jjido has quit [Quit: My MacBook Air has gone to sleep. ZZZzzz…]
Dcow has quit [Remote host closed the connection]
Dcow has joined #asahi
Etrien has quit [Ping timeout: 480 seconds]
thelounge7571340 has joined #asahi
thelounge7571340 has quit [Remote host closed the connection]
Dcow has quit [Remote host closed the connection]
Dcow has joined #asahi
Dcow has quit [Ping timeout: 480 seconds]
Dcow has joined #asahi
kazuki has joined #asahi
chadmed has joined #asahi
Dcow has quit [Ping timeout: 480 seconds]
<chadmed>
ChaosPrincess, scardracs: if you havent already, it's probably worth your while to use at least the portage overlay just to reduce your own maintenance burden and outsource that to me
<chadmed>
kazuki: they came straight out of the cluster and core joule counters and match a cruder mechanism i was using previously based on the SMC's idea of power consumption
<chadmed>
i think i mightve said it on m a r c a n's stream, but we actually arent *entirely* sure theyre correct (the counters are weird) but he merged them anyway to prove that EAS works in principle
<chadmed>
m p s, j a n n a u and i smoke tested them for a while when i first worked on it and there were no regressions at the very least :P
<kazuki>
I see, I was curious since benchmarks from others show a very different picture
<chadmed>
like, i sincerely doubt specint 06 has any awareness (or access to under macos) the core's perf counter registers...
<kazuki>
Iirc the person who did the benchmark hooked up the phone to a power meter
Etrien has joined #asahi
<chadmed>
so thats really going to be whole of soc power for the load
<chadmed>
which doesnt actually tell us much about what the core clusters are doing
<kazuki>
That's true but I think your values are too low even considering that
<chadmed>
ill spend some time refining the methodology when i have a few days to dedicate to it. fwiw i dont entirely disagree, and like i said we werent really sure ourselves if they were legit or not.
<chadmed>
basically what i did was read the joule counters, run a mix of integer and floating point instructions, measure how long they took, read the joule counters again, take deltaE and then divide by the time the instructions took
<chadmed>
(in m1n1)
<chadmed>
i dont think the methodology itself is _particularly_ flawed, for example there is a very noticeable increase in power use when you light up the FPUs
<chadmed>
iirc the max opp-microwatt value for firestorm was ~27k or something ridiculous without floating point instructions
thelounge7571340 has joined #asahi
thelounge7571340 has quit [Remote host closed the connection]
<kazuki>
Hm that seems super low
<chadmed>
yeah i know, at first i thought we might not be initialising the counters properly, but like i said it lines up with an earlier methodology where i would use the SMC's integrated power meter. the values are within a few % of each other
<chadmed>
not really willing to scratch up traces on the logic board of my daily driver to find out though :P
<kazuki>
I think the CPU capacities were calculated with this benchmark
Dcow has joined #asahi
Etrien__ has joined #asahi
<chadmed>
ok no this isnt going to work for power sampling
<Tramtrist>
Is there a large asahi presence on Mastodon? I see a large presence on twitter but cant seem to find equivalent mastodon post volume.. Im on fosstodon myself
<chadmed>
or rather, if it _did_ work, it would give the value out of the SMC which will include the entire system power draw and not just one core
<chadmed>
but it wouldnt work because it has no conception of the SMC and im not going to add it just to arrive at the same conclusion i did in m1n1
Etrien has quit [Ping timeout: 480 seconds]
Dcow has quit [Ping timeout: 480 seconds]
<kazuki>
Yes but does it matter if *all* the power metrics have the entire system power draw included?
<chadmed>
probably not, but if its going to get the power value out of the SMC then its going to show the same thing i got in m1n1 when i did it, which matched the values from the joule counters, which is what is in the DT
<chadmed>
and like if we're going to say that what matters is the scaling between cores for the EM and the actual value doesnt matter, then it doesnt matter if theyre too low either since all cores and pstates were tested exactly the same way
<chadmed>
so the scaling between them should be consistent
<kazuki>
Yeah true
<chadmed>
youve nerd sniped me though, building a kernel now :P
<kazuki>
Heh, what for though?
