marcan changed the topic of #asahi to: Asahi Linux: porting Linux to Apple Silicon macs | General project discussion | GitHub: https://alx.sh/g | Wiki: https://alx.sh/w | Topics: #asahi-dev #asahi-re #asahi-gpu #asahi-stream #asahi-offtopic | Keep things on topic | Logs: https://alx.sh/l/asahi
<Serentty[m]>
I’ve partitioned space for it and installed M1N1. How should I go about installing Linux. Is there a write-up somewhere? I’m okay with something manual like extracting a tarball and setting it up from the command line. I’m just not sure what I should download in terms of a kernel, if I need any patches, and stuff like that.
<Serentty[m]>
This guide seems to be quite out of date.
<alyssa>
jmr2: Anwyay, so there's no problem parsing
<jmr2>
Could the macOS version used to install m1n1 have an impact?
<alyssa>
No
* alyssa
is typing
<jmr2>
Good.
<alyssa>
Er wait yes
<alyssa>
DCP firmware version
<alyssa>
slightly diferent from what you said but
<alyssa>
You're on 11.4 right?
<alyssa>
Everything should be broken otherwise.
<jmr2>
11.6 on MacOS, 11.5.2 I think on the Linux stub.
<alyssa>
uh
<alyssa>
yes that could have an impact..
<chadmed>
Serentty: if youve already got m1n1 installed then the developer quickstart guide should still be accurate to follow on from. nothing about that process has changed materially, marcan just made a script to automate some of it
PhilippvK has joined #asahi
<marcan>
alyssa: we need to figure out how to encode this in the DT
<alyssa>
marcan: thanks for volunteering to write the m1n1 patch
* alyssa
doesn't care if the DT is purple and upside-down
<marcan>
I guess I should scrape all existing IPSWs and first determine if iBoot version always bumps together with DCP version
<alyssa>
marcan: the other question is, how do I know the physical size of the display?
<marcan>
if it doesn't then we need to use something else
<marcan>
didn't DCP tell you?
<alyssa>
for HDMI and presumably DisplayPort, this is in the productattributes dict
<jmr2>
I could try to install the 11.4 stub tomorrow, and redo the dump.
<alyssa>
At least on whatever version jmr2 is on, DCP doesn't tell you for the internal panel on the macbook
<alyssa>
So either it gets hardcoded in the linux DT, or maybe it's passed through the ADT
<alyssa>
Minor thing but apparently is breaking the DCP driver on Macbooks right now
phiologe has quit [Ping timeout: 480 seconds]
<marcan>
I don't see it in any obvious place in the DT
<marcan>
*ADT
<alyssa>
Alright.
<alyssa>
I guess that needs to go in the macbook air / pro DTs resprectively then.
<marcan>
ah wait
<jmr2>
Confirmed - my Linux stub is 11.5.2.
<marcan>
yeah I don't see it
<alyssa>
width-mm/height-mm panel props are a thing on linux
<alyssa>
^DT
<marcan>
yeah, we should probably just use that
<alyssa>
still annoying as hell
<chadmed>
would that mean a separate DT for each size macbook?
<marcan>
that's already going to be a thing
<marcan>
they are different models
<marcan>
we will have a DT for each model
<marcan>
it's not just the screen size
<chadmed>
ah right
rohin has quit [Quit: Konversation terminated!]
<alyssa>
marcan: I suppose on the Mini side, I'm currently hardcoding some stuff for HDMI
<alyssa>
so that all has to get described in the DT too. meh.
rohin2 has joined #asahi
<alyssa>
Ooh, this is what drm_bridge is for.
<alyssa>
I see how to do this now. and with 100% generic code if I don't screw it up :]
<alyssa>
ok, yes
<Serentty[m]>
<chadmed> "Serentty: if youve already got m..." <- Can I boot it without sending commands over a serial connection? That’s what the instructions seem to say to do.
<jmr2>
alyssa: just installed a 11.4 stub, and the three lengths are the same. I'll need a bit more time to compare the contents.
<alyssa>
switch to drm_bridge, use display connector bridge, describe HDMI in the mini DT, describe panels in the macbook DTs, everyone is happy.
<alyssa>
jmr2: did you apply the rm diff I sent?
<jmr2>
Not yet. That's my next step.
<alyssa>
handling the macbook case properly takes a lot more work (see ^^), but for now you can just skip it tbh
<chadmed>
Serentty: you can concatenate the device tree, kernel and m1n1 into a single macho binary and have m1n1 chainload the kernel. its in the quickstart, a bit further down i believe
<Serentty[m]>
Alright. I’m going to build the kernel now. It’s complaining about the lack of a GNU toolchain targeting AArch64 Linux. Can I use Clang and LLD instead?
<chadmed>
yeah just add LLVM=1 LLVM_IAS=1 after make
<chadmed>
m1n1 has some problems with clang i think though so it might be worth investing the time into building an AArch64 gnu toolchain
nobodynada has joined #asahi
<jmr2>
alyssa: ah, beat you to it! If you check in my "for when you're bored" patch & dmesg from earlier today, I was already forcing non-zero values into width_cm and height_cm.
<jmr2>
That leads to a frozen display, an unhandled D211 callback, vblank wait timouts and lots of interesting stuff.
<jmr2>
(same result currently - I'm still forcing those, even if I no longer do the corruption that you pointed out)
<jmr2>
But still - let me try commenting that out instead.
<Serentty[m]>
Hm... It says “restart config” and it wants me to go through the kernel configuration. I copied the file to .config like it said.
<chadmed>
you should run menuconfig or nconfig and set the kernel up properly. it will stop it from doing that. theres probably some new options in the kernel that arent in your .config or vise versa
<jmr2>
alyssa: same with your patch - frozen screen, D211 and vblank timeout.
<Serentty[m]>
chadmed: Now I’m going through the kernel configuration. Is page size something worth thinking about?
<chadmed>
the M1 uses 16K pages so as long as that's set you should be fine. not sure on the status of sven's 4K patches rn.
<Serentty[m]>
scripts/sorttable.c:27:10: fatal error: 'elf.h' file not found
<Serentty[m]>
That’s strange.
<jmr2>
alyssa: and unless I screwed up, the ColorElements/TimingElements/DisplayAttributes are the same regardless if I boot from an 11.4 or an 11.5.2 stub.
marvin24 has joined #asahi
<Serentty[m]>
Oh man, I’m running into all of the errors described in this article which I found by searching.
<marcan>
... how did that even happen. something something right click menu
<marcan>
Serentty[m]: I was going to say, yeah, *usually* you'd be building linux kernels on linux
<marcan>
doesn't have to be ARM linux though
<Serentty[m]>
Oh, it just occurred to me. Does anyone have a Mach-O image with M1N1, the kernel, the device tree, and the initramfs? I just tried it with the kernel I built, but it didn’t detect it.
<marcan>
did you use the gzipped kernel?
<Serentty[m]>
Yes.
<marcan>
this is for running under the hypervisor?
<chadmed>
damn konversation doesnt play nice with fcitx...
<Glanzmann>
Serentty[m]: You only need a usb-c to usb-c cable and a second linux box.
