#
✕ nick:Sobek[m]
#asahi · 2022-03-24 10:47
Sobek[m]Hmm, I hope the discussion with the distros will end up in a state where distro are will to support 16k kernel (or mostly 16k kernel, if eventually a solution is found so that odd balls like FEX can run on the kernel while using 16k pages for everyone else). But sorting out these final bits is way less important than the driver work, isn't it ?
#asahi · 2022-03-24 10:43
Sobek[m]Stable Debian, on M1, call me in 2030 perhaps ?
#asahi · 2022-03-24 10:43
Sobek[m]Some people seem to forget this is still alpha and all the moving parts that have been dealt with yet, and all that remais to be done on the other front.
#asahi · 2022-03-24 10:42
Sobek[m]What is he referring to, by "don't ship a working kernel" ?
#asahi · 2022-03-24 10:19
Sobek[m]Is there like #asahi-offtopic ?
#asahi · 2022-03-23 13:21
Sobek[m]Well, Japan is getting away pretty cheaply :o
#asahi · 2022-03-23 13:20
Sobek[m]1 144,00 € in France too, but all tax included (including our 20% VAT), if I remove the taxes I get down to 950€ which is still 100€ than Japan. (Assuming the Japan one is without tax)
#asahi · 2022-03-23 13:17
Sobek[m]Hmm, are taxes included in both cases ?
#asahi · 2022-03-23 11:02
Sobek[m]Glanzmann: The asahi installer expects some kind of ISO to dump on the newly created root, or just a bootable live media ? I'll be interested to read about the steps you took to create a gentoo system that boots on my for sure :)
#asahi · 2022-03-23 10:59
Sobek[m]What do you mean by package gentoo ?
#asahi · 2022-03-23 10:56
Sobek[m]That's what I thought :D
#asahi · 2022-03-23 10:45
Sobek[m]packages + the one two special ones, as long as you have a stage3 that runs)
#asahi · 2022-03-23 10:45
Sobek[m]However, gentoo is supposed to be installed by hand, so aside from documenting a reliable way of installing gentoo, a bootable media that can boot the M1 mac with network is the only thing that needs to be provided, along with a working stage3 (I hope the gentoo stage3 work on M1s) ^^' (However documenting a known to work install process will be great. I'd say an asahi linux overlay probably is enough to add the various patches on top of
#asahi · 2022-03-23 10:39
Sobek[m]https://wiki.gentoo.org/wiki/Initramfs
#asahi · 2022-03-23 10:36
Sobek[m]I'd say either by hand, or using tool such as genkernel (or dracut if i don't misremember ?). How do you do it on arch ?
#asahi · 2022-03-22 22:20
Sobek[m]aem64[m]: (Hmm MoltenVk is a thing on macOS, I think the Vulcan tutorial more or less targets macOS using it as part if the SDK, and there are perhaps 1-2 obscure corners that don’t work but otherwise it works well)
#asahi · 2022-03-22 22:18
Sobek[m]By the way, speaking of the GPU side of things, I’ve seen stuff about OpenGL, but is Vulkan also on the roadmap ? High performance nowadays tend to also involve vulkan (and then stuff like DxVk are built on top of it). (@alyssa ?)
#asahi · 2022-03-22 13:39
Sobek[m]mps: I believe you can go around and manually fix stuff, but fixing the original keyboard layout shipped by distros and making sure the default one selected is the sane one is still nice :)
#asahi · 2022-03-22 13:36
Sobek[m]mps: "What is the best default keyboard layout for the various languages, and what bug are still hidden". I'll probably find a few stupid thing with the Fr (mac) layout once my backup situation is sorted out and I can tweak stuffs. I remember quite a few headaches with odd characters not being typed by the right keys on my previous mac linux dual boot a couple of years ago.
#asahi · 2022-03-22 13:32
Sobek[m]Already the fact that Asahi has tackled the keyboard mess will be a huge contribution for many linux users (if hopefully this is a success).
#asahi · 2022-03-22 13:28
Sobek[m]mps: Uh, what is that driver for ?
#asahi · 2022-03-22 13:27
Sobek[m]802.1X are a regular pain to configure on various OSes might be troubles only remotely related to Asahi (like just the usual arch + wifi + 802.1X troubles)
#asahi · 2022-03-22 13:20
Sobek[m]sven: But is there any way we can get this sort of thing to work ? I could see a very inefficient method which is deny real access to the whole and do the subpage check and if legal modification in teh fault handler (like ZFOD and CoW design tend to do), but it sounds very very inefficient perf wise ?
#asahi · 2022-03-22 13:18
Sobek[m]j`ey: I love it. Actually it might be a great solution to the problem. The kernel pretendes it deals only in 4k pages, while it in fact tries very hard to use as many 4k pages as possible :D
#asahi · 2022-03-22 13:18
Sobek[m]marcan: Is this something that could be brought up to some linux maintainer for advice / opinion on what they think should be done long term. Like Asahi linux alone can't decide on these sort of thing, but perhaps we can get the people in charge thinking a bit and provide guidance on how to make the part fall in place long term ?
