#
✕ nick:marcan
#asahi · 2021-10-15 03:02
marcanfor arch I'm going to do both a barebones ALARM image (same spirit) and a proper desktop
#asahi · 2021-10-15 03:02
marcanwell yeah, it's gentoo, I'm assuming users will know what they're doing and know how to install the rest
#asahi · 2021-10-15 03:01
marcanthe installer will have an "update my m1n1 please" option
#asahi · 2021-10-15 03:00
marcanstage3 is an image after all, the script just automates the partitioning, then you can follow the handbook from first boot
#asahi · 2021-10-15 03:00
marcanchadmed: why not just do what we do for all the other distros and use an image?
#asahi · 2021-10-15 02:25
marcanthere's no "default" other than, er, whatever order we show them in
#asahi · 2021-10-15 02:24
marcanwhatever people want to maintain; I'm going to maintain ALARM
#asahi · 2021-10-15 02:24
marcanI just need to work out some firstboot script to copy the wifi firmware and resize the ext2fs, though we might be able to just bundle resize2fs built for mac with the installer to take care of that
#asahi · 2021-10-15 02:24
marcan_alice: the installer will do the distro install for you, the rootfs image way (which is popular these days anyway, see e.g. ALARM itself)
#asahi · 2021-10-14 17:04
marcan(then after that it turned out the bandwidth was the same either way, including on macos)
#asahi · 2021-10-14 17:04
marcan(I was going to redo it on the stream today, but I wanted to get bandwidth numbers... spent a ton of time learning that apparently the M1 is *50% slower* doing memory ops if you unroll more than 4 stp/ldp per iteration?!)
#asahi · 2021-10-14 17:03
marcanseems it is mostly useless
#asahi · 2021-10-14 17:03
marcanyes, the mcc stuff I already did
#asahi · 2021-10-14 16:53
marcanthe overall scores are within 1% or so
#asahi · 2021-10-14 16:53
marcanso far I found one geekbench subtest where it shows a ~8% performance gain, and I'm not so confident about it not being just noise
#asahi · 2021-10-14 16:53
marcanso apparently that dramcfg thing is almost entirely bullshit, lol
#asahi · 2021-10-14 13:03
marcanlol
#asahi · 2021-10-14 12:10
marcangoing to stream more MCC / cpufreq shenanigans in a bit (#asahi-stream)
#asahi · 2021-10-14 11:57
marcanfair
#asahi · 2021-10-14 11:35
marcansure, but if undefining that is an easy fix, might as well just make it a flag :)
#asahi · 2021-10-14 11:32
marcanfair
#asahi · 2021-10-14 11:32
marcan(when certain MMIO is hooked)
#asahi · 2021-10-14 11:32
marcanit's important when running macOS because otherwise timeouts happen
#asahi · 2021-10-14 11:32
marcanif that works I should just make it runtime configurable
#asahi · 2021-10-14 11:32
marcanpovik: you can undefine TIME_ACCOUNTING at the top of hv_exc.c and try
#asahi · 2021-10-14 11:31
marcanI didn't particularly care for now, but if we're doing audio it starts mattering...
#asahi · 2021-10-14 11:31
marcanmight be worth implementing :)
#asahi · 2021-10-14 11:31
marcanI think to do this properly requires paravirtualized shenanigans
#asahi · 2021-10-14 11:31
marcanpovik: ah yes, I would not be at all surprised if the HV breaks the passage of time
#asahi · 2021-10-14 10:30
marcanI guess I could probably just powergate DISP0, lol
#asahi · 2021-10-14 10:29
marcanalyssa: did you have the DISP0 regs handy somewhere? I want to just kill the framebuffer in m1n1 (to remove it from the bandwidth equation for testing)
#asahi · 2021-10-14 10:28
marcanI guess I'll have to run some m1n1 tests...
#asahi · 2021-10-14 10:28
marcan... but that means I now need memory bandwidth numbers if I don't want to fake it
#asahi · 2021-10-14 10:28
marcanso I guess *that* is what the memory controller should use?
#asahi · 2021-10-14 10:28
marcanand OPP also interops with that
#asahi · 2021-10-14 10:28
marcanso apparently there's this *other* performance config framework called "interconnects" intended for memory bandwidth and such
#asahi · 2021-10-14 10:03
marcanmight end up being the first NUMA ARM64 laptop? :p
#asahi · 2021-10-14 10:02
marcandepending on how that's implemented
#asahi · 2021-10-14 10:02
marcanthat said, there's rumors of multi-die stuff which might be... interesting
#asahi · 2021-10-14 10:02
marcanassuming it's the "M1X" then it's probably largely compatible
#asahi · 2021-10-14 10:02
marcandepends on whether this is the "M1X" or the "M2", but yeah
#asahi · 2021-10-14 09:58
marcanmostly just new DTs, adding chickens to m1n1, fixing any hardcoded device issues (e.g. if UART0 moved), and driver changes where incompatible changes happened
#asahi · 2021-10-14 09:57
marcanunless something really radical changed
#asahi · 2021-10-14 09:57
marcantbh we can probably get it booting to a terminal in a day
#asahi · 2021-10-14 09:57
marcanyeah, though expect a nice bring-up stream the day mine arrives :)
#asahi · 2021-10-14 09:24
marcannow I can largely just make all the review feedback changes in my working tree, run that and rebase -i, and be done with the next version
#asahi · 2021-10-14 09:24
marcan*very* handy
#asahi · 2021-10-14 09:24
marcanjust discovered https://github.com/torbiak/git-autofixup
#asahi · 2021-10-14 06:32
marcanof course there's also plenty of room for testing, writing docs, etc :)
#asahi · 2021-10-14 06:31
marcankernel programming isn't particularly strange as far as systems programming goes, you get used to the style fairly quickly

plus ›