<Serentty[m]>
I don’t think it is. Just booting in general.
<Serentty[m]>
Glanzmann: I don’t actually have any other computers in the house with a USB-C port.
<Glanzmann>
Serentty[m]: USB-C also does it.
<Glanzmann>
Serentty[m]: Do you have a USB-A to USB-C Cable than?
<marcan>
Serentty[m]: you usually would not embed Linux kernels into m1n1 for booting stand-alone
<Serentty[m]>
How can I boot standalone, then?
<marcan>
for an end-user system, you'd embed U-Boot into m1n1 and use that as a second stage, but we don't really have guides for that nor is it in the installer because right now we only expect kernel devs to be booting things and developing using another system as a host
<Glanzmann>
marcan: How can someone pass linux command line parameters with a concatted m1n1? So how can I run this with concatted m1n1? M1N1DEVICE=/dev/ttyACM0 proxyclient/tools/linux.py --bootargs="net.ifnames=0 rw root=/dev/nvme0n1p4 rootwait rootfstype=ext4" Image.gz t8103-j274.dtb
<marcan>
Glanzmann: you can't, you need to put them in the device tree right now. I've been meaning to add some mechanism for that.
<Glanzmann>
marcan: But I think the linux kernel can embedd a linux default command line, can't it?
<marcan>
yes, either in the kernel or the DT
<marcan>
I'm just saying there's no way to tell m1n1 to do it separately
<marcan>
but you can use either of those methods
<marcan>
I use the DT
<marcan>
just put the commandline in /chosen/boot-args
<Serentty[m]>
So right now, the only way to boot is with serial every time? Or making M1N1 binary with U-Boot?
<Glanzmann>
Serentty[m]: No, as marcan said you put the kernel command line in the dt under /chosen/boot-args
<marcan>
I mean you *can* embed a kernel into m1n1 and boot it, but it'd be silly because you'd have to go to 1TR and run kmutil every time you want to change the kernel
<Glanzmann>
Than you concat m1n1, kernel and dt and copy to the boot volume and bless it with kmutil.
<marcan>
that's why the production setup will use U-Boot
<Glanzmann>
In your case you have to do that two times, one for booting from usb, and than for the final one booting from the nvme.
<marcan>
in my opinion there really isn't a good reason for people to be putting kernels straight into the boot object, especially at this point in the game... but I won't stop you if you really want to do it that way :)
<Glanzmann>
Serentty[m]: But it is much easier if you just go to amazon and buy a cable, if you don't have one. You just need a second linux box with a usb port and the right cable to hook it up to the m1.
<marcan>
Glanzmann: macos works too
<Glanzmann>
True.
<marcan>
and yes, any USB-C to USB-A or USB-C cable (depending on your other side) will work, nothing fancy
<marcan>
you don't need a serial cable, just a USB cable will do
<marcan>
serial is a special thing
<Glanzmann>
marcan: I mean the m1n1 as desktop is usable, I'm typing on it right now. And if you don't play a lot with the kernel, maybe you just want to be able to turn it on and use it. But yes, probably you will compile kernels every few days.
<marcan>
sure, but given how things are in flux...
<Glanzmann>
True.
<marcan>
like, I won't stop people from doing it, but it's going to make debugging a nightmare and just isn't terribly useful, so you're kind of on your own if you really want to do that
<marcan>
what this is telling me is I need to get the U-Boot flow sorted out because enough people want to use this already :p
<chadmed>
yeah i dont see a reason for directly embedding the kernel unless you really, _really_ hate u-boot or something
<Glanzmann>
marcan: How will that work? m1n1 loads uboot, uboot has a nvme driver and reads the kernel from the nvme?
<Glanzmann>
chadmed: Or you just want a workstation that boots when you turn it on.
<Serentty[m]>
And then I put the payload file on my APFS volume and call it m1n1.macho.
<marcan>
did you run kmutil?
<Serentty[m]>
Okay, thanks.
<chadmed>
Glanzmann: you can build u-boot so that it runs non-interactively, when was the last time you had to point your android phone to its device tree and kernel?
<Serentty[m]>
marcan: Nope. But it did find M1N1 and boot it.
<marcan>
nope, it didn't
<marcan>
that was your old m1n1
<marcan>
m1n1.macho doesn't do anything on its own, you need to install the boot object from 1TR (booting with the power button held down). This is why this is a pain in the ass and not recommended.
<marcan>
when you run step2.sh it copies m1n1.macho into the proper place and updates the boot policy
<marcan>
this is part of secureboot and requires authenticating against the SEP and only works from 1TR
<marcan>
so it is absolutely required to go through this process every time
<marcan>
that is why I think it makes no sense to embed kernels this way
<Serentty[m]>
marcan: Can I run the step2.sh file on my Linux volume from recovery to update this?
<marcan>
yes, as long as you put m1n1.macho next to it (you might need to remount read-write as macos likes to mount system volumes read-only by default)
<Glanzmann>
marcan: What is missing for uboot? Is kettenis not already doing it for OpenBSD?
<Glanzmann>
marcan: Where do I get the dt for the macbook air?
<marcan>
there is no canonical MBA devicetree yet
<Glanzmann>
I see.
<Serentty[m]>
Well, now it’s saying that the custom kernel failed to boot, so that’s progress.
<marcan>
u-boot works, it's just not upstream yet and I haven't integrated it with the installer / build stuff
<marcan>
Serentty[m]: how big is your image? there's a payload size limit
<marcan>
64MB by default right now
<Serentty[m]>
It’s only about 25 MiB.
<Glanzmann>
marcan: I see. Where do I get the uboot from, are there any instructions I can try?
<marcan>
don't ask me, getting that part sorted out is on *my* TODO list :)
<marcan>
there's a reason it's not in the installer yet
<Glanzmann>
I see.
<marcan>
again, this is why I'm only targeting kernel devs willing to boot tethered right now...
<Serentty[m]>
By the way, there’s no filesystem or anything for it to boot off of. That shouldn’t be the issue, right? Like, if that were the issue, I would see Linux try to boot but complain, right? It wouldn’t just reboot without showing anything?
<Glanzmann>
Serentty[m]: Correct.
<Serentty[m]>
I think I’ll leave it here for today.
<Serentty[m]>
I’ve learned a bit more about the boot process.
<Serentty[m]>
I could try the USB method next time. The funny thing is that the only other Linux machine that I have in this house (other than mobile stuff) is a PowerMac G3. Using that would be kind of funny.
<Glanzmann>
povik: I would like to try the audio driver, where can I get it from and do you have any instructions for it?
<Serentty[m]>
Using a PowerMac to boot an ARM Mac, completely skipping over x86. There’s some poetry to it.
<marcan>
you can use any Linux or macOS machine as the host side
<marcan>
probably windows too if you try hard enough
rohin2 has quit [Ping timeout: 480 seconds]
<Serentty[m]>
I’ll probably use the PowerMac G3 then. It’s not fast, but it should be fast enought to run this I would imagine.
<Serentty[m]>
It’s kind of neat how this machine which came with Mac OS 8 can still be useful thanks to Linux. :D
<marcan>
I imagine compiling kernels on that won't be a great experience, but it should work
<Serentty[m]>
Oh, I compiled the kernel earlier in a Linux VM on my M1 Mac.