#asahi · 2022-03-22 13:14
Sobek[m]What's the memory folio thing ? (I can take a pointer to an explanation if that would bring the channel to far off topic for too long)
#asahi · 2022-03-22 13:13
Sobek[m]Also, at that point will it be possible for that kernel to offer 16k page to processes by default, unless requested otherwise ? (Or is that impossible to do in linux because it would require some special userland interface that cannot be added). That way only processes that require 4k page will opt-in ?
#asahi · 2022-03-22 13:10
Sobek[m]Plus, don't you need 16k pages for some PCIe / hardware stuff because of the IOMMU ?
#asahi · 2022-03-22 13:09
Sobek[m]realmoot: Pages, not buffers.
#asahi · 2022-03-22 13:08
Sobek[m]Oh well, I was wrong about U-boot (hopefully it is not also a Yellow kind of submarine, unless I'm very wrong :D)
#asahi · 2022-03-22 13:07
Sobek[m]And each piece has a simpler job to do. Also U-boot and grub are pieces that are more or less off the open source shelve that required few patches. The big work is m1n1 to interface with how macs want to boot stuff (they expect something that looks like macOS and quacks like macOS), and linux kernel to have all the drivers.
#asahi · 2022-03-22 13:06
Sobek[m]realmoot https://github.com/AsahiLinux/docs/wiki/Introduction-to-Apple-Silicon and https://github.com/AsahiLinux/docs/wiki/Open-OS-Ecosystem-on-Apple-Silicon-Macs tell the full story
#asahi · 2022-03-22 13:05
Sobek[m]It can also contains a fun hypervisor that is used to help reverse engineering what macos does.
#asahi · 2022-03-22 13:05
Sobek[m]realmoot: It is the thing marcan wrote that pretends to be a macOS kernel and then goes around and loads the rest of the open OS boot chain
#asahi · 2022-03-22 10:44
Sobek[m]Or you can dump the files on a dumb USB key with a fs that both OS like (FAT32 for small files, I don't remember off the top of my head what to use for big files).
#asahi · 2022-03-22 10:43
Sobek[m]EvadingEvasions[m]: So, I think there are some tools that allow mount ext2 on macos (using fuse or something like that), if they have made the jump to ARM land. And the other way round, there's a fuse APFS fs that can be used (but in readonly, be careful).
#asahi · 2022-03-22 10:42
Sobek[m](But the page size stuff is likely going to cause headache to a bunch of people before it gets better I fear)
#asahi · 2022-03-22 10:41
Sobek[m]kaprests: Mostly because a lot of the mac software is closed source + macos never ran on ARM before, whereas linux is a. open source so recompile is quite a bit easier, and linux already runs on ARM so software have already been ported to ARM.
#asahi · 2022-03-22 10:27
Sobek[m]Matrix IRC might have mangled who I was replying to, I was asking enchiladas, wrt his rust / cargo question.
#asahi · 2022-03-22 10:23
Sobek[m]<mps> "enchiladasconpinguino: by the..." <- What exactly are you interested in ?
#asahi · 2022-03-22 09:02
Sobek[m]Ah, well, I got confused ><
#asahi · 2022-03-22 08:54
Sobek[m](At least it appears M1 is not the only aarch64 chip out there that has some page size subtleties, Loongson is another one from what appears earlier in the chat)
#asahi · 2022-03-22 08:53
Sobek[m]This is where we should probably have a chat with distro maintainers to see what's the favoured solution for them ?
#asahi · 2022-03-22 08:52
Sobek[m](Larger page size or page size detection)
#asahi · 2022-03-22 08:47
Sobek[m]jemalloc is generally linked statically / vendored or is it a package that asahi linux could override on M1s with a version suitable for it ?
#asahi · 2022-03-22 08:37
Sobek[m]It sounds like the page size should be part of the target triple or something though, given the impact it has on portability. Is it deemed a part of the ABI or not btw ?
#asahi · 2022-03-22 08:36
Sobek[m]Hmm, embedded people are the one who can likely add in the switch to specify which arm page size they want. (They know their hardware).
#asahi · 2022-03-22 08:32
Sobek[m]The main cost is obviously making the library take up three time the space it used to (which will feed the troll about rust binaries sizes obviously ^^')
#asahi · 2022-03-22 08:31
Sobek[m]If you set-it up this way yes. But they pointed out that some compile time structure relied on page size, so you have to do it that way and you can't just have a rather synamic implementation where the same code handles the three case.
#asahi · 2022-03-22 08:29
Sobek[m]The issue has been mentioned like twice this year btw

plus ›