#uACPI - a portable and easy-to-integrate ACPI implementation
left orbit
1.23M
north holly
@fiery turtle if my spinlock kernel api aren't stubbed, but instead my allocator has no locks when it should, does that score not count?
left orbit
1.48M
north holly
amazing
left orbit
north holly
now I need to fix my fucking spinlocks™️
north holly
<:galaxybrain:701707619651420251>
this is how I feel
left orbit
it just keeps getting better
left orbit
ok i give up trying to run it and get a better score
1506330 ops/s is the final score
left orbit
maybe u should ifdef the minIrql checks lmao
north holly
minIrql is a parameter
north holly
almost
gentle peak
idk how much of what you're doing is necessary since im not familiar with the design of your kernel
so the only advice I can give is using acq rel memory order instead of seq cst
north holly
so the only advice I can give is using acq rel memory order instead of seq cst
that tanks the performance
nvm
gentle peak
also I seriously doubt spin locks are the bottleneck
torpid root
north holly
also I seriously doubt spin locks are the bottleneck
when I stopped locking in my allocator, the score went up by many
numbers
fiery turtle
<@353593948523986945> if my spinlock kernel api aren't stubbed, but instead my a...
direct or indirect stubbing is whats forbidden
idk if this falls under that
kinda does
frank canopy
@fiery turtle btw, does early table access use synchronization calls? i need early table access done so i can set up prcbs for synchronization
also if thats arguably wrong then my score is debatable because while my old PMM is synchronized iirc, the zig gpa is being given no mutex type atm
slim panther
north holly
<@353593948523986945> btw, does early table access use synchronization calls? i ...
no
frank canopy
perfect
fiery turtle
<@353593948523986945> btw, does early table access use synchronization calls? i ...
only map/unmap & log
frank canopy
cool
ill just have to figure out how to swap out my map/unmap once the MM is initialized lol
fiery turtle
u can check the current uacpi init level to decide if u want to call early map or normal map
frank canopy
oh yeah tru
i was gonna do my own init level but that also works
before my MM gets initialized my map is an addition for the bootloader hhdm, afterwards its a completely different function lol
but i need the madt to get sync stuff online which the mm needs so whatever
fiery turtle
yeah thats pretty normal
uacpi warns you if there's any unmapped memory that was mapped early
left orbit
uacpi warns you if there's any unmapped memory that was mapped early
wait wdym?
north holly
left orbit
tghdefrnbexF ed
north holly
could someone test this and give me the results
left orbit
of Course
north holly
meh
north holly
well, triple faults then does that
yeah I observed that
left orbit
1.335M
north holly
rebooting has stopped working in obos
left orbit
why not use the 1.5M from earlier
north holly
why not use the 1.5M from earlier
because it wasn't legit
since the allocator had no locks while it should've been locked
north holly
lol
left orbit
tahts even better
left orbit
no worries
left orbit
i use qemu-system-x86_64 -serial stdio -cpu host,migratable=off,+invtsc,tsc-frequency=100000000 -M q35 -accel kvm -cdrom obos.iso
north holly
qemu-system-x86_64 obos.iso -M q35 -m 1G -accel kvm -cpu host -debugcon file:/dev/stdout
I initially sent
that
left orbit
might be missing 1 zero in the tsc frequency
but it doesn't affect anything
the vm might just run a bit slower
north holly
mainly because I don't have TSC
north holly
the vm might just run a bit slower
so my score becomes slower 
smh I have been boycotted
north holly
🤯
calm latch
tahts even better
What CPU is that?
left orbit
just dont tell anyone that they can add that extra zero to the cmdline 
left orbit
What CPU is that?
13700k
north holly
try rebooting
calm latch
It's like 5x slower on 5900X
Which it shouldn't be
Though I'm testing on Windows in WSL
north holly
AMD skill issue
north holly
Though I'm testing on Windows in WSL
so is he iirc
calm latch
pmOS has gotten 1.8 million on that PC
left orbit
because hyperv is trolling
calm latch
1 million points are missing
On my PC
too much trolling
5900X should be on par with 12700K or something
north holly
OBOS is broken on AMD
left orbit
this is kinda bad because aml is different
north holly
remember
calm latch
Still
left orbit
wont give you the same number as qemu aml
calm latch
Should give something similar
fiery turtle
wait wdym?