<Serentty[m]>
I’ll just transfer it over.
<Serentty[m]>
But that’s for another day at this point.
<Glanzmann>
jmr2: If you do any progress on the macbook air, let me know, I would like to reproduce that.
<Glanzmann>
maz: Can you send me the URL to your m1 kernel tree, please? I would like to try it out.
<sven>
povik: those are "documented" somewhere in the logs of this channel :(
<sven>
but i'm about to finally get back to nvme
<sven>
just want to submit another version of that 4k iommu series first
<sven>
povik: but can just use marcan's reset controller to make ANS work again after it's been initialized once
<sven>
it's not the "proper" way without sending the shutdown commands first but it'll work fine
<Glanzmann>
sven: What are the next steps for NVMe?
<sven>
clean it up and then submit it
<sven>
then figure out all the things i did wrong during review and clean those up as well
<Glanzmann>
I see, nice. :-)
<Glanzmann>
sven: I only noticed that random writes under Linux feel slow. I compiled fio yesterday under macos to compare performance, but it does not run.
<sven>
there's a reason i want to run linux on these machines ;)
gabuscus_ has joined #asahi
gabuscus has quit [Ping timeout: 480 seconds]
<Glanzmann>
sven: Allocating 80GB of ram when you have only 16 might be a bad idea. If I only allocate 1 GB of RAM, fio works on macos. There is a hugh performance difference: 4k random writes: 50 vs 34.4k. See here: https://ab34.de/u/m1-nvme-performance-linux-vs-macos-2021-10-18.txt
<Glanzmann>
But you're right, lets get that driver upstream first.
<Glanzmann>
Because a slow nvme driver is much better than none. :-)
<povik>
^ this is .config i am building audio-w-dcp with
<sven>
Glanzmann: huh, that's a surprisingly large difference
<sven>
especially because all that performance stuff should be handled by the regular linux nvme code
<maz>
Glanzmann: try *what* out? my tree is at git://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git , but there is no single M1 branch. rather a collection of series.
<Glanzmann>
j_ey: Than I'll do that do, do you have x11 running on the air?
<j_ey>
Glanzmann: nope
<Glanzmann>
maz: I heard you have a frankenstein kernel with some corellium parts and asahi, I would like to try that.
<Glanzmann>
j_ey: But the framebuffer is working?
<j_ey>
Glanzmann: yes
<Glanzmann>
Perfect.
<Glanzmann>
sven: Yes, i noticed when I installed my first deb using apt, but could not pinpoint it until I got fio working on macos.
<Glanzmann>
sven: I'm doing some linear tests as well.
rross101 has joined #asahi
<Glanzmann>
sven: Maybe it has something todo with large pages? Linear is better and has higher iops even with 1MB block size. So for linear writes with 1MB block size it is 2.7GB/s (macos) vs 1.7 GB/s (Linux). https://ab34.de/u/m1-nvme-performance-linux-vs-macos-2021-10-18.txt
<sven>
no
<sven>
the nvme page size is 4k
<sven>
it also doesn't go through DART
<maz>
Glanzmann: no, I have no such thing. I don't touch the corellium stuff at all.
<Glanzmann>
maz: I see.
<Glanzmann>
maz: Than I confused you with someone else.
<maz>
Glanzmann: no worries
<povik>
i used the same words to describe pipcet's kernel, maybe you mixed that up
<Glanzmann>
Basically I did linear read/write with 1MB block size and random read/write with 4k block size. all queue depth 1.
<Glanzmann>
povik: [RE audio branch and config] I'm currently sitting in the car waiting for my wife to finish some doctors appointment but once I'm back on my desk, I'll try that, thank you.
<povik>
great, let me know how it goes. expect only the speaker on mac mini working
<povik>
no other output devices will be there
<Glanzmann>
j_ey: About the m1 mabook air: Do you have the built in keyboard?
<povik>
although i have a working prototype of audio_jack.py :-p
<j_ey>
Glanzmann: yes, using corelliums spi+keyboard driver
<maz>
sound? and I was trying to prevent the mini from making that sound at on reset... ;-)
<maz>
s/at//
<j_ey>
maz: boot into macOS, you can turn it off!
<sven>
i hear a dremel or similar tools should work fine for that
<povik>
maz: yeah. i had an opposite reaction, i wanted more of it!
<maz>
j_ey: booting into macos involves finding a keyboard and a screen... I think I like sven's approach better.
<j_ey>
maz: his approach may void your warranty
<maz>
j_ey: I'll survive that.
<sven>
you could also write a spi driver and write to that nor flash to disable the bootup sound from linux - but where's the fun in that?
<j_ey>
sven: well we'd have a SPI driver for one thing :P
everslick_ has quit [Remote host closed the connection]
<maz>
I'm not indifferent to non-invasive surgery, but given that I already have the tooling... :D
everslick_ has joined #asahi
rross101 has quit [Quit: Page closed]
<Glanzmann>
j_ey: Do you have your tree somewhere?
<Glanzmann>
j_ey: Do you have wifi running on the mabook air? If so do you have instructions how to get it running?
<j_ey>
Glanzmann: I dont have wifi, but I think jmr2 did. alyssas branch has the code but you need the firmware and some way to set it up
<kov>
j_ey, btw, real life got in the way, but I managed to do a few builds and poking at config, I couldn't get the keyboard to work even after figuring out all of the dependencies to get KEYBOARD_APPLESPI enabled properly, would you mind sharing your .config? maybe there is some hidden dependency that I failed to enable
<j_ey>
kov: can you share yours? I can take a look
<kov>
heh, but now kov@uva ~/P/A/linux (m1-keyboard)> grep PINCTRL_APPLE_GIO .config
<kov>
kov@uva ~/P/A/linux (m1-keyboard) [1]>
<kov>
erm doh
<kov>
yes it was the missing P
<j_ey>
:P
<kov>
ok let's try building that and see if it works
<j_ey>
kov: if not, paste the dmesg output
<kov>
j_ey, k
bgb has joined #asahi
<bgb>
j_ey: just tried the IPI fix of m1n1, seems 100% boot to macos(hv) :)
<kov>
j_ey, it worked! I wonder if it was COMMON_CLK_APPLE_CLKGATE all along but menuconfig somehow fixed it for me
<kov>
I think I was only using olddefconfig before
<kov>
ok, now to figure out the cherry picking =)
<j_ey>
kov: definitely easier to cherry pick keyboard onto DCP
<j_ey>
bgb: great! its in the main branch now
rross101 has joined #asahi
<alyssa>
jmr2: Next thing I guess is enabling the debug messages in DCP
<alyssa>
good to know the TimingElements/DisplayAttributes dicts are the same
rross101 has quit []
<alyssa>
povik: my kernel has some corellium code in it but not much
<alyssa>
reboot, spmi, rtc, and wifi are corellium code
<alyssa>
of those, I've made changes to... I think all of them
<alyssa>
reboot will be replaced entirely by a wdt driver eventually
<alyssa>
wifi needs work
<alyssa>
spmi and rtc, idk, the code is there, I fixed a lockdep issue (🤔) but yeah
<arnd>
alyssa: do you have a link to your version of those drivers?
bps has quit [Ping timeout: 480 seconds]
<arnd>
I just looked at the corellium rtc driver, the code looks "obvious", but I think we'd need to replace the two DT properties with normal "reg" properties, and of course it needs the spmi driver first, and that seems likely to need more changes, or at least a more detailed review
jmr2 has joined #asahi
bgb has quit [Ping timeout: 480 seconds]
<jmr2>
Glanzmann: The only real progress I made on the air is to get the wifi working by reproducing alyssa's steps. Just need to activate with smc.py prior to booting, and putting the correct firmware files in an initramfs. See the T2linux wiki.