<chadmed>
dont have the userspace scheduler enabled in my normal build, and my timer frequency is 1khz
<chadmed>
im setting the options recommended in the readme
chadmed has quit [Quit: Konversation terminated!]
chadmed has joined #asahi
LunaFoxgirlVT has joined #asahi
marve has joined #asahi
das_j has quit [Remote host closed the connection]
marve_ has joined #asahi
das_j has joined #asahi
thelounge7571340 has joined #asahi
thelounge7571340 has quit [Remote host closed the connection]
<chadmed>
yeah this is utterly useless, its just noise
<chadmed>
sorry
<chadmed>
-300mW uh-huh yeah sure
<chadmed>
jim keller found a way to create energy from running instructions
marve has quit [Ping timeout: 480 seconds]
<kazuki>
chadmed: 300mw at lowest freq?
<chadmed>
no as in NEGATIVE 300mW
<chadmed>
the core creates energy
<kazuki>
Yeah idk how you got that
dviola has quit [Ping timeout: 480 seconds]
<chadmed>
because its like i predicted
<chadmed>
the system is doing a load of other stuff, and the work the single core is doing at 600mhz just falls into background noise
<kazuki>
Results still shouldn't be negative though...
<chadmed>
no they shouldnt, which is why i use m1n1 and hit the energy counters inside the cores directly to avoid this exact thing happening, where the total system power turns the work a single core is doing into noise
<chadmed>
i think the cores really just are this efficient
<chadmed>
"that doesnt look right, the snapdragon uses more power" really doesnt invalidate the methodology i checked over multiple times and cross referenced with an entirely different one tbh
<kazuki>
The pic I sent contains power metrics for A14
<chadmed>
there is divergence at the top end of the pcore results at the highest pstates but below that its literally just noise and there is no way to verify that from userspace at this time
<chadmed>
yeah i understand that but again, that measures whole soc power on cores that are tuned for different pstates at different V-P curves
<chadmed>
its not really comparable
chadmed has quit [Quit: Page closed]
chadmed has joined #asahi
<kazuki>
There's no way the P-core at max freq only uses 0.127827W
<kazuki>
You missed a digit or smth
<chadmed>
im going to run it one more time
chadmed has quit []
<kazuki>
I can understand if it's 12W though
chadmed has joined #asahi
<chadmed>
ok now things are making a little more sense
<kazuki>
How was it?
<chadmed>
results ranging from 100mW to 5W
<chadmed>
so a little more sense
<kazuki>
Ah so more inline with what I expected
<kazuki>
Mind sending run.log?
<chadmed>
yeah one second
<chadmed>
nvm it was saved in tmpfs -.-
<chadmed>
dont have time to do it again but i captured the results
<kazuki>
Can you send those please?
kazuki has quit [Quit: Konversation terminated!]
kazuki has joined #asahi
<chadmed>
yep
thelounge7571340 has joined #asahi
thelounge7571340 has quit [Remote host closed the connection]
Etrien__ has quit [Read error: Connection reset by peer]
Etrien has joined #asahi
Etrien has quit [Ping timeout: 480 seconds]
SSJ_GZ has joined #asahi
chadmed has joined #asahi
julio7359 has quit [Ping timeout: 480 seconds]
mxz has quit [Quit: cya]
mxz has joined #asahi
thelounge7571340 has joined #asahi
thelounge7571340 has quit [Remote host closed the connection]
marve_ has quit [Ping timeout: 480 seconds]
<marcan>
chadmed: the first couple lines there for the E cores are suspicious, I wouldn't expect efficiency to drop at the lowest Pstate (the lowest pstate needs to be the peak efficiency, otherwise it makes no sense)
<marcan>
the rest look reasonable though!
<chadmed>
yeah im still not happy with the lowest pstates given the wide variance/noise in the smc power measurements. sorry i didnt save the run logs
<marcan>
ah, these are from SMC? or Ecounters?
