uACPI - a portable and easy-to-integrate ACPI implementation | Operating System Development | Page 66

frank canopy Mar 12, 2026, 4:48 PM

#

(also not a huge deal because im still a ways away from actually implementing pci interrupt routing in my kernel lmao)

slim panther Mar 12, 2026, 5:43 PM

#

why not take a trait ref

rustic compass Mar 12, 2026, 6:44 PM

#

slim panther why not take a trait ref

thats what i wanted to do

frank canopy Mar 14, 2026, 2:21 AM

#

zuacpi update: added binds for object_get_type, and eval and get for string, buffer, and string_or_buffer

#

@fiery turtle why is it eval_buffer_or_string and get_string_or_buffer

#

why are the two different

fiery turtle Mar 14, 2026, 9:04 AM

#

frank canopy why are the two different

Aml defines them as separate object types

#

A string is guaranteed to be null terminated in uacpi

fiery turtle Mar 14, 2026, 9:42 AM

#

(internally it uses the same object storage ofc)

fiery turtle Mar 14, 2026, 10:27 AM

#

Some feature plans (or ideas) for uacpi:

Ability to precalculate the space needed for the early table buffer
(behind an option) An ability to attach a user object to a namespace node
An extended version of for_each_subtable that allows arbitrary headers via a callback (not just madt/srat entries)
table.h has a lot of naming inconsistencies, so these should be fixed for the next major release
Fix uacpi_generate_absolute_path to no include trailing _, or have a separate function that outputs prettier paths
Fix random scattered globals, at least follow the g_ convention for them
Maybe add helpers for static/compile-time spinlock/mutex creation without allocations
Consider offloading uacpi state into a context
UACPI_DISABLE_PCI
An overridable 64-bit division helper
Check how NT handles missing arguments for methods

calm latch Mar 14, 2026, 11:10 AM

#

Are you planning for for_each_subtable to just deduce the type based on the signature?

#

(so like the api won't even change)

fiery turtle Mar 14, 2026, 11:13 AM

#

thats an interesting idea

#

i could have a for_each_subtable_auto or something

rustic compass Mar 14, 2026, 2:33 PM

#

@fiery turtle given

uacpi_status uacpi_object_resolve_as_aml_namepath(
    uacpi_object*, uacpi_namespace_node *scope, uacpi_namespace_node **out_node
);

do i need to pass a ptr to the uacpi_object that i currently have there or the ptr that i get from uacpi when creating a new a new object?

fiery turtle Mar 14, 2026, 2:33 PM

#

you pass in a string object that you already have

rustic compass Mar 14, 2026, 2:36 PM

#

you mean uacpi_namespace_node?

fiery turtle Mar 14, 2026, 2:39 PM

#

no i mean an object

#

this function takes a string object that the aml provided you

#

which contains a path

#

that path is then resolved against a scope

#

and gives you back the namespace node that it meant

rustic compass Mar 14, 2026, 2:40 PM

#

the problem i have is does the function want its input as call by value or by reference?

fiery turtle Mar 14, 2026, 2:40 PM

#

wdym by that?

#

in C all arguments are by value

#

do you mean which arguments here are output?

#

only the out_node argument

#

even if you wanted, u cant make a uacpi_object manually and pass it by reference (pointer) because its an opaque type

rustic compass Mar 14, 2026, 2:42 PM

#

i can pass a ptr that points to the pointer i got from uacpi that represents the object or i can just directly give the function a copy of that ptr

fiery turtle Mar 14, 2026, 2:42 PM

#

ngl i dont understand the question

#

if you pass it a uacpi_object** it will crash

rustic compass Mar 14, 2026, 2:43 PM

#

//Call by reference
fn call(thing: &MyStruct) 

//Call by value
fn call(thing: MyStruct)

fiery turtle Mar 14, 2026, 2:45 PM

#

the object is not modified by the function

#

if thats what you mean

#

like you literally cannot pass an object by value in uacpi

#

its an opaque type

#

all you have is a handle

#

at all times

frank canopy Mar 14, 2026, 2:47 PM

#

fiery turtle Aml defines them as separate object types

no I know that, why is one buffer or string and the other function string or buffer

#

the order in the name is different between the two functions

fiery turtle Mar 14, 2026, 2:48 PM

#

ohh

#

i guess one more inconsistency to the list

frank canopy Mar 14, 2026, 2:49 PM

#

also stuff like the object type here is why I love zig's opaque type, which just let's me define a thing that's allowed to have a pointer to it but not be dereferenced

fiery turtle Mar 14, 2026, 2:49 PM

#

Added here #1217009725711847465 message

rustic compass Mar 14, 2026, 2:49 PM

#

fiery turtle all you have is a handle

i found my error, also do i have to check that the handle aka the ptr to the object/namespace/.. that i get from uacpi is never null?