<jmr2>
alyssa: OK, I'll try later today. Would you mind sharing your best trampoline_* guess for D211 ? Is the tool you're using to lookup the function names accessible?
<povik>
alyssa: i know, i took reboot from you actually
<alyssa>
jmr2: m1n1/proxyclient/hv/trace_dcp.py has the function names from 11.4
<alyssa>
^ this is what I'm running as a daily driver
<arnd>
thanks
<alyssa>
and agreed... rtc looked just about there, spmi looked kinda scary
<alyssa>
or maybe I'm just saying that because their spmi driver broke the boot because of a serious locking failure
<alyssa>
(did they just disable lockdep?)
<arnd>
I have no business in drm-misc, or at least I keep saying no whenever I'm offered commit access
<sven>
^-- smart! :p
<alyssa>
arnd: I kept trying to do that
<alyssa>
but I kept asking for someone to take a look at a patch
<alyssa>
and the maintainer kept telling me to push it to drm-misc-fixes
<alyssa>
and i'm like I don't have push
<alyssa>
and he's like well fix taht.
<alyssa>
do I have class today? idk!
<alyssa>
it's a review session. do I go late or skip it or stop yapping on irc and sprint
<brentr123[m]>
Vote for skip it?
jmr2 has quit [Quit: Page closed]
Misthios has quit [Ping timeout: 480 seconds]
bgb has quit [Ping timeout: 480 seconds]
<arnd>
alyssa: I had not even noticed that MFD regmap bit, only the main driver. Any idea what the MFD side is good for? For the actual spmi driver, it's mostly the irqchip bits that I suspect need changes after a thorough review. The interrupt-hopping data transfer looks inefficient, but there is probably not much to be done about that in the driver.
bgb has joined #asahi
bgb has quit [Ping timeout: 480 seconds]
bgb has joined #asahi
bgb has quit [Ping timeout: 480 seconds]
<chadmed>
[22:44] <alyssa> do I have class today? idk! <-- me every day for ~26 weeks of the year
<chadmed>
i have phsyiology lab time booked at 10am tomorrow, it is currently 11:30pm and the mac event is at 3am. my dilemma right now is do i make myself another green tea and stay up to watch it, or go to the lab well rested and ready to sit around collecting data?
bgb has joined #asahi
Leidenfrost has joined #asahi
<Leidenfrost>
Hello guys. Question: Will HD video (via vdpau or vaapi) will come after 3d acceleration or is it a project expected to be started in parallel?
<Leidenfrost>
Oh sorry wrong channel
bgb has quit [Ping timeout: 480 seconds]
bgb has joined #asahi
bgb has quit [Ping timeout: 480 seconds]
bgb has joined #asahi
Misthios has joined #asahi
bgb has quit [Ping timeout: 480 seconds]
Misthios has quit []
Misthios has joined #asahi
bgb has joined #asahi
bps has joined #asahi
bgb has quit [Ping timeout: 480 seconds]
alyssa_ has joined #asahi
bps has quit [Ping timeout: 480 seconds]
<alyssa_>
alyssa: where are you
<sven>
you actually cloned yourself? impressive!
rimani has joined #asahi
bgb has joined #asahi
bgb has quit [Ping timeout: 480 seconds]
kenjigashu has joined #asahi
rimani has left #asahi [Leaving]
bgb has joined #asahi
alyssa has quit [Remote host closed the connection]
alyssa_ is now known as alyssa
<alyssa>
arnd: haven't looked too much at the SPMI stuff, not sure what the MFD shim was for
<alyssa>
DCP has been consuming all my (already limited) dev hours
* everslick_
is gone, searching for himself. if you see him, before he gets back, keep him here.
<marcan>
alyssa, sven, maz, povik, kettenis, j_ey, arnd: how about we try to shift more dev chat to -dev, since this channel is getting a bit noisy lately? This is why I made them separate initially :)
<povik>
why not
<sven>
works for me
<alyssa>
wfm
<alyssa>
in that case i'm /parting this one because I don't need the noise 😅
<j_ey>
what about helping people, that can stay here I suppose
<povik>
alyssa: bye bye
bgb has quit [Ping timeout: 480 seconds]
<povik>
j_ey: no! all your attention will be -dev now
<alyssa>
will find somewhere else to dump my inane scholastic ramblings, maybe twitter :-p
<ar>
marcan: do you want to supersize that m1macmax?
<fridtjof[m]>
m1x2, m1x4? :P
<j_ey>
lol chop
<marcan>
"chopped in half" I guess? :p
<j_ey>
marcan: mentioned secure boot
<marcan>
yeah heh
<j_ey>
I guess thats it for the fun stuff
<dottedmag>
Why do the app developers sound like a Apple marketing department employeees?
<dottedmag>
stupid question, I know
<brentr123[m]>
j_ey: I’m sure we can get around that somehow
<marcan>
so the thing is, the rumor is there's a third one
phire_ has joined #asahi
<marcan>
which is 2x M1 Max dies together
<marcan>
... which I guess is not getting announced today
<j_ey>
m1 maxmax
<marcan>
but that would be one monster chip
<marcan>
M1 Max²
<opticron>
so current tally on that chopped off bit is 2x memory controllers, 2x display controllers, 32x functional GPU cores
<opticron>
fun
bgb has quit [Ping timeout: 480 seconds]
<j_ey>
brentr123[m]: assuming they give us bputil/kmutil etc, it just "works" not working around
doggkruse has joined #asahi
<i509vcb[m]>
dottedmag: "use our technology and don't go to others"
<j_ey>
i missed it, is there a touchbar?
<marcan>
pretty sure no touchbar
<j_ey>
4.7 pounds? thats cheap!
<kov>
didn't see it but they added that thing at the top of the screen like on the iphone
<j_ey>
kov: the notch
<opticron>
the notch
<i509vcb[m]>
hol up
<kov>
guess it's the camera
<psykose>
no touch
<i509vcb[m]>
notch on laptops?
<marcan>
notch.
* marcan
puts "add notch support to kwin" on the TODO list...
<dottedmag>
"brought it to laptop" lol
<i509vcb[m]>
concern
<brentr123[m]>
YESSS
<kov>
it'll nicely cover gnome's date/time
<brentr123[m]>
REGULAR F1 KEYSS
<j_ey>
sdcard is weird
<marcan>
but will it do SDExpress?
<marcan>
external PCIe ports are fun :p
<jix>
how would a notch even work with macos? they still have the menu bar don't they?