if u forget to unref an early access table
left orbit
proxima did 9M in qemu with some optimizations
and on bare metal it did ~1M
just because it runs like 26k aml instructions
instead of 1.7k
it varies wildly between hardware
fiery turtle
its more like, what that aml does
left orbit
qemu mainly stresses allocator + sychronization primitives
fiery turtle
yea
fiery turtle
<@353593948523986945> I updated my score in the pinned message
sure
north holly
credits to marca for helping me optimize spinlocks
man told me like one thing and this happens
he isn't on this server anymore
tho
fiery turtle
What did he tell you
north holly
You could also see if forcing alignment on the atomic_flag to something like 8 bytes helps
#350158983316766730 message
(beware before clicking on the link, it's in hotgarbage)
((come prepared))
(((gl)))
fiery turtle
Yeah I ain't reading that channel
zenith crescent
Yeah I ain't reading that channel
for good reason
north holly
lol
fiery turtle
alright leaderboard update time
north holly
lesgo
north holly
lol
for anyone who doesn't know, think of that channel like breaking every rule of this server at once
left orbit
also please don't use . as the separator in the numbers ☠️
calm latch
❁
north holly
I can't wait for my position in the leaderboard to not change at all because everyone else optimized their stuff too
calm latch
I need to retest pmOS urgently
zenith crescent
I need to get my kernel to a state where I can use uACPI...
north holly
No.
fiery turtle
also please don't use `.` as the separator in the numbers ☠️
y
fiery turtle
https://en.wikipedia.org/wiki/Decimal_separator#:~:text=In English-speaking countries%2C the,%2C)%20placed%20on%20the%20baseline.
left orbit
yeah comma is a more proper separator
fiery turtle
it depends on the country
fiery turtle
i stick to US english in the docs, so . is consistency
zenith crescent
it's not just US english that does. British english also does it
north holly
exactly
calm latch
I've seen people in Spain use '
left orbit
even better
north hull
how is this a surprise
north hull
yes
north holly
and . is decimal separator
fiery turtle
thats the same thing
north hull
lmao
left orbit
or well, 1.500 would be 1+1/2 whereas 1,500 is 1500
median crest
local discord users learning about different languages, life is beautiful
north holly
or well, 1.500 would be 1+1/2 whereas 1,500 is 1500
yeah
I'm in canada, so idk how that differs from US english except for the fact that canada wants to try and be different
zenith crescent
or well, 1.500 would be 1+1/2 whereas 1,500 is 1500
yes
north hull
I'm Finnish and it always trips me up that we reverse them
calm latch
zenith crescent
I'm in canada, so idk how that differs from US english except for the fact that ...
canada is sort of a merge of us and british english isn't it?
north holly
pretty much
fiery turtle
I think oberrow was right
left orbit
1,234,567.89
fiery turtle
thousands separator is ,
fiery turtle
lol sorry maybe u did
north holly
so did I
fiery turtle
. is decimal and , is thousands
north holly
US english
fiery turtle
its the same in europe i think
zenith crescent
I SAID THAT
I think there was a bit of confusion about what separator was being discussed
north holly
in "canadian english" the thousands separator is a space
zenith crescent
US english
not just us english
north holly
but I have never seen any canadian do that
fiery turtle
in "canadian english" the thousands separator is a space
💀
left orbit
I think there was a bit of confusion about what separator was being discussed
true lmao
north hull
its the same in europe i think
eh it depends, in Finland people often reverse them even in English text
calm latch
bruh limine is refusing to boot...
left orbit
no uacpi logs
north holly
same thing in "canadian french"
left orbit
help
north holly
except the decimal separator is ,
median crest
just use scientific notation and fractions
north holly
il sera "100 000,50" dans le francais canadienne
median crest
no fighting about separators
zenith crescent
in "canadian english" the thousands separator is a space
australian english can be either, depending on what region you live in. I normally just use a space
fiery turtle
@hollow elm im updating the leaderboard and managarm is still using the old measurement with old uacpi, do u want to quickly do an update to latest to check if its better? (some people got a 5x speedup from just the update)
left orbit
i got it
hollow elm
<@183510341894930432> im updating the leaderboard and managarm is still using th...
can you remind me later? currently working on something else
left orbit
man pty or smth
median crest
there was a man page explaining TTYs but I forgor what it was
man 3 termios
left orbit
oh there u go
fiery turtle
can you remind me later? currently working on something else
ah alright, ill queue it for whenver the next update happens then
median crest
man 7 pty is for ptys themselves iirc
hollow elm
ah alright, ill queue it for whenver the next update happens then
would also be good to get leo to run the test on his cpu
zenith crescent
man pty or smth
pty is the virtual consoles for in a window manager
fiery turtle
would also be good to get leo to run the test on his cpu <:meme:5754450945896612...
true
median crest
though termios doesnt go in depth about what they are, you probably should read that one article hold on
left orbit
what score have u got queued up for proxima?
median crest
@north holly https://www.linusakesson.net/programming/tty/
fiery turtle
what score have u got queued up for proxima?