fiery turtle Mar 14, 2026, 2:50 PM

#

it's guaranteed to not be null as long as the return status is OK

frank canopy Mar 14, 2026, 2:53 PM

#

oh for the subtable iteration, any plans to support the bonus stuff in the GTDT? or leave that to custom headers

#

doesn't impact me either way because zuacpi can't use c macros and anything that takes a callback is to be avoided if possible because I lose some good zig qol that way (zig uses iterators as the main idiomatic pattern anyway for that)

fiery turtle Mar 14, 2026, 2:54 PM

#

What is gtdt again?

frank canopy Mar 14, 2026, 2:55 PM

#

generic timer description table

#

it's for arm

#

it is in the proper acpi spec though

fiery turtle Mar 14, 2026, 2:55 PM

#

Perhaps could be handled by new for_each_subtable_auto

frank canopy Mar 14, 2026, 2:55 PM

#

the usual chapter for tables

#

the thing with the gtdt is it's got two lists of different subtable headers iirc

#

well slightly different

#

it has one list of subtables but one of the subtable entry types has its own list in it

#

thankfully the sub-sub-entries are fixed size now that im lookin at it again lmao

#

https://uefi.org/specs/ACPI/6.6/05_ACPI_Software_Programming_Model.html#generic-timer-description-table-gtdt

rustic compass Mar 14, 2026, 3:33 PM

#

fiery turtle it's guaranteed to not be null as long as the return status is OK

and if i only get the object as a return and no status?

fiery turtle Mar 14, 2026, 3:34 PM

#

Can you give an example of an api

rustic compass Mar 14, 2026, 3:34 PM

#

uacpi_object *uacpi_object_create_package(uacpi_object_array in);

fiery turtle Mar 14, 2026, 3:34 PM

#

Null is returned on OOM

frank canopy Mar 14, 2026, 4:44 PM

#

oh also, zuacpi technically has api autodocs now

#

insofar as my kernel uses it and the autodocs i generate for the kernel include all the modules it imports 😆

#

https://khitiara.codeberg.page/imaginarium/docs/api/#zuacpi

fiery turtle Mar 14, 2026, 4:47 PM

#

Damn lol

#

Did you make the website yourself?

frank canopy Mar 14, 2026, 4:48 PM

#

yep

#

the autodocs are generated at compile time by zig

fiery turtle Mar 14, 2026, 4:48 PM

#

Thats cool

frank canopy Mar 14, 2026, 4:48 PM

#

note that this is technically a bit out of date because i dont have any ci set up so the website has to be manually updated by me

frank canopy Mar 14, 2026, 4:49 PM

#

fiery turtle Thats cool

so how the autodocs for zig actually work is kinda bonkers tbh. theres a zig module in the source for zig itself that imports the literal actual zig parser and compiles to wasm, and then it has a small js wrapper and loads a tarball of the sources and the wasm module parses the doc comments out of the actual literal source tarball on the fly

#

and then that generates html using the dom api stuff afaik

#

so when you load those autodocs its literally loading a source tarball using the zig std tar api and the actual from-the-compiler zig parser to extract the autodocs

#

https://khitiara.codeberg.page/imaginarium/docs/api/#src/zuacpi/gtdt.zig but that also means it can just pull up the source from the tarball whenever

#

(i did replace the handmade html and css parts of the autodocs to match my site but i left the wasm and js fully intact other than finding and fixing a bug in it that made it break if you build the docs on a platform that uses backslash as the file separator)

fiery turtle Mar 14, 2026, 4:54 PM

#

Damn

frank canopy Mar 14, 2026, 4:54 PM

#

the old-school html purist in me hates this wasm nonsense. the lazy in me is like meh its fine i leave it

#

i dont really think the site is that that impressive but ive been doing osdev for like two and a half years now and webdev for literally like 5-7 times longer than that

#

tfw i realise ive been doing webdev for longer than @north holly has been alive 💀

hallow marten Mar 14, 2026, 4:58 PM

#

fiery turtle Damn

uacpi api reference when

fiery turtle Mar 14, 2026, 4:59 PM

#

Header comments

hallow marten Mar 14, 2026, 4:59 PM

#

maybe but

#

you could use my documentation tool

#

meme

fiery turtle Mar 14, 2026, 4:59 PM

#

What is it

hallow marten Mar 14, 2026, 4:59 PM

#

https://github.com/rdmsr/duck

fiery turtle Mar 14, 2026, 5:00 PM

#

Cool

#

If someone was willing to actually host a website or something that contains the docs that would be cool, but I have no idea how that stuff works and have nowhere to host so

hallow marten Mar 14, 2026, 5:01 PM

#

fiery turtle If someone was willing to actually host a website or something that contains the...