<j_ey>
marcan: thatts what your setup should look ike
<VinDuv>
it’s split looks like
<jix>
lol they just add a gap to the menu bar?
<marcan>
j_ey: lol
<kov>
lol the menu items spilling to the right
<psykose>
wonder how many times they've said 'entirely new level'
<marcan>
not gonna lie, I thought I'd end up with a new iMac but at this rate my new main box might be the big MBP replacing my old laptop and driving a 4K display or so
<opticron>
ugh, and now we have two current techs called "magsafe"
<kov>
if it were left side app, right side system it would be annoying, this is just dumb
<marcan>
which means going back to my old "laptop is main" setup
<i509vcb[m]>
Please other laptop manufacturers DO NOT start using notches
<VinDuv>
I wonder what happen to the cursor around the notch
<kov>
marcan, that's what I'm pondering doing as well
<kov>
MBP as main, but the 14" one
<Emantor>
They fixed the keyboard, SD slot, HDMI, touchbar gone, more display controllers than current gen intel… not bad.
<marcan>
ooh 120hz
<marcan>
Emantor: they've apparently learned to admit to mistakes, which is new for Apple
<marcan>
half of it is going back to 2015 MBPs
<i509vcb[m]>
120hz isn't the worst thing
<i509vcb[m]>
Arguably should be standard
<j_ey>
marcan: yeah, i really thought they'd never 'admit' touchbar was bad
<opticron>
I like variable refresh rate...
<i509vcb[m]>
(with vrr for battery life)
<opticron>
I guess I'm getting a MBP m1p in the nearish future
<VinDuv>
When they brought back the FireWire and the SD card slot on the 2009 MBP I don’t think they mentioned it so clearly, at least
<marcan>
if only HDR worked well on linux
<marcan>
:p
<j_ey>
marcan: add it to the TODO list
<Emantor>
HDR is coming slowly with the wayland color management extensions.
<marcan>
you know, I kind of hope once we get all the core stuff working smoothly, this inspires more devs to care about things like color management and HDR
<dottedmag>
j_ey: That's Wayland TODO list, or at least compositors' one
<marcan>
since right now it's still... questionable
<opticron>
I have no idea how well linux handles VRR
<i509vcb[m]>
You do need to get some protocol for everyone to agree to in wayland or else it will be just like screen capture
<j_ey>
isnt it weird that speakers are basically just very controlled fans
<opticron>
I still have a dream of eventually building an exotic home theater system with at least a rotary sub and a plasma tweeter
<opticron>
probably some panel exciters for the mid range
<j_ey>
Im a caveman, just use my macbook air's speakers
<marcan>
now the question is how many configurations of these do I have to buy...
<VinDuv>
:D
<JTL>
lolol
<marcan>
can I get away with a big max one and a small pro one... or do I need both configs for the small one too...
<i509vcb[m]>
I do wonder has anyone ever figured out the touch bar or is it still a black box
<JTL>
I was wondering the same for your sake
Glanzmann has joined #asahi
<j_ey>
is there going to be a new mini?
<jeh[m]>
I'm hating the notch design :p I'll stick with my MBA only xD.
<marcan>
I swear I'm going to end up with a shelf dedicated to stocking macs
<j_ey>
marcan: finish cleaning your apartment!
<marcan>
j_ey: definitely, but I'm smelling it not getting announced today... but I guess we'll see?
<marcan>
j_ey: ;_;
<marcan>
i509vcb[m]: on what macs?
<marcan>
it's been figured out on T2s
<brentr123[m]>
this is why people need to keep donating to marcans patreon
<marcan>
on M1s it's just some other display hardware nobody has bothered to look at yet
<phire_>
"you can record a guitar with the built-in mic array"
<Glanzmann>
Did the already say how many cpu cores the air and pro have?
<phire_>
which is optimistic
<i509vcb[m]>
So the drivers exist on T2 but not M1?
<marcan>
Glanzmann: 8+2
<j_ey>
Glanzmann: 10
<kov>
I may need to double my patreon donation =P
<Glanzmann>
wow.
<Glanzmann>
macbook pro has 64 GB ram, right?
<marcan>
max
<kov>
I bet the 16" with max will be expensive
<Glanzmann>
what about macbook air?
<marcan>
they didn't announce any airs
<marcan>
I don't expect them to
<mini>
the air will be unchanged, I suspect
<marcan>
the air will be M1-only
<mini>
the air is a base model after all
<marcan>
there's no way they can put these chips in a fanless laptop
<Glanzmann>
I see.
<j_ey>
i want a new mba!
<kov>
yeah, I'd expect them to do Air refresh next year with "M2"
<marcan>
on the plus side: m1n1 is still appropriately named
<Glanzmann>
Anything new on the mini?
<kov>
not yet
<marcan>
no(t yet?)
<Glanzmann>
I already hate the notch.
<kov>
they still talking about macbooks
<Glanzmann>
I see.
<j_ey>
kov: yeah jeez, we get it now
<marcan>
"We’re making updates to the Apple Store. Check back soon."
<marcan>
wonder when they go on sale
<j_ey>
f5 f5 f5 f5 f5
<i509vcb[m]>
I do wonder is there a future where all the iDevices and macs share the same cpus? Given the M1 ipad has come to exist
<opticron>
lort, up to 8T of storage
<mini>
I don't want a 30W+ chip in my phone. I like my hands unmelted
<i509vcb[m]>
sure cores and gpus varying
<j_ey>
Glanzmann: the notch will be.. interesting
<marcan>
tbh I hope they take a bit of time to ship because this week I'm swamped
<rohin>
hardware looks good, too bad macOS feels like MEGA blocks
<marcan>
let me do the bringup stream next monday plz
<marcan>
:p
<kov>
well
<kov>
next week it is marcan
<kov>
haha
<j_ey>
no mini i guess
<jeh[m]>
j_ey: I wonder how it'll be handled in Linux
<kov>
yeah, looks like they're wrapping up on M1 Pro
<marcan>
jeh[m]: that's the window manager's job
bgb has quit [Ping timeout: 480 seconds]
<kov>
would it be possible to just make the display controller "ignore" that part of the screen?
<j_ey>
jeh[m]: i would maybe just be ok with having it act like a full border?
<opticron>
"magic keyboard" did I miss an explanation somewhere of that?
<j_ey>
kov: ^
<i509vcb[m]>
jeh[m]: "Pretend it exists when rendering"
<marcan>
yes, we could make DCP have a "just pretend the notch isn't there" mode
<j_ey>
opticron: magic as in.. not touchbar lol
<opticron>
hah
<marcan>
at least as a stopgap
<marcan>
that's easy
<kov>
because it will be ok for gnome and kde I think, but as soon as you go fullscreen on a game it's game over
<jannau>
done and the store changed to almost ready
<JTL>
Is there prior art for the Linux dispaly stack or other components handling notches in displays?
<marcan>
kov: the window manager will make sure fullscreen stops at the notch
<Glanzmann>
no children were harmed in this video.
<Glanzmann>
Okay, very nice.
<marcan>
no sale date?
<kov>
I guess they can do that unless they do compositor pass-through?
<jeh[m]>
No new Mac Mini - very interesting!