6,406,643
left orbit
nice
north holly
I got 2 days and 5 hours to get bash
fiery turtle
some osses have shifted places
median crest
I probably went like down a place or two
fiery turtle
but proxima is still undefeatable
left orbit
i just got 2.7M
north holly
I got 2 days and 5 hours to get bash
-16 hours assuming I go to sleep
fiery turtle
New pmOS score?
if u want to sumbit this instead of 3M then sure 
calm latch
if u want to sumbit this instead of 3M then sure <:LULW:543420553021882371>
When was that?
left orbit
bro actively ignoring me 😭
calm latch
Oh
north holly
damn it
left orbit
lmaoo
north holly
obos hasn't shifted
calm latch
I didn't know it was pmOS
fiery turtle
lmao
left orbit
or well i replied to ur message
calm latch
I was looking at other monitor and wasn't paying attention
left orbit
the logs scroll by so fast because of all the other servers that launch
if u can give me an image with just devicesd that would be nice
maybe i can get a better run
calm latch
It's even doing some dumb calculations in background
fiery turtle
do u have a Intel Core i5-13600K @left orbit ?
calm latch
bro actively ignoring me 😭
Yeah thanks for the score ofc
left orbit
do u have a Intel Core i5-13600K <@796462670995390524> ?
13700k
fiery turtle
an i5?
calm latch
I think it's i7
left orbit
yeah i7 sorry
fiery turtle
when iretq os on the leaderboard
median crest
you're third right now
wtf
median crest
astral phd os confirmed
fiery turtle
nah managarm is just using the old uacpi
north hull
I am gonna implement uACPI this week ❤️
fiery turtle
pmos is currently behind obos but i guess that will change shortly after the final iretq remeasure
calm latch
How much time do I have?
fiery turtle
how much do u need
calm latch
Idk 20 minutes
fiery turtle
thats fine
calm latch
one exit(0) I guess
calm latch
left orbit
fiery turtle
damn
calm latch
It's because it has sse
fiery turtle
astral fell off of third place
calm latch
And my pthread_lock is just a couple of atomic ops
fiery turtle
i did scores for like 4 people on the list
huh, u sure?
i have your cpu recorded only for pmos
fiery turtle
the one oberrow submitted was using Intel i5-4590T
fiery turtle
1.3M
left orbit
tahts even better
this
median crest
astral fell off of third place
I mean yeah
left orbit
i did it on my 13700k :p
fiery turtle
ah ok
left orbit
i also did ilobilix
fiery turtle
This is what I have atm: https://github.com/UltraOS/uACPI/tree/leaderboard-v2?ta...
does this look correct
calm latch
Does this test do raw memory accesses?
fiery turtle
no
fiery turtle
ah right
slim panther
astral fell off
fiery turtle
yeah
calm latch
Hmm this doesn't use lto
calm latch
fiery turtle
@left orbit didnt your os score insane as well?
calm latch
\s
slim panther
i will remove timers
fiery turtle
how do u have so many ops
fiery turtle
no i mean 2418
fiery turtle
do u pass a bunch of weird options to qemu?
calm latch
sudo qemu-system-x86_64 -serial stdio -m 512M -cpu max,+hypervisor,+invtsc,+tsc-deadline -M q35 -accel kvm -cdrom limine/pmOS.iso -smp 4
fiery turtle
whats your qemu verison
calm latch
QEMU emulator version 8.2.93
slim panther
@calm latch by the way you lose a lot of perf my running with smp
left orbit
<@796462670995390524> didnt your os score insane as well?
yeah but it's missing some stuff + not open source
fiery turtle
ah damn
slim panther
[ 0.006877] [ 0] acpi: successfully loaded 1 AML blob, 1732 ops in 0ms (avg 3082947/s)
[ 0.006756] [ 0] acpi: successfully loaded 1 AML blob, 1711 ops in 0ms (avg 3909784/s)
left orbit
and i'm probably gonna rewrite like 90% and implement missing stuff so it might drop a mil or so
fiery turtle
`[ 0.006877] [ 0] acpi: successfully loaded 1 AML blob, 1732 ops in 0ms ...
??
hows is there a different amount of ops
between runs
left orbit
smp?
fiery turtle
ah, its the per-cpu processor objects
fiery turtle
yeah definitely run with -smp 1 if u want the biggest score
slim panther
okay even with -smp 32 i'm still getting around 2.5 mil
zenith crescent
`sudo qemu-system-x86_64 -serial stdio -m 512M -cpu max,+hypervisor,+invtsc,+ts...
why do you run it with sudo?
slim panther
(i only have 16 harts)
fiery turtle
why do you run it with `sudo`?
not in kvm group
left orbit
not in kvm group
slim panther
zenith crescent
ah ok
left orbit
ilobilix Yet another monolithic Linux clone wannabe. Currently under a rewrite
nice description
fiery turtle
do u ship floppies or something
calm latch
<@350748255929499649> by the way you lose a lot of perf my running with smp
Hyper-V memes?
