#asahi · 2022-03-02 04:08
chadmedthere you go try that#asahi · 2022-03-02 04:08
chadmedoop one second#asahi · 2022-03-02 04:06
chadmedthe configs here should work for all apple silicon macs released thus far, so long as you pair them with a gzipped initramfs#asahi · 2022-03-02 04:05
chadmedtpw_rules: https://github.com/chadmed/asahi-gentoosupport#asahi · 2022-03-02 01:36
chadmedby all means go ahead#asahi · 2022-03-02 01:29
chadmedobviously the long term goal is to get everything important upstreamed, but this should do nicely until then #asahi · 2022-03-02 01:29
chadmedmy hope is that they behave almost identically to a kernel+initramfs youd get from any distro, so that the kernel itself behaves as a user would intend it to. ill have these up on my github at some point this week, and then eventually i'll ship them with a shell script that builds them and an initramfs for ALARM and its flavours and one for gentoo as well #asahi · 2022-03-02 01:27
chadmedive been working on a couple of .configs that behave like generic binary distro kernels but trimmed down to support only these machines to go with linux-asahi. one for gcc and one for clang/thinlto for people that way inclined #asahi · 2022-03-01 02:13
chadmedespecially if we ever get VRR so the screen stops refreshing when theres nothing happening on it #asahi · 2022-03-01 02:12
chadmeditll be far nicer once we can offload display and rendering stuff to the dcp and gpu respectively #asahi · 2022-03-01 02:11
chadmedyeah its great, even using llvmpipe the battery life in linux is fantastic now with working cpufreq#asahi · 2022-02-28 02:13
chadmedmost of the problems you have described can be fixed by just following the guides#asahi · 2022-02-28 02:12
chadmedputust[m]: please read the wiki, specifically https://github.com/AsahiLinux/docs/wiki/Developer-Quickstart-2.0 and https://github.com/AsahiLinux/docs/wiki/DesktopKernel. trying to take shortcuts or doing it without having read these pages will just result in headaches#asahi · 2022-02-27 07:11
chadmedhmmm that might be nice to have a graphical installer written for Cocoa at some point down the road ;)#asahi · 2022-02-27 01:54
chadmedwelp uni is cancelled for the next week because half the state is flooded, whod be interested in a gentoo overlay and install shell script#asahi · 2022-02-25 10:07
chadmedbut thatd be something rosetta does, not the kernel#asahi · 2022-02-25 10:07
chadmedyeah it does, otherwise binaries compiled for amd64 assuming 4k pages wouldnt work. but those pages still need to be mapped into memory by the kernel and addressable by the mmu, which means you either need to pad a 16k page with 12k worth of 0s or pack 4 pages into a 16k page and track which 4k page is which somewhere else in the process's memory#asahi · 2022-02-25 09:59
chadmedi mean its not a kernel thing, its a malloc/jemalloc/tcalloc thing. it would be nice if they had support for packing smaller pages into large ones, i.e. put 4x 4k pages into a single 16k page and have a lookup table in memory of the addresses corresponding to each 4k page, which i assume is what apple does for rosetta 2? #asahi · 2022-02-25 09:42
chadmedi left my spell checker in my other pants#asahi · 2022-02-25 09:41
chadmedas j`'ey said though, jemalloc compiled for larger page sizes works on systems with smaller page sizes than it was compiled for (but not vise versa) so if you could prove that there is no performance penalty you might be able to convince distros to at least patch their versions for 16kB pages on aarch64#asahi · 2022-02-25 09:39
chadmedyeah someone mentioned that there are just so many places where they rely on it being hard coded that it would be nigh on impossible to change it at this stage, which sucks#asahi · 2022-02-25 09:36
chadmedits actually kind of hypocritical and pathetic imo, they maintain these gargantuan software packages that exist almost entirely to facilitate the execution of code in the most inefficient, slow and energy intensive manner possible #asahi · 2022-02-25 09:35
chadmedwhere they feel like they literally cannot spare the handful of cycles to check the system page size#asahi · 2022-02-25 09:34
chadmedthat being said i dont like the state of the web and the mammoth effort required by web browser developers to support it all. it leads to problems like this #asahi · 2022-02-25 09:34
chadmedhonestly in some ways, its good for enterprise. HR managers and other assorted corporate mooks should not and cannot be trusted to secure their own data, so use webapps to hide the data from them#asahi · 2022-02-25 09:33
chadmedbecause web browsers in 2022 are just funky VMs #asahi · 2022-02-25 09:30
chadmedi mean the point about compile time optimisations is still true, idk what optimisations they think are so important that they cant just check for page size at runtime. they used to do it in fact#asahi · 2022-02-25 09:28
chadmedvalve for example ship their own version of tcalloc with all their source engine games#asahi · 2022-02-25 09:28
chadmedgames being the major group#asahi · 2022-02-25 09:28
chadmedmalloc is slow as hell, virtually every program that is memory latency dependent uses a custom allocator#asahi · 2022-02-25 09:19
chadmedmy main gripe is just fex not working on 16k pages lmao, i wanted to do some performance testing and tracing over the holidays but my holidays ended today so...#asahi · 2022-02-25 09:18
chadmedheh yeah#asahi · 2022-02-25 09:16
chadmedi cant imagine theyd be willing to if it meant a performance regression on 4k systems#asahi · 2022-02-25 09:16
chadmedyeah but we have to convince them to do so#asahi · 2022-02-25 09:11
chadmedor rather, nothing the developers are willing to merge ;) #asahi · 2022-02-25 09:11
chadmedyeah i came across that in my investigation too, shame theres not a lot that can be done about it #asahi · 2022-02-25 09:07
chadmedthe only major show stopper ive run into on 16k pages is some things that distribute a precompiled jemalloc#asahi · 2022-02-24 08:31
chadmedoh right, yeah it was add .gitignore but github doesnt like dotfiles in commits i guess#asahi · 2022-02-24 08:28
chadmed?#asahi · 2022-02-24 08:25
chadmedthe commit history is rough, please dont look at it ;) #asahi · 2022-02-24 08:23
chadmedhello all, ive been working on an Asahi Linux Plymouth theme for the last two days. it very closely emulates Apple's own boot splash screen. i've been testing it with Artix Linux, however I feel like it would be good to have some wider spread testing. https://github.com/chadmed/asahi-plymouth#asahi · 2022-02-22 04:35
chadmedoh good stuff then #asahi · 2022-02-22 04:33
chadmedhow do we reliably expand the dd-ed image into whatever free space the user actually specified? ik some rpi images do it but it looks like these assume any space below the root partition is free to bulldoze? #asahi · 2022-02-12 10:46
chadmed+1 #asahi · 2022-02-12 10:42
chadmedyep#asahi · 2022-02-12 10:33
chadmedoh right so to use a uuid as your root you change root=/dev/sdX to root=UUID=[uuid of the partition] #asahi · 2022-02-12 10:29
chadmedif youre building a payload for an untethered boot you will need to build the kernel with the arguments built in. the "official" way will be to only chainload m1n1 and u-boot and then use GRUB to pick a kernel and pass its boot arguments to it#asahi · 2022-02-12 10:26
chadmedyou also need to specify which minor on /dev/sda has the root partition. i would always use the partitions UUID instead of its name in /dev/ since the uuid never changes#asahi · 2022-02-12 10:25
chadmed if you are using an initrd you need to have those modules packed into it#asahi · 2022-02-12 10:25
chadmedcheck your kernel config. if youre not using an initrd you need to have usb, scsi support, usb mass storage and usb attached scsi set to y not m