<chadmed>
SMC, freqbench has no way to access the counters since it relies on voltage_now and current_now from power supply class devices
<chadmed>
when im at home and have a couple of days ill dig out the m1n1 experiment from the backup of my dev machine (RIP) and try to see what we were doing wrong there
<chadmed>
i reckon the benchmark asm just needs to light up more of each core
Etrien has joined #asahi
chadmed has quit [Quit: Konversation terminated!]
chadmed has joined #asahi
<marcan>
chadmed: make sure you enable the MMU on the seconaries, otherwise all mem accesses will be super slow
<marcan>
*secondaries
marve_ has joined #asahi
rvalue has quit [Remote host closed the connection]
rvalue has joined #asahi
julio7359 has joined #asahi
Etrien has quit [Read error: No route to host]
Etrien has joined #asahi
thelounge7571340 has joined #asahi
thelounge7571340 has quit [Remote host closed the connection]
chadmed has quit [Quit: Konversation terminated!]
marve_ has quit [Quit: Leaving]
LunaFoxgirlVT has quit [Read error: Connection reset by peer]
Etrien has quit [Ping timeout: 480 seconds]
Etrien has joined #asahi
mps has quit [Quit: Lost terminal]
Etrien has quit [Read error: Connection reset by peer]
Etrien has joined #asahi
will[m] has quit [Write error: connection closed]
nyanpasu64 has quit [Write error: connection closed]
thelounge7571340 has quit [Read error: Connection reset by peer]
kazuki has quit [Quit: Konversation terminated!]
kazuki has joined #asahi
<joske>
This might just be the completely wrong fix of course, just pointing out that out-of-the box tas2770 doesn't load on M1 mba
guillaume_g has joined #asahi
<marcan>
those errors are normal
<joske>
thx, I don't recall seeing them before, but I guess I was just not looking :-D
r0ni has quit [Quit: Leaving]
<marcan>
(we still need to do something about them...)
greguu has joined #asahi
bestouff has quit [Ping timeout: 480 seconds]
bestouff has joined #asahi
joske has quit [Quit: Leaving]
bestouff has quit [Read error: Connection reset by peer]
Etrien__ has joined #asahi
kazuki has quit [Quit: Konversation terminated!]
<rann>
marcan: quick question. In your experience diving in the internals of the M1 and M2, is there support for ECC memory hidden in there? Just curious!
bestouff has joined #asahi
Etrien has quit [Ping timeout: 480 seconds]
Etrien has joined #asahi
thelounge7571340 has joined #asahi
thelounge7571340 has quit [Remote host closed the connection]
Etrien__ has quit [Ping timeout: 480 seconds]
<scardracs>
ChaosPrincess, chadmed: did you emerge Firefox? How much time it takes?
<povik>
joske: hah. seems like a bug in tas2770.c triggered by a change in macaudio.c
<povik>
and your change could be the right fix!
<povik>
do you want to give me your name/email address so i can tag you in the commit?
<povik>
heh, that will be the *eighth* fix to that driver
<marcan>
rann: I don't think traditional ECC is or can be supported in LPDDR systems
* povik
checks where else TI copied this faulty code
<marcan>
the channels are too narrow
<povik>
it's there in tas2764.c
<marcan>
however, there could be ECC internal to the RAM itself, and some form of ECC used at the protocol level
<povik>
tas2780.c too
<ChaosPrincess>
scardracs: iirc it took long enough for me to kill it and use my server to build it.
<marcan>
the caches do have ECC though, that I know
<scardracs>
ChaosPrincess: as i fear. Unfortunately i don’t have a server for the pourpose
<marcan>
but yeah, traditional ECC is dead with LPDDR, though conversely I think ECC is becoming a necessity with LPDDR5 densities, so there's a good chance there's RAM-internal ECC support
<marcan>
but I wouldn't know how to find out
<marcan>
this is deep memory controller shenanigans
<marcan>
nor whether there is anything for the link
<ChaosPrincess>
scardracs: leave it overnight, or if you are ok with pre-built versions, iirc there are arm64 snaps.
<ChaosPrincess>
hot take - ram internal ecc is not 'real' ecc, its just a chip implementation detail that exists because it is impossible for the memory maker to achieve the capacity w/o it.