fiery turtle
tape drives
slim panther
there is no image building step
fiery turtle
how do u even boot it
fiery turtle
bruh
calm latch
I just xorriso pmOS
fiery turtle
fair but u should have a way to build it for testing on real hw
fiery turtle
well if u send a tarball with your os and paste the command im sure people can test
fiery turtle
missing out on free points
zenith crescent
whatever cmake release is
that is O3 normally
fiery turtle
the lb score is 3.9M soo
calm latch
Nah my pmm is bad and it's still fast
calm latch
Mine is segregated fit
hallow marten
calm latch
Mine is segregated fit
But it's doing lazy allocations and giving it one page at a time as it pagefaults
On first access
Even though it can give contiguous ranges of any size in O(1) and map that immediately
So like there is room for improvement
slim panther
yea, i'm not getting higher than 3.9m
north holly
I'll continue my optimizations soon™️
fiery turtle
iretq might get a higher score, but w/e we can queue that up for the next update
north holly
rn I got bash to worry about
fiery turtle
you're second anyway and i doubt you'll beat proxima
left orbit
i will one day.
fiery turtle
not happening unless fpu
the currently recorded proxima score is no fpu
median crest
One day I will rewrite astral and all of you will have 10x less score 
slim panther
14m is insane
calm latch
You can do jit in uACPI 2.0 
fiery turtle
that was LTO + O3 + AVX2 + some stuff stubbed
fiery turtle
You can do jit in uACPI 2.0 <:trl:829316641191886888>
yeah i want to try
just generating hardcoded x86 instructions for each aml op might already speed it up by an order of magnitude
although its kinda stupid and a waste of time since it takes like a few ms to load even on most insane firmware
slim panther
If anyone could test and verify score, please do.
qemu-system-x86_64 -machine q35,smm=off -cpu host -accel kvm -name "menix x86_64" -serial stdio -no-reboot -no-shutdown -m 2G -smp 1 -drive format=raw,file=fat:rw:install,if=none,id=fatfs -device virtio-gpu -netdev user,id=net0 -device virtio-net,disable-modern=on,netdev=net0 -device nvme,serial=deadbeef,drive=fatfs -bios /usr/share/ovmf/x64/OVMF.4m.fd
needs ovmf*
median crest
My score is with old uacpi but I douvt itd matter that much
fiery turtle
If anyone could test and verify score, please do.
```
qemu-system-x86_64 -machin...
are u saying it wont boot without ovmf?
fiery turtle
My score is with old uacpi but I douvt itd matter that much
heavily depends on the allocator
median crest
heavily depends on the allocator
Mine is slab so probs not much difference
slim panther
My score is with old uacpi but I douvt itd matter that much
it moved me from 1.5 to almost 4m
with no diff on my side
median crest
I can send an iso ig
qemu-system-x86_64 -cdrom astral.iso -M q35 -enable-kvm -cpu host,migratable=off
slim panther
fiery turtle
3rd place reclaimed (but I already merged the pr so next time ig)
fiery turtle
``qemu-system-x86_64 -cdrom astral.iso -M q35 -enable-kvm -cpu host,migratable=o...
did u do LTO +O3 for this?
median crest
no -O2
slim panther
fiery turtle
but why no o3
slim panther
why u no optimize
slim panther
what
median crest
idrk why
slim panther
what happens
median crest
I could try looking at why lto doesnt work
fiery turtle
solution: disable LTO instead of fixing the bug
median crest
which struct pages
the ones in my kernel
median crest
solution: disable LTO instead of fixing the bug <:meme:575445094589661215>
I never built with lto in the first place soooo
fiery turtle
lesgo
dusky glade
median crest
once I finish this cogmind run I'll see why lto breaks
fiery turtle
astral phd os confirmed
slim panther
zamn
left orbit
astral not beating menix in this lb update 
median crest
soon
slim panther
can you try menix on your cpu
left orbit
sure
slim panther
If anyone could test and verify score, please do.
```
qemu-system-x86_64 -machin...
.
left orbit
entire leaderboard gonna be tested on 13700k soon
slim panther
based
fiery turtle
lmao
left orbit
why zip
slim panther
i'm getting my 265k on monday
left orbit
i will eat u
fiery turtle
thats better than having random cpus
slim panther
sorry
slim panther
fent
fiery turtle
/usr/share/ovmf/x64/OVMF.4m.fd
works on my machine
slim panther
works on my arch
but yea, no non uefi boot
i'm making an efi stub equivalent
for the sake of it
slim panther
what
fiery turtle
bruh
left orbit
unlucky wsl
left orbit
i did like 10 times
fiery turtle
menix is not hyperv friendly
slim panther
woe
left orbit
990K-1.1M
slim panther
that's cursed
that's worse than what i had before
@fiery turtle what cpu you got
maybe it's an intel thing
wait @dusky glade you have a good cpu, right?
dusky glade
yea
fiery turtle
<@353593948523986945> what cpu you got
i9 9900k
left orbit
i got 2M once lmao
slim panther
bruh
left orbit
[ 0.057703] [ 0] acpi: successfully loaded 1 AML blob, 1708 ops in 6ms (avg 255050/s)
new top score for menix
update leaderboard rn
slim panther
💀
left orbit
[ 0.012915] [ 0] acpi: successfully loaded 1 AML blob, 1708 ops in 0ms (avg 2012157/s)
yeah no i cant lol
gentle peak
entire leaderboard gonna be tested on 13700k soon
maybe once i figure out why proxima's mutexes die on the 13700k
fiery turtle
wdym by die
gentle peak
absolutely tank the performance
calm latch
Score invalidated?