I'm not sure how your comments are written atm, but if it follows the rustdoc/duck convention then I could set it up for you

frank canopy Mar 14, 2026, 5:01 PM

#

fiery turtle If someone was willing to actually host a website or something that contains the...

github pages

hallow marten Mar 14, 2026, 5:01 PM

#

its static pages so github pages works fine

fiery turtle Mar 14, 2026, 5:01 PM

#

hallow marten I'm not sure how your comments are written atm, but if it follows the rustdoc/du...

Nah it doesn't follow any convention as far as im aware

frank canopy Mar 14, 2026, 5:02 PM

#

technically i dont think it needs to be static for gh pages, just client-side only if theres dynamic stuff

fiery turtle Mar 14, 2026, 5:02 PM

#

That could be fixed ofc but thats effort

hallow marten Mar 14, 2026, 5:02 PM

#

well

#

if you'd like I can set it up

fiery turtle Mar 14, 2026, 5:02 PM

#

Will it work without proper convention?

hallow marten Mar 14, 2026, 5:02 PM

#

yeah, or I can adapt the comments

#

theres no convention

#

just

#

/**
 * Something
*/

fiery turtle Mar 14, 2026, 5:03 PM

#

Sure I mean thatd be cool

hallow marten Mar 14, 2026, 5:03 PM

#

two stars

#

or ///

#

but like it doesnt try parsing doxygen-style annotations like @param or whatever

fiery turtle Mar 14, 2026, 5:03 PM

#

It just does stars without the second one atm

hallow marten Mar 14, 2026, 5:40 PM

#

@fiery turtle

#

that's kinda cool

frank canopy Mar 14, 2026, 5:40 PM

#

neat

fiery turtle Mar 14, 2026, 5:40 PM

#

Nice

#

Although Len is not an int I think

frank canopy Mar 14, 2026, 5:42 PM

#

no its a uacpi_size

hallow marten Mar 14, 2026, 5:47 PM

#

yeah that's likely a libclang problem

#

not sure yet

#

ok no I forgot to add -Iinclude

#

now it works

left orbit Mar 14, 2026, 6:11 PM

#

maybe you should make sure it parses without errors

hallow marten Mar 14, 2026, 6:36 PM

#

ok it's all working perfectly now

#

I will try integrating it into the CI

#

do you like that CSS @fiery turtle or do you prefer something more like this: https://rdmsr.github.io/duck/

fiery turtle Mar 14, 2026, 6:48 PM

#

is that just white vs black theme or is something else different also

hallow marten Mar 14, 2026, 6:50 PM

#

the fonts are different

#

and colors

#

obviously

fiery turtle Mar 14, 2026, 6:50 PM

#

i mean idk but i think your website looks better

torpid root Mar 14, 2026, 6:51 PM

#

indeed

hallow marten Mar 14, 2026, 6:51 PM

#

well okay imma set it up with the light theme for now, and I'll let you look at it live and change it if you want

hallow marten Mar 14, 2026, 6:55 PM

#

fiery turtle i mean idk but i think your website looks better

https://rdmsr.github.io/uACPI/

#

lmk if you still prefer the other theme (I personally think that one looks nice, but I can easily change it so its no big deal)

#

obviously some comments are rendered weird, but that's because they were not written with the tool in mind. Now if you want prettier comments, you'd need to write them using markdown syntax

fiery turtle Mar 14, 2026, 7:11 PM

#

Yeah looks good

#

Looks like it indexed internal api as well

hallow marten Mar 14, 2026, 7:12 PM

#

is that a problem?

#

I can exclude it if you want

fiery turtle Mar 14, 2026, 7:12 PM

#

Well other than it being unstable guess not

fiery turtle Mar 14, 2026, 7:12 PM

#

hallow marten I can exclude it if you want

It will have way less noise I suppose

hallow marten Mar 14, 2026, 7:13 PM

#

yeah that's what I figured too

#

I think it's as simple as changing the glob from **/*.h to *.h

fiery turtle Mar 14, 2026, 7:13 PM

#

Yeah

hallow marten Mar 14, 2026, 7:15 PM

#

fiery turtle Yeah

what kinda symbol is internal?