<scardracs>
ChaosPrincess: i have already checked and no snapshots available
<rann>
marcan: thanks, yeah, I have similar thoughts w/regards to current/future densities.. People running 128GiB without an ECC like meganism, it becomes statistically more likely for one to encounter issues
joske has joined #asahi
<joske>
povik: Jos Dehaes jos.dehaes@gmail.com
<joske>
I noticed only left channel works with that fix today though (although yesterday evening I thought sound was coming out of both speakers)
<joske>
yes the code in tas2770.c looked obviously wrong
le0n_ has joined #asahi
r0ni has joined #asahi
<povik>
joske: thanks! i think we should drop that slots==1 branch altogether, i will tag you as suggested-by
<povik>
not sure why would have left channel only, let me read the code again
<joske>
rebooted to macOS to check right speaker wasn't b0rked, it's not
greguu has quit [Quit: WeeChat 3.0]
<povik>
joske: can't figure out what would be the issue. how are you testing it?
le0n has quit [Ping timeout: 480 seconds]
<povik>
ah, we are missing fixup_controls for j313. in that case you will get left channel routed to both speakers on default settings of card controls
<povik>
could that be what's happening?
<joske>
ah yesterday I added that, but I removed it again for the diff
<povik>
what you mean? the fixup_controls? were you setting ASI1 Sel in there?
<povik>
in that case it's all explained
<joske>
I added the fixup_controls yes
chadmed has joined #asahi
thelounge7571340 has joined #asahi
thelounge7571340 has quit [Remote host closed the connection]
<joske>
but I removed it as I wasn't sure about that
<scardracs>
I cannot update grub because /boot/efi is read-only… any thought ChaosPrincess ?
<scardracs>
(I used to port genfstab to gentoo around 2 years ago)
<ChaosPrincess>
1. use LANG=C before posting screenshots :P, 2. it think the problem is not that /boot/efi is not writable, but rather that those systems are not 100% real efi, and it can't write to efi variables.
<ChaosPrincess>
i think there is a way to make it only install the binaries and skip writing to efi vars
<scardracs>
Is there something i can do to update the kernel?
<chadmed>
the overlay has the latest kernel, use it :)
<scardracs>
chadmed: i already did it but stucks here
<chadmed>
oh right yeah dont do grub-install
<chadmed>
or you can i guess, but you need to specify that its a removeable drive
<r0ni>
if I don't run grub-mkconfig but update the kernel and initramfs and they the same filenames... would it still boot?
<chadmed>
yeah
<r0ni>
ahh good thats what i was hoping
<chadmed>
grub-mkconfig really only scans /boot for changes and makes entries for each kernel/initramfs in there
<chadmed>
if the entry didnt change you dont have to run iot
<chadmed>
it*
<scardracs>
chadmed: it seems in build 102, which is stable, they (libs) are enabled by default
<r0ni>
well my setup is a little messy and every time i run grub-mkconfig it dies and i have to fix it from macos... and i'm tired
<r0ni>
so skipping that sounds like a winning plan
<j`ey>
r0ni: would still be good to figure out why :P
<r0ni>
well i'm sure it's the diff between slackware grub and arch grub, i just havent gotten the arch pkg yet to see how the /etc/default/grub is different
<r0ni>
i'm the king of cutting corners ;)
joske has joined #asahi
<joske>
povik: works fine!
<povik>
ey!
<povik>
off to the list with it
int16h has joined #asahi
<j`ey>
r0ni: nah its just about prioritising :)
danielnechtan has quit [Ping timeout: 480 seconds]
<chadmed>
work smarter, not harder
<r0ni>
well iirc grub isn't a required bit for booting anyway, so when i figure out how to bypass grub all-together, i'll likely go that route cuz slackware is KISS philosophy anyway
<r0ni>
i haven't had any time to play with this at all for days
<chadmed>
you can concatenate your kernel and initramfs to the m1n1 payload
<chadmed>
$EDITOR $(which update-m1n1) to see roughly how
<j`ey>
r0ni: leaving out grub doesnt necessarily equate to KISS though
<r0ni>
i'll look at it, i knew you just cat them together, just didn't know the finer details
<chadmed>
no thats pretty much it, you just cat them :P
<j`ey>
if you cat them, you have no menu if you want multiple kernels etc
<r0ni>
but what do i name it? where do i put it? (those details)
<chadmed>
having a bootloader is good though, for those reasons
<chadmed>
im looking into using sd-boot, my current rootfs uses ESP-as-boot so it should be painless
<j`ey>
it's pretty easy to reboot into recoveryOS of course, but a menu is nice too
<r0ni>
bootloader is good, but keeping this thing booting is 90% of my issues lol
<r0ni>
and grub menu is worthless for me, i have no kb inputs anyway on it
<j`ey>
huh, what machine?