gentle peak
on my cpu, normal proxima is 6.4M, but on the 13700k the exact same build is like 5M
if I stub out the locks my cpu gets a lower score than the 13700k
gentle peak
maybe
left orbit
menix does 3.87M normally but it does 1M on my pc
soooo yeah..
also it triple faults on my pc
on a write to the STAR msr
gentle peak
I still think it could be caused by the kvm-on-wsl testing setup
north holly
Under VT-x or AMD-V virtualization, executing PAUSE many times in a short time interval may cause a #VMEXIT. The number of PAUSE executions and interval length that can trigger #VMEXIT are platform-specific.
I think I might know why obos runs a bit slower
on amd
virtualized
it uses the pause instruction for all spinlocks
which causes a vmexit
actually
nvm it also happens on intel
gentle peak
are your locks that contended?
gentle peak
if the lock isn't able to be taken immediately upon calling acquire it's contended
the more threads are trying to acquire the lock at the same time, the more contended it is
north holly
then no locks are contended at all
everything I've been sending have been uniprocessor builds
gentle peak
then pause shouldn't be executed at all either
slim panther
it uses the `pause` instruction for all spinlocks
same
fiery turtle
smh no mwait and c/p states or wahtever
median crest
just dont have locks
north holly
*might cause a vmexit
lilac knoll
why
north holly
because cpus are stoobid
fiery turtle
why pause causes vmexit
same as halt i guess, so that the host can schedule something else instead of wasting cpu time
fiery turtle
depending on how the host handles this vmexit, it could be either very cheap or super expensive
also what else do u do, like literally start spinning like crazy?
that would tank the perf even more
north holly
aren't vmexits in and of themselves expensive
like the cost of one vmexit is very big
even if the actual handler doesn't do much
lilac knoll
also what else do u do, like literally start spinning like crazy?
thas what a spinlock is
fiery turtle
if u spin without relaxing the cpu ur likely hurthing the pair thread, wasting enery, etc etc
slim panther
just dont have locks
dont have smp
fiery turtle
aren't vmexits in and of themselves expensive
if u do pause() on a spinlock its likely contended so you're okay with it "pausing" for N microseconds
lilac knoll
if u spin without relaxing the cpu ur likely hurthing the pair thread, wasting e...
true but if the hypervisor is shitty it might be beneficial to avoid
median crest
#ifndef WILL_RUN_IN_VM
fiery turtle
well theres the hypervisor bit in cpuid
north holly
it's been nearly a year since the last LAI commit
vagrant hull
lai is kill
vagrant hull
@fiery turtle please fix these warnings else streaks will kill me if i try to upstream the port of uACPI to Ironclad lol
/base_dir/ironclad/source/uacpi/interpreter.c: In function 'handle_create_field':
/base_dir/ironclad/source/uacpi/interpreter.c:1740:17: warning: 'field_data.region' may be used uninitialized [-Wmaybe-uninitialized]
1740 | uacpi_shareable_ref(field->bank_region);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/base_dir/ironclad/source/uacpi/interpreter.c:1595:32: note: 'field_data.region' was declared here
1595 | struct field_specific_data field_data;
| ^~~~~~~~~~
/base_dir/ironclad/source/uacpi/interpreter.c:1751:17: warning: 'field_data.field0' may be used uninitialized [-Wmaybe-uninitialized]
1751 | uacpi_shareable_ref(field->index);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/base_dir/ironclad/source/uacpi/interpreter.c:1595:32: note: 'field_data.field0' was declared here
1595 | struct field_specific_data field_data;
| ^~~~~~~~~~
/base_dir/ironclad/source/uacpi/interpreter.c:1754:17: warning: 'field_data.field1' may be used uninitialized [-Wmaybe-uninitialized]
1754 | uacpi_shareable_ref(field->data);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/base_dir/ironclad/source/uacpi/interpreter.c:1595:32: note: 'field_data.field1' was declared here
1595 | struct field_specific_data field_data;
| ^~~~~~~~~~
/base_dir/ironclad/source/uacpi/interpreter.c:1745:35: warning: 'field_data.value' may be used uninitialized [-Wmaybe-uninitialized]
1745 | field->bank_value = field_data.value;
| ~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~
/base_dir/ironclad/source/uacpi/interpreter.c:1595:32: note: 'field_data.value' was declared here
1595 | struct field_specific_data field_data;
|```
strong heath
I indeed will
fiery turtle
<@353593948523986945> please fix these warnings else streaks will kill me if i t...