<kov>
marcan, they said today
<marcan>
"Just a few finishing touches and the Apple Store will be ready. See you soon."
<marcan>
ah
<kov>
marcan, well pre-orders
<jannau>
order today. shipment next week
<marcan>
oh
<marcan>
it's up
<kov>
but available from next week
<balrog>
I expect a desktop Mac event in Feb or so, with iMacs and Mac Minis
<kov>
so yeah
<marcan>
it just opened now
<kov>
you got your week marcan
<kov>
xD
<phire_>
I don't have that much of a problem with the notch. I see it as the top border being pushed up along most of the top
<kov>
let me see if brazilians were also blessed
<marcan>
welp
<marcan>
how many configs is this now...
<Glanzmann>
How much do think the pro or max or whatever it is called (the best machine) will cost with everything maxed out?
<phire_>
but I am going to have to move gnome's clock
<marcan>
8-core CPU? is this a cut down 6+2 model?
<j_ey>
marcan: i support not buying everything
<kov>
yep, available here as well
<opticron>
whoah...yeah
<kov>
costing a couple of kidneys
<opticron>
I'm seeing that, too, no 32core GPU on the 14"
<fridtjof[m]>
shipping with a 140W USB-C power adapter??
<fridtjof[m]>
doesn't PD only go up to 100W?
<opticron>
ehhhh
<kov>
14" base model is priced at 4.9kUSD here, welcome to Brazil ;D
<opticron>
the new PD goes up to 240W or something like that
<Emantor>
2249€ in germany
<Glanzmann>
I see the store is ready.
<fridtjof[m]>
oh, also USB-C to magsafe 3
<fridtjof[m]>
interesting
<marcan>
I guess it's going to be a Pro 14-inch and a Max 16-inch for now
<opticron>
the previous PD technically goes up to 150W or so, but I think it requires non-standard power profiles
<Emantor>
GPU also scales up, multiple M1 & M1 Max variants.
<marcan>
... and then we'll see if I need more in the future...
<j_ey>
marcan: tell us the final $ amount :3
<marcan>
definitely going to be a total idiot on the 16-inch though, because hey this actually sounds like a machine I'd use
<jannau>
store is slow for me in Germany
<sven>
aw.. no new mac minis
<Glanzmann>
For me it is fast.
<marcan>
$8200 in the cart...
<j_ey>
sven: save your money!
<jannau>
there is 24 core gpu on thr 14" model
<sven>
j_ey: oh yeah, i won't buy any of these
<marcan>
wait, can I even put this on my credit card? :D
<Glanzmann>
The 16-inch macbook pro maxed out costs 6099 USD.
<VinDuv>
ok that’s a bit deceptive, going from M1 Pro to M1 Max adds 230€ but it forces 32 GB RAM which adds 460€
<marcan>
I'm hoping 4TB SSD is enough
<Glanzmann>
marcan: Depends on your credit card.
<brentr123[m]>
Most should
<j_ey>
take out a small business loan
<marcan>
yeah not on mine lol
<marcan>
not at my current usage anyway
<sven>
pretty sure i'd get a call from bank if i put that on my card just because it's unusual activity :D
<marcan>
japanese CCs are kinda... crappy
<j_ey>
sven: like when I got a call from my bank cos AWS billed me $1
<VinDuv>
lol
<sven>
lol
<JTL>
marcan's card gets declined from ordering several M1 systems at 2 in the morning
<JTL>
</joke>
bgb has joined #asahi
<marcan>
I actually have a second card, but I'm not sure that one is much better
<marcan>
OTOH pretty sure there are other payment methods?
<marcan>
let's see
<j_ey>
offer to do a linux port, if they donate some machines
<sven>
my bank is pretty reasonable. i think i went over the card limit once because i had to put some business travel on it and all they did was call me and ask me to confirm
<marcan>
also I'm not even in the right shop, lol
<Glanzmann>
In Germany they ship Nov 3 - 5 if you order right now.
<ah-[m]>
ordered :)
<j_ey>
ah-[m]: which one?
<ah-[m]>
Delivers Oct 29 - Nov 2
<ah-[m]>
16 inch 64GB
<j_ey>
big boi
<marcan>
961,600円
<marcan>
lol
<Glanzmann>
I'll not order, I like the macbook air. And I have other systems if I need bigbois.
<marcan>
I swear I'm doing this for the project :p
<marcan>
seems like a resonable combination for "computers I'd actually use"
<marcan>
if I end up needing more variants I'll get minimal specs for those...
<Glanzmann>
Previously I always spend approx. 2500 EUR every 3 or so years on a new thinkpad.
<jannau>
14"/pro/10c24g/32GB/2TB ordered
<Glanzmann>
Now that the macboos have so many gpus, we need a gpu bringup stream.
<marcan>
16" arriving next tuesday, 14" arriving a bit later ish
<marcan>
if I can get them ordered..
<j_ey>
jannau: nicenice
<marcan>
blergh, card only
<marcan>
maybe I can put one on each card...
* tophevich[m]
is curious how much impact these newly released macbook pro models will have on developers getting itchy to contribute to this project, as they get hungry to have that hardware as daily drivers :D
<opticron>
the website was having a really bad time for me earlier
<psykose>
do you normally pay cash? :D
<Glanzmann>
marcan: Just try it, if it is a debit credit card, it probably will just work.
<marcan>
psykose: I normally don't put things on my card that exceed the limit
<marcan>
Glanzmann: it won't
<marcan>
checking the limit on the other one
<psykose>
ahh, yes that would not be very good
bgb has quit [Ping timeout: 480 seconds]
<marcan>
the problem is precisely that I've been paying for *everything* with card lately, and due to the way they count the limit over more than one month, that kinda sucks for it
<marcan>
yeah I can't even put a single machine on the other card, lovely
<rohin>
i think i will give it a couple months, like the ipad pro hype, it went 200 bucks off 3 months after launch lol
<j_ey>
I dunno what my limit is, i guess that means I never got close :P
<marcan>
why doesn't apple allow paying via bank transfer like most big ticket japanese businesses, sigh
<marcan>
I guess this is the part where I start putting "get better credit cards" on the business expenses list... :/
<tophevich[m]>
hm apple online shop seems to struggle right now (at least for me)
<kov>
meh, can't pre-order yet it seems
<j_ey>
marcan: I thought you just go to the bank and theyre happy to raise it
<marcan>
this is japan.
___nick___ has quit []
<Glanzmann>
marcan: I used to have good credits card, but they required a not rooted phone to purchase in the internet as only option, as a result I cancled it. These days I have to debit visa cards, both never let me down even for very high charges.
<marcan>
so no.
<marcan>
anwyay, I ordered just the 14"...
<j_ey>
marcan: never been, I should visit
<marcan>
maybe I can put the 16" on there in a couple weeks after card pay day
<marcan>
お届け予定日 2021/10/28 - 2021/11/0114インチMacBook Pro
<Glanzmann>
marcan: In Germany I had also problems getting a decent credit card for a very long time, but than I financed a flat with one bank and from then on, it is private banking. If I want something from them (credit card, credit, ...) I call them and two days later I have it.
bgb has joined #asahi
<jix>
I have a credit card where I can extend the limit on the fly by transfering funds into a separate account, essentially turning it into a debit card... but it still has a normal credit limit when I don't do that and it's still accepted at places that don't accept debit cards
<Glanzmann>
jix: DKB?