<r0ni>
mini
<j`ey>
oh
<r0ni>
the dreaded "oh" :(
<j`ey>
:)
<r0ni>
I had hoped it was my fault lol
<chadmed>
oh as in your usb keyboards dont work?
<scardracs>
Any idea on when the inital m1/m2 support will reach the mainline kernel?
<chadmed>
m1 boots, m2 is still a long while away
<r0ni>
no, i had read it works, but i thought maybe it was usb-c only or smth else, hardware i don't have
<chadmed>
m1 pro/max/ultra DTs should be in for 6.2
<scardracs>
Lucky me when i choose to pay less and take the m1 🤣
<j`ey>
you still will want to use the asahi kernel for quite a while yet
<chadmed>
boots =/= useable
<r0ni>
what dtbs do i need for mini if I cat this?
<j`ey>
r0ni: just the same ones you currently use
<r0ni>
oh ok so all of them
<scardracs>
I know really well
<r0ni>
i see now, half reading the script
<scardracs>
I used to have an HP so I’m used to things not working as they should
<j`ey>
r0ni: you can just pick arch/arm64/boot/dts/apple/t8103-j274.dts, but catting all also works
<r0ni>
ya i didn't really want to go all in, just what i need... now for distrobution, that would be diff
<chadmed>
i mean youre saving 200kb on a 500mb partition
<chadmed>
m1n1 just picks the correct one anyway, it doesnt slow boot down or anything
thelounge7571340 has joined #asahi
thelounge7571340 has quit [Remote host closed the connection]
chadmed has quit [Quit: Konversation terminated!]
<r0ni>
ok so is the resulting file go /boot/efi/m1n1/boot.bin ? is that what i'm comprehending?
<j`ey>
yeah
<r0ni>
oh awesome, if this works you all get KISSes
<j`ey>
currently you use u-boot there, so instead of u-boot you can just put the kernel
<Illya>
does asahi have working hardware performance counters etc? thinking of trying it out if I can more easily benchmark arm64 code
borisk has quit [Quit: Page closed]
chadmed has joined #asahi
<chadmed>
someone was working on a perf counter driver iirc, i think it might already be in linux-asahi?
<mkurz>
OK, thanks. I was just interested if there is a "USB 3" repo that I can follow for current status.
nicolas17 has joined #asahi
<opticron>
the joke being that every time someone asks (especially in the wrong place) a month gets added to the release date
<nicolas17>
hello
<opticron>
hi nicolas17
<mkurz>
ah ok now I get it
<nicolas17>
I have an MBP M2 with a Spanish keyboard, it has a <> key next to Z, and a ªº\ key next to 1, but in asahi (as opposed to macos) they're swapped, any idea whose fault it is? :P
<sven>
joke? you think i'm joking? :P
<opticron>
mkurz, you can probably watch what's happening on the various branches in the asahi-linux repo
<j`ey>
ok only seems to have fixes for jp variants
<nicolas17>
there's issues for other layouts already
mkurz has quit [Quit: Leaving]
amarioguy2 has joined #asahi
<kettenis>
nicolas17: I believe that is a common quirk of Apple keyboards with ISO layout
thelounge7571340 has joined #asahi
thelounge7571340 has quit [Remote host closed the connection]
<kettenis>
so there probably is already support for this somewhere in Linux
<_jannau_>
nicolas17: in doubt apple's. we should set a quirk flag for apple-hid but unfortunately we have not enough information to make that decisions
<_jannau_>
the keyboard I'm using (iso internal is correct without the flag) and we can't distinguish that keyboard from other layouts for european languages
chadmed has quit [Remote host closed the connection]
<_jannau_>
I don't really and would prefer the keys swapped as well. I thought apple's international layout would be more less what pc international layout is
<nicolas17>
afaik the <> physical key location matches the spanish keyboard on other computers
<nicolas17>
what arch package includes xev?
amarioguy2 has quit [Remote host closed the connection]
<nicolas17>
keyboard settings had keyboard model set to "generic PC", I tried the multiple Apple variants, no change
<nicolas17>
then I went to layout (rather than model) settings, tried Spanish and Spanish (Macintosh) and Spanish (Windows) variants, no change
<nicolas17>
if I click preview layout, the image that shows up doesn't even have a <> key
<_jannau_>
nicolas17: set the iso_layout module parameter to 1 for hid-apple
<nicolas17>
yes that fixed it!
mrkajetanp has quit [Remote host closed the connection]
<_jannau_>
I'll chnage the default at the next time I look at the HID changes. I think users of the iso international layout are in the minority and it needs to swapped for most or all other layouts
<nicolas17>
it was -1 before I changed it
<_jannau_>
I think that's automatic meaning, respecting APPLE_ISO_TILDE_QUIRK from the device table
<scardracs>
ChaosPrincess, chadmed: how did you fix the build error of mmpeg? What’s the correct CHOST? Or did you remove —march=native?