I've seen them posted 999 times so yeah ill fix them
They're completely bogus tho
flat badge
Wmaybe-uninitialized is quite unreliable
strong heath
If mint finishes porting it, it could be a good addition until I find uACPI but written in Ada ™️
fiery turtle
Ask chatgpt for c to Ada converter
fiery turtle
Wmaybe-uninitialized is quite unreliable
Yeah and what's worse is they change the logic in every gcc version so it doesnt even reproduce on my gcc
flat badge
the simple fix is just adding a memset or similar to every initialization
fiery turtle
I was thinking of something smarter because its asking me to zero a 64 byte struct for no reason
hmm maybe if i update my wsl2 to ubuntu 24
sterile egret
-Wno-maybe-uninitializedthere, fixed
fiery turtle
yeah i wish, but i dont control the flags used by whoever compiles uacpi 
median crest
yeah i wish, but i dont control the flags used by whoever compiles uacpi <:sadca...
__attribute__((initialized)) every single variable
median crest
Apparently it is
fiery turtle
Ada's one is much more accurate
btw hows ada c compat?
vagrant hull
Wmaybe-uninitialized is quite unreliable
well, gcc 14 has it enabled by default sadly
fiery turtle
yeah gcc 11.4 (that comes with ubuntu 22) doesnt produce those warnings even with uninitialized explicitly enabled
daily gcc compile from source session incoming i guess
or maybe i should just run docker
strong heath
btw hows ada c compat?
good
north hull
nixos ftw
fiery turtle
well what does that mean
do u have to make bindings yourself, or are they natively imported
like zig
or
median crest
yeah gcc 11.4 (that comes with ubuntu 22) doesnt produce those warnings even wit...
inb4 -Wno-maybe-unitialized in the spec file
strong heath
gcc can autogenerate bindings for Ada
which I forgot to tell to @vagrant hull
oops
fiery turtle
nixos ftw
its just ubuntu with tools from 20 years before the distro release year
fkn annoying
north hull
its just ubuntu with tools from 20 years before the distro release year
yeah its why I switched to arch
and then eventually to nixos
strong heath
gcc can autogenerate bindings for Ada
gcc -c -fdump-ada-spec -C header.h does it
vagrant hull
bruh
fiery turtle
which I forgot to tell to <@379163868364275712>
💀
vagrant hull
thank god i am lazy and i did not start writing them
strong heath
hell yeah
vagrant hull
lazyness saved me once again
strong heath
clutch
fiery turtle
employee of the month
north hull
So do I gotta implement something else besides the function prototypes in kernel_api.h
fiery turtle
So do I gotta implement something else besides the function prototypes in kernel...
if u want to go fast
fiery turtle
readme has all the headers listed, but platform/libc.h is probably the most important for that
fiery turtle
yeah
fiery turtle
good luck!
north hull
so I noticed like obos doesnt seem to implement uacpi_kernel_pci_device_open and uacpi_kernel_pci_device_close
or am I dumb
fiery turtle
obos is probably on an older uacpi commit?
north hull
ahh
fiery turtle
or you specifically are looking at an old commit
north hull
no thats prolly it im guessing
fiery turtle
bruh are u fucking kidding, ubuntu 24 gcc (GNU 13.3.0) still doenst produce the goddamn uninitialized warnings
median crest
Its like 3 lines of code in that function so its not that vomplicated
median crest
bruh are u fucking kidding, ubuntu 24 gcc (GNU 13.3.0) still doenst produce the ...
Try building it from source
fiery turtle
that was my next step i just didnt really want to do it
let me try clang before i go there
median crest
Just do a static build and stash it in a random server for ci or something
fiery turtle
its kinda hard to ship random binaries into github ci
because u start from scratch every time
mortal yoke
just push the binaries to a throwaway branch 
north hull
so whatsup with this
median crest
I get ptsd from looking at that
fiery turtle
is whats up with this
fiery turtle
I get ptsd from looking at that
north hull
well ig my question is, is it optional?
median crest
Yeah
fiery turtle
i mean its behind an ifdef
north hull
well yeah
fiery turtle
should i add an (OPTIONAL) comment perhpas
fiery turtle
btw, if not sure about some option, there's platform/config.h
north hull
and then impl it
fiery turtle
it explains each option in depth
north hull
ok
fiery turtle
so do I like just define that macro at the top
no no u absolutely dont do that
mortal yoke
so do I like just define that macro at the top
in cflags is probably the best place
fiery turtle
u use your build tool to propagate the define everywhere
north hull
yeap that makes sense
fiery turtle
if u modify kernel_api.h directly u have to ship uacpi's entire source in your kernel
so u cant use a submodule
fiery turtle
its designed so that u dont have to modify uacpi's source directly
north hull
though the way I am building it with my project is rather scuffed
north hull
ah config.h is helpful indeed
median crest
Define it properly, dont do like I did
left orbit
bruh are u fucking kidding, ubuntu 24 gcc (GNU 13.3.0) still doenst produce the ...
haha ubuntu user
how are you liking that llvm 15 and qemu 4?
median crest
Real osdevs use lfs wsl
left orbit
also does ubuntu like abuse snap?