<Glanzmann>
c/elar
<jix>
Glanzmann: yeah DKB
<Glanzmann>
jix: I hated that, because the credit card account was also the one where you got the interest. So I had a lot of money on that account and so I never used the credit card because I was afraid of abuse. And once I used it in russia, guess what ...
<Glanzmann>
Got the money back, only a few thousand bugs, but I had to go the police and file a report, proof that I was not in russia when the transactions happened, etc, etc. cost me whole day.
<marcan>
apparently I can get a prepaid Visa from my bank, maybe that's the safe option for this kind of thing
<marcan>
what's the limit on it though...
<marcan>
ok, yeah, ¥2m max balance, that'll work
<marcan>
though ¥1m/mo max spending, lol, what's the point of more than that then?
bgb has quit [Ping timeout: 480 seconds]
<marcan>
still, would've covered these two macs at least
<kov>
apparently brazil's telecom agency hasn't approved the new macbooks yet so they aren't even allowed to accept pre-orders gah
<sven>
uh. i got a credit card in .de with like a 2k eur limit as a student without any kind of income. there are enough sketchy/awesome banks around :D
<marcan>
sven: the problem is this one ends up applying the limit over 2 months or so
<marcan>
just due to the stupid way billing works
<sven>
that sounds dumb :/
<tophevich[m]>
hm ... airpods pro cost the exact same in my country as the airpods 3rd gen o_O
<marcan>
so I have a $15k equiv limit, but in the end it's more like $7k/mo
<marcan>
and I've been using that card for almost everything (because it gives me some points), so most of that limit is already gone
<marcan>
(and now after this purchase it *really* is on the edge)
___nick___ has joined #asahi
<Glanzmann>
sven: When I was 18 I was declined from every bank I applied for a credit card. But then came barcalay card and others around. 1600 EUR/ month ...
<marcan>
I wish I could just press a button and pay off the whole balance early and reset everything
<sven>
yeah barclays was the bank that gave me that credit card when i was 19 or so as well *shrug*
<marcan>
but alas I don't think that's an option
<marcan>
anyway, pay date is in a week
<marcan>
so it won't be too long
<jix>
Glanzmann: ah well, I'm not leaving any money in that account, and being able to quickly raise my limit when I need it was really convenient a few times
* tophevich[m]
can't buy anything directly at apple because they only support credit cards for payment ...
phire_ has quit [Ping timeout: 480 seconds]
<dottedmag>
I remember having a credit card without a limit (probably they'd call me if I tried to buy an island with it though), was quite convenient.
<dottedmag>
Too bad Eurocard is only in nordics nowadays
<milek7>
>at places that don't accept debit cards
<milek7>
huh, does such places exist?
<Glanzmann>
marcan:
<Glanzmann>
marcan: When does the one that you ordered, arrive?
<marcan>
I couldn't set an exact date like with the 16"
<marcan>
so sometime next week probably
<marcan>
depending on timing, since I'm busy that weekend, the bring-up might be the following monday
<kode54>
don't forget
<kode54>
Max isn't just a giant GPU
<jix>
milek7: "Virtual or prepaid credit or debit cards aren't accepted as payment for Azure subscriptions." is one place where I ran into that
<nobodynada>
I get mine a week from tomorrow
<kode54>
it's also the only 64GB config
<marcan>
nobodynada: that's what it would've been like for the 16"
<marcan>
kode54: I know, that's why I'm getting a max too
<marcan>
... as soon as my CC lets me
<kode54>
$3500 fricking eh
<marcan>
more with more SSD...
<kode54>
true
<kode54>
I bet that $3500 config only comes with what 256GB?
<kode54>
I didn't read that much into it, oops
<marcan>
nah, 1TB I think?
<kode54>
oh damn
<jannau>
the max has also twice the memory bandwidth
<kode54>
that's slightly cheaper than the overkill iMac 5k I bought back in 2014
<kode54>
had so many troubles with that thing
<kode54>
not the least of which was that the display controller told not-macOS that it was a single channel 4k panel
<kode54>
Linux never fixed that until after I sold it
bgb has joined #asahi
<kode54>
even Windows had trouble keeping the two display halves synced
___nick___ has quit []
<kode54>
for instance, the onboarding wizard with the fading background, you could see the two halves were fading out of sync
___nick___ has joined #asahi
<jannau>
oh, the base 14" model has just 6 performance cores (14 GPU core)
<kode54>
yes, it costs as much as the base 16" model just to get a properly binned 14" model
<kode54>
at least they're not offering 8GB configs
<nobodynada>
question, how likely am I to need a UART adapter? it looks like that USB-C controller board on the Asahi website has been chip-shortaged out of existence
<nobodynada>
can I run Linux in the hypervisor and use its vUART?
<jannau>
thunderbolt 4
<marcan>
kode54: linux hasn't fixed that yet
<marcan>
I have one of those
<j_ey>
nobodynada: yes
<marcan>
been meaning to fix the 5k thing... some day...
<marcan>
when I get bored... if that ever happens...
<j_ey>
nobodynada: I use a normal USB cable with linux+vuart
<nobodynada>
awesome, thanks
<marcan>
they 5K support is in the driver, linux one needs patches
<marcan>
gpu performance switching also doesn't work... that's another one I need to annoy the amdgpu guys about one day...
<nobodynada>
I have a Cypress PSoC devkit whose USB controller might be...repurposable, but I really don't want to do that if I don't have to
<marcan>
but at this rate this computer will be retired for an M1 before tha happens
<marcan>
*that
<j_ey>
marcan: but your precious rootfs :P
bgb has quit [Ping timeout: 480 seconds]
<marcan>
j_ey: this will be the end of hostname raider, after all these years
<kode54>
I already retired the 5k, it was throttling too much
<Glanzmann>
nobodynada: You can also use a second mac, if you have one around.
<kode54>
I actually on numerous occasions managed to make it hit the thermal cutoff of 106°C
<marcan>
I have a 2015 one that I maxed out the CPU on last year (got it from a friend) and it's not throttling too badly
<marcan>
but yeah it's not exactly amazing either
<kode54>
neither Apple nor AASP documented or knew of this cutoff
<j_ey>
nobodynada: did you buy one of the current M1s?
<kode54>
when it hits 106, it goes to sleep
<nobodynada>
I just preordered a pro
<kode54>
AASP swore it was a defective board and replaced it (under applecare)
<j_ey>
nobodynada: but one of the previous revisions I mean'
<ah-[m]>
Imean 106 is a little toasty
<kode54>
4790K and R9 M295X *easily* hit 100C under load
<nobodynada>
j_ey: I don't have one, though maybe I can ask some friends who do if they'd be down for kernel hacking sessions
<kode54>
AASP also never heard of retina desktops before
<kode54>
managed to stress fracture the display getting it off to do the board replacement
<j_ey>
nobodynada: because what Glanzmann was alluding to, was that two M1 machines can act like a hardware uart
<j_ey>
nobodynada: but it works fine with jus teh vUART mostly
<kode54>
I'll take further discussion of the 5k to offtopic
<nobodynada>
j_ey: yeah, I saw that on the website
<nobodynada>
j_ey: awesome, thanks!