<ChaosPrincess>
remove march-native, yes, it uses eor3 or sth else that m1 does not support
<scardracs>
Ok
eroc1990 has quit [Ping timeout: 480 seconds]
<ChaosPrincess>
use -march=armv8.2-a
<scardracs>
Perfect, thanks :)
<_jannau_>
-march=armv8.4-a would work as well if you do not care about binary compatibility with non-apple silicon CPUs
julio7359 has quit [Remote host closed the connection]
julio7359 has joined #asahi
eroc1990 has joined #asahi
<scardracs>
Not really interested on it
<scardracs>
Anyway Clang show the worst of it. The entire chassis is warm
<scardracs>
I don’t know if it’s a good idea to use Gentoo here without a proper refrigerator system
julio7359 has quit [Ping timeout: 480 seconds]
Gaspare has quit [Quit: Gaspare]
___nick___ has joined #asahi
julio7359 has joined #asahi
___nick___ has quit []
___nick___ has joined #asahi
thelounge7571340 has joined #asahi
thelounge7571340 has quit [Remote host closed the connection]
<opticron>
ChaosPrincess, does gcc just misdetect the cpu flags when using march-native? or does apple misreport its cpu flags?
<opticron>
or clang
<ChaosPrincess>
i dont know, i think there is a disagreement in the build pipeline, since gcc emits those instructions, but binutils refuses to assemble them
<_jannau_>
I think the issue is that gnu as doesn't support -march=native on arm64 and gcc doesn't adjust the -march value with the detected one
<_jannau_>
clang should be fine with the integrated assembler. with -fno-integrated-as it might run into the same problem as gcc
<opticron>
fun, I've definitely seen broken behavior on gcc with -march=native on older versions with virtualized intel CPUs being detected as supporting 3dnow instructions
<opticron>
that one's pretty historical at this point, though
<scardracs>
At the end I've removed the gentoo partition. It's fun to see linux on my mac but it only has around 40/50 days and I don't want to see it blow up
<ChaosPrincess>
thats kind of a weird take imo, those machines are capable of running at 100% cpu load, and they dont really overheat from that.
unicordian has quit [Remote host closed the connection]
Etrien__ has quit [Read error: Connection reset by peer]
Etrien has joined #asahi
DarkShadow44 has quit [Ping timeout: 480 seconds]
thansen has joined #asahi
uur has joined #asahi
unicordian has joined #asahi
unicordian has quit []
jjido has joined #asahi
uur has quit [Remote host closed the connection]
Etrien has quit [Ping timeout: 480 seconds]
SSJ_GZ has quit [Ping timeout: 480 seconds]
bestouff has quit [Read error: Connection reset by peer]
bestouff has joined #asahi
unicordian has joined #asahi
Gaspare has quit [Ping timeout: 480 seconds]
Etrien has joined #asahi
thelounge7571340 has joined #asahi
thelounge7571340 has quit [Remote host closed the connection]
unicordian has quit [Quit: unicordian]
unicordion has joined #asahi
assman has joined #asahi
<assman>
hi, general question, is there a way to have a mac mini with asahi turn after power loss? i know you can configure it from macos but doing it via linux would be nicer
<assman>
its probably faster to just go into macOS to configure it but I figured I would ask
jacksonchen666 has joined #asahi
millenialhacker has joined #asahi
<millenialhacker>
Has anyone tried -march=armv8.4-a in Gentoo ? I'm rebuild my whole system with these new CFLAGS but wanted to be sure if anyone has already done to tell me if it's safe.
redlegion_ has joined #asahi
redlegion has quit [Read error: Network is unreachable]
redlegion_ is now known as redlegion
millenialhacker has quit [Ping timeout: 480 seconds]