left orbit
Real osdevs use lfs wsl
i actually wanted to do that
for like the longest time
its such a retarded idea i might do it
left orbit
well no its probably not hard but annoying lol
you need linux to build it so thats one thing
fiery turtle
how are you liking that llvm 15 and qemu 4? <:tf:796755156888059954>
lies
im on latest gcc
north hull
@fiery turtle I hate to plague u with questions but should I just ignore UACPI_PRINTF_DECL(2, 3)
fiery turtle
<@353593948523986945> I hate to plague u with questions but should I just ignore...
its a printf attribute
left orbit
you need linux to build it so thats one thing
well, wsl works but ideally youd wanna bootstrap it so you probably dont have a distro yet
north hull
ah
left orbit
im on latest gcc
try adding the offending warning flag yourself?
fiery turtle
try adding the offending warning flag yourself?
nah i mean im compiling it right now
left orbit
ahh
thorny orbit
mortal yoke
https://tenor.com/view/hello-fgteev-duck-cute-duck-ducktales-fgteev-gif-54137896...
I hate those gifs 💀 I have already banned one duck for posting too many of them 
thorny orbit
I hate those gifs 💀 I have already banned one duck for posting too many of them...
who
fiery turtle
tbh i didnt even understand what this was in reference to or what it means
fiery turtle
will i break my apt if i do make install for gcc compiled from source
thorny orbit
will i break my apt if i do make install for gcc compiled from source
why
mortal yoke
if you did it with the default prefix of /usr/local probably not?
thorny orbit
it will rewrite /usr/bin/gcc
fiery turtle
if you did it with the default prefix of /usr/local probably not?
ah so i dont need to do anything?
thorny orbit
or dpkg can return error
mortal yoke
ah so i dont need to do anything?
yeah I don't think, though ig you can run make install as non-root first to make sure that it indeed puts it to there
fiery turtle
well if it bricks my wsl i know who to blame
thorny orbit
for removing gcc i always using this command: sudo rm -rf /boot
left orbit
will i break my apt if i do make install for gcc compiled from source
just install without sudo first
see where it tries to install it
if it doesnt do /usr/local then have fun reconfiguring and rebuilding i guess
but it should
mortal yoke
yeah I did remember right, configure puts it by default in /usr/local
left orbit
yeah even better
i was trying to think if there is a way to tell install to do a dry run
without thinking about telling make to do a dry run
median crest
Just use astral to compile and try gcc out
fiery turtle
somehow compiling via wsl takes longer than compiling via astral
left orbit
i WAS gonna give mintia amd64 port a go but i guess i'm doing lfs wsl now
fiery turtle
lmao
median crest
Once I make my I/O better and rewrite my scheduler it could unironically be kinda fast I think
median crest
i WAS gonna give mintia amd64 port a go but i guess i'm doing lfs wsl now
Keep me updated on how it goes
left orbit
Keep me updated on how it goes
i'll just build lfs systemd and tar it up the import as a wsl distro
hopefully it just works™️
left orbit
idk wasting my time i guess
median crest
Science
median crest
u also need ms specific drivers i think
You can download the kernel source from ms
left orbit
nah the kernel is like separate from the distro
fiery turtle
ah
left orbit
every wsl distro shares the kernel iirc
but i built one anyway
since you can download it as mathew said
fiery turtle
bruh libbacktrace also takes 10 years to compile
fiery turtle
every wsl distro shares the kernel iirc
yeah that makes sense
because its like highly specialized for wsl
left orbit
it's nothing special though, just some weird drivers i guess
but all in all i can skip the kernel and a bootloader
fiery turtle
im assuming they also ship the config that they need for wsl?
with like all stuff disabled they dont care about
left orbit
i found out you can even make wsl spawn a console with earlyprintk
yeah they do
its in the repo
left orbit
with like all stuff disabled they dont care about
also no ☠️
the kernel has like most of the unused shit turned on anyway
i still built like amdgpu and intel gpu drivers for some reason
nouveau too
fiery turtle
i found out you can even make wsl spawn a console with earlyprintk
its over virtio-console right
fiery turtle
i still built like amdgpu and intel gpu drivers for some reason
💀
least lazy ms project
left orbit
its over virtio-console right
no idea, you can put [wsl2] debugConsole = true
in %userprofile%/.wslconfig
and the next time you launch wsl it will spawn a console
dont do it now because you gotta wsl --shutdown but yeah its a thing
fiery turtle
interesting
fiery turtle
❯ lspci -vt
-+-[c5af:00]---00.0 Red Hat, Inc. Virtio console
+-[b8de:00]---00.0 Microsoft Corporation Device 008e
+-[a648:00]---00.0 Microsoft Corporation Device 008e
+-[7cdb:00]---00.0 Red Hat, Inc. Virtio file system
\-[0000:00]-
wsl pci topology is insane
they use pci segments exclusively
left orbit
[ 0.000000] printk: bootconsole [uart8250] enabled interesting
fiery turtle
they use pci segments exclusively
so your average hobby os wont boot
they also use aml
median crest
Omw to make astral bootable on wsl
left orbit
Omw to make astral bootable on wsl
shouldn't be too hard
assuming you make your kernel compatiblke with linux boot proto
left orbit
that is kinda cool
fiery turtle
there's a special VMOD device which is wsl specific
which handles all wsl stuff i guess
and i dont see them exposing PCI topology in any way
so i wonder how linux figures out the pci segments to look for
left orbit
nah
fiery turtle
no idea, you can put```
[wsl2]
debugConsole = true```
in %userprofile%/.wslconf...