<j_ey>
nobodynada: the good thing about the hardware uart is that you can reboot the target machine remotely without using the hypervisor
<nobodynada>
j_ey: can you do that with a 2nd mac, or do you need the hardware adapter?
<jannau>
both
<j_ey>
^
<j_ey>
the only time ive booted without the hypervisor so far is to test kvm patches
<jannau>
a second m1 mac has the same functionality as serial adapter
<Glanzmann>
nobodynada: But you need one of these good usb cables. The one that shipped with my eizo works for me.
<Glanzmann>
The charging one from apple, doesn't.
doggkruse has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<nobodynada>
Glanzmann: ugh, guess I'm making a trip to best buy then
bgb has joined #asahi
<Glanzmann>
nobodynada: The good usb cable you only need to use a second m1 as a uart replacement. vuart works with a normal cable.
___nick___ has quit [Ping timeout: 480 seconds]
<marcan>
anyway, I should sleep
<nobodynada>
wow, it looks like they're already selling out
<nobodynada>
some of the models are showing shipment dates in November now
<ah-[m]>
michael fast as always
<Leidenfrost>
Yeah I prefer to wait until the frenzy ends and the price drops
<marcan>
nobodynada: yeah, the 16" is already on the same timescale as the 13" here now :/
bgb has quit [Ping timeout: 480 seconds]
tomtastic_ has joined #asahi
tomtastic has quit [Ping timeout: 480 seconds]
tomtastic_ has quit [Remote host closed the connection]
tomtastic has joined #asahi
<mini>
the custom build 14" I ordered is second week of november for ship date already
<mini>
but I imagine there's a *lot* of demand
bgb has joined #asahi
nehsou^ has quit [Ping timeout: 480 seconds]
nehsou^ has joined #asahi
jelly-hme is now known as jelly
bgb has quit [Ping timeout: 480 seconds]
<balrog>
mini: custom built always have longer lead times.
<mini>
balrog: yeah, I know. but that's already slipped by ~2 weeks since the store opened
<mini>
or at least a week and a half
tomtastic_ has joined #asahi
bgb has joined #asahi
bps has joined #asahi
doggkruse has joined #asahi
tomtastic has quit [Ping timeout: 480 seconds]
doggkruse has quit []
tomtastic has joined #asahi
bgb has quit [Ping timeout: 480 seconds]
tomtastic_ has quit [Ping timeout: 480 seconds]
bgb has joined #asahi
<kettenis>
heh, thaat certainly caused a lot of noise ;)
<balrog>
kettenis: what did?
<kode54>
the page design is kind of broken
<kode54>
there is no 14" on the main page
<kode54>
you have to click 16" to get to the 14" option
<mini>
the store is working pretty well in the UK now at least, it melted quite badly immediately after the stream finished
<mini>
although for a while it was still showing the *intel* 16" after they reopened it
<mini>
on some pages anyway
<kettenis>
the release of the new macs
bgb has quit [Ping timeout: 480 seconds]
malvo has quit [Ping timeout: 480 seconds]
<Glanzmann>
kettenis: What is the status of OpenBSD on m1?
<kettenis>
still issues with SMP :(
<kettenis>
but quite solid with a non-MP kernel
<kettenis>
going to see if I can get the laptop keyboard work in the next week
<j_ey>
kettenis: cool!
<Glanzmann>
kettenis: I see. I used the mini on Linux for a few days, works good so far. Did not had any issues beside the dcp, but once the display is up it is stable.
<Glanzmann>
kettenis: Did you do performance benchmarks on the nvme?
<Glanzmann>
Under OpenBSD?
<kettenis>
not systematically
<kettenis>
the driver uses a queue depth of 1, so that impacts performance I bit
<Glanzmann>
kettenis: Could you try this command and let me know how many iops do you get? fio --ioengine=sync --direct=0 --fsync_on_close=1 --randrepeat=0 --nrfiles=1 --name=randwrite --rw=randwrite --bs=4K --size=1G --end_fsync=1 --fallocate=none --overwrite=0 --numjobs=1 --sync=1 --directory=/mnt
<jmr2>
I don't plan to maintain that doc - feel free to merge it in yours - or in the project's wiki.
bgb has quit [Ping timeout: 480 seconds]
<Glanzmann>
kettenis: Yes, but you can also paste the whole output.
<Glanzmann>
jmr2: Thank you.
<mjg59>
marcan: I'm sure you mentioned this at some point, but I can't find anything online - does M1 implement Trustzone?
bgb has joined #asahi
<j_ey>
no, there's no secure ELs
<sven>
mjg59: they don't have EL3. the only part of trustzone that they implement is the memory protection.
<sven>
they can setup a carveout area that only their secure co-processor can access
<mjg59>
Thanks!
bgb has quit [Ping timeout: 480 seconds]
bgb has joined #asahi
bps has quit [Ping timeout: 480 seconds]
bgb has quit [Ping timeout: 480 seconds]
bgb has joined #asahi
bgb has quit [Ping timeout: 480 seconds]
bgb has joined #asahi
<Glanzmann>
sven / jey: I did not pin the core, but the point is the random write 4k are really bad.
<sven>
did you run cpufreq before?
jmr2 has quit [Remote host closed the connection]
<sven>
i don't know how that benchmark works, it might be heavily cpu bound
<sven>
the number is still low but if it was running on a heavily underclocked performance core who knows
bps has joined #asahi
malvo has joined #asahi
bgb has quit [Ping timeout: 480 seconds]
<balrog>
These M1 Pro and M1 Max don't look like multi-die
<landscape15[m]>
Do you think the less efficiency cores in the new chips will affect a lot the CPU drivers?
<kettenis>
Glanzmann: I will do once I actually run this on an m1 machine
bgb has joined #asahi
<Glanzmann>
sven: I ran it multiple times. I notices the bad performance immediatly after I was installing my first debian package using apt. I ran 80 threads in parallel. I get 1000 IOPS. Look at the CPU: https://ab34.de/u/screenshot-nuc-2021-10-18-23_16_33.png
<sven>
again: did you run cpufreq before?
<sven>
or does your tree have the cpufreq driver?
<Glanzmann>
sven: No, I did not. I ran alyssas tree.
<Glanzmann>
Let me check.
<Glanzmann>
Yes, the driver is in.
<sven>
hrm, ok. so it's actually running with full speed probably
<Glanzmann>
sven: Even if it didn't its way to bad.
<sven>
*sigh*
<Glanzmann>
But, if you want I can also run the m1n1 cpufreq experiment.
<sven>
<sven> i don't know how that benchmark works, it might be heavily cpu bound
<sven>
but if the cores are running at full speed it's a different issue
<Glanzmann>
sven: You saw my screenshot. This is 80 threads, I ran on of them.
Leidenfrost has quit [Quit: Connection closed]
<j_ey>
you can probably look at some /sys values to find the frequency