this option probably sets the AML variable to 1
median crest
assuming you make your kernel compatiblke with linux boot proto
We need limine for wsl
left orbit
this option probably sets the AML variable to 1
ahhh that is kinda cool
median crest
this option probably sets the AML variable to 1
Wait wtf acpi has DOS device names? 😭
fiery turtle
nah its an iasl bug
its Displayed Device Name
aka how to show it in UI
there's also a virtual TPM device
Name (_STR, Unicode ("Enclave Page Cache 1.0")) then wahtever this is
Hyper_V_Gen_Counter_V1 and this
Name (_HID, "ACPI0012" /* NVDIMM Root Device */) and whatever this is
median crest
its Displayed Device Name
Ohhh
I really need to do acpi stuff in astral but before that I have like 3 other things I need to do
fiery turtle
astral suspend to S3 support 
func fact WSL supports up to 1000 processors
and includes all of them in the aml blob under if statemets
left orbit
wtf lol
left orbit
least scuffed aml
fiery turtle
they were too lazy to generate this at runtime
median crest
Does aml not have a for
fiery turtle
their entire aml is under if statements
fiery turtle
Does aml not have a for
yes but u cant alter these names
like C variable names
if u know what i mean
fiery turtle
yeah u cant
well with safe modifying code u absolutely can
no one is that creative
especially since AML supports dynamically loading other aml code
left orbit
why can't you Name ( "CPU" + Counter, ... )
fiery turtle
TLDR: who knows
they could allow it
atm all names are stored literally
there isnt a StringOp prefix to them
its just literally 4 bytes in a row
left orbit
ah okay that makes more sense
fiery turtle
and that 4 byte literal could mean a lot of things depending on context
could be a field read, could be a method call, could be a ref
depending on context
but they could say e.g. NameOp accepts either a name literal or a TermArg that resolves to a name
thats definitely implementable
and if you're curious
im still compiling gcc
or libstdc++ ig
fiery turtle
i9 9900k
left orbit
hmmm
fiery turtle
its on an nvme too
median crest
At this rate astral will beat wsl
fiery turtle
i really need to reinstall this windows because its very old and has accumulated a ton of crap
left orbit
ur guys' wsl are cursed in a way where u dont get invtsc
and me wsl is cursed in a way where i get 99% variance in uacpi scores
cant win either way
fiery turtle
wsls are cursed no matter where u look
median crest
ur guys' wsl are cursed in a way where u dont get invtsc
Iirc that is like because you have a specific thing for your tsc
I think thats like scaling
And hyperv likes it
fiery turtle
i have no idea what that means
lack of skill issue
median crest
Like if your cpu doesnt support scaling the tsc hyperv just assumes it wont be invariant cuz of vm migrations and doesnt expose the bit
Its annoying because you cant disable this
left orbit
i wonder how fucked would trying to boot limine on wsl be
fiery turtle
u would just stub the linux boot protocol i guess
left orbit
i would love if it was somewhat upstreamable tho
fiery turtle
but then limine wont be able to talk to any of the disk drives
it would need a virtio-fs driver
fiery turtle
there isnt a bios or anything
and the memory map it can figure out from the nvdimm aml device somehow
left orbit
virtio-fs driver for limine wouldn't be too hard maybe
fiery turtle
finally, a uacpi integration for limine
mortal yoke
virtio-fs driver for limine wouldn't be too hard maybe
then its just a matter of time when limine ends up bloated like grub where it has all kinds of drivers starting from usb
fiery turtle
yeah i think:
Device (\_SB.NVDR.N000)
{
Name (_ADR, Zero) // _ADR: Address
Method (_DSM, 4, NotSerialized) // _DSM: Device-Specific Method
{
Return (CDSF (Arg0, Arg1, Arg2, Arg3, _ADR))
}
Method (_LSI, 0, NotSerialized) // _LSI: Label Storage Information
{
Return (LSIM (_ADR))
}
Method (_LSR, 2, NotSerialized) // _LSR: Label Storage Read
{
Return (LSRM (Arg0, Arg1, _ADR))
}
}
u use _LSI and _LSR
and that returns some nvdimm garbage that probably contains the phys address
of where ram exists
fiery turtle
hardcoded like the rest of this info probably
left orbit
this whole thing is only slightly cursed