#

just wanna check whether it got included

fiery turtle Mar 14, 2026, 7:16 PM

#

Like it literally listed uacpi_memcmp etc

hallow marten Mar 14, 2026, 7:16 PM

#

yeah ok good

#

I think it's ready now? do you want me to PR?

fiery turtle Mar 14, 2026, 7:17 PM

#

Can osdev wiki stuff be copied there also?

hallow marten Mar 14, 2026, 7:17 PM

#

yeah

#

you can embed mdbook-style docs

fiery turtle Mar 14, 2026, 7:17 PM

#

Would it be a separate repo?

hallow marten Mar 14, 2026, 7:17 PM

#

no

#

a different branch

#

it can be a different repo tho

#

https://github.com/rdmsr/uACPI/commit/4ae753d16481cb9efef3024ff92fe07982c1b96b

GitHub

docs: generate documentation · rdmsr/uACPI@4ae753d

fiery turtle Mar 14, 2026, 7:18 PM

#

I think a separate repo would be better, I dont think a separate branch for literally a different thing is how git is supposed to be used

hallow marten Mar 14, 2026, 7:19 PM

#

you'd have to use your secret token and stuff in the workflow but I think it'd work

fiery turtle Mar 14, 2026, 7:19 PM

#

hallow marten https://github.com/rdmsr/uACPI/commit/4ae753d16481cb9efef3024ff92fe07982c1b96b

A lot of comments changed here are in internal/ also

fiery turtle Mar 14, 2026, 7:20 PM

#

hallow marten you'd have to use your secret token and stuff in the workflow but I think it'd w...

Hmm

hallow marten Mar 14, 2026, 7:20 PM

#

fiery turtle A lot of comments changed here are in internal/ also

does it really matter though? meme

#

ok yeah maybe

fiery turtle Mar 14, 2026, 7:20 PM

#

Less code churn

hallow marten Mar 14, 2026, 7:21 PM

#

I will try restoring them

fiery turtle Mar 14, 2026, 7:21 PM

#

That would be nice

#

Ill think about whether it should be a sep repo, idk yet

#

How do you add new stuff to the website?

#

Like stuff from the wiki etc

hallow marten Mar 14, 2026, 7:22 PM

#

if you have a markdown file you just add it

#

do you?

fiery turtle Mar 14, 2026, 7:22 PM

#

Its all on the wiki so not yet

hallow marten Mar 14, 2026, 7:22 PM

#

ok I reverted the files

#

I'll see if I can get the clanker to rewrite as markdown

fiery turtle Mar 14, 2026, 7:23 PM

#

Nice

#

You can make a pr for now I guess

hallow marten Mar 14, 2026, 7:24 PM

#

I'll do the wiki thing then I'll do it

fiery turtle Mar 14, 2026, 7:24 PM

#

hallow marten a different branch

Why cant it be on master tho?

hallow marten Mar 14, 2026, 7:24 PM

#

because you dont want 8 million HTML files in your master branch

#

ideally

fiery turtle Mar 14, 2026, 7:24 PM

#

You have like 2 there?

hallow marten Mar 14, 2026, 7:25 PM

#

and github pages works really well with "deploy from branch"

hallow marten Mar 14, 2026, 7:25 PM

#

fiery turtle You have like 2 there?

see my gh-pages branch

fiery turtle Mar 14, 2026, 7:25 PM

#

Hm

hallow marten Mar 14, 2026, 7:25 PM

#

the workflow generates HTML and pushes it to that branch

#

github pages publishes it

fiery turtle Mar 14, 2026, 7:25 PM

#

I think other projects use separate repos usually, I can make uacpi a submodule there or something ig?

#

Or it can clone it itself or smth

hallow marten Mar 14, 2026, 7:25 PM

#

yes, you could do that

#

but then youd need to trigger it on every uACPI push?

#

or maybe on every release you regenerate the docs

fiery turtle Mar 14, 2026, 7:26 PM

#

Yeah on release could make more sense

hallow marten Mar 14, 2026, 7:26 PM

#

mhm

#

well, I'll add the wiki stuff and PR for now

fiery turtle Mar 14, 2026, 7:27 PM

#

How hard would it be for you to make a repo where it automatically clones uacpi and then we can just move it into the uacpi org?

hallow marten Mar 14, 2026, 7:27 PM

#

not hard

fiery turtle Mar 14, 2026, 7:27 PM

#

I think that would be ideal

north hull Mar 14, 2026, 7:28 PM

#

hallow marten the workflow generates HTML and pushes it to that branch

is there a reason to push them to git at all

fiery turtle Mar 14, 2026, 7:28 PM

#

And I guess comment changes can be in the main repo ofc

hallow marten Mar 14, 2026, 7:28 PM

#

actually I think I can still PR, but I'd move the workflows to another repo

hallow marten Mar 14, 2026, 7:28 PM

#

north hull is there a reason to push them to git at all

github pages

fiery turtle Mar 14, 2026, 7:29 PM

#

Yeah you pr the comment changes right

north hull Mar 14, 2026, 7:29 PM

#

hallow marten github pages

the html files dont need to be in git for pages to work afaik

#

at least my workflow doesnt do that

hallow marten Mar 14, 2026, 7:29 PM

#

fiery turtle Yeah you pr the comment changes right

and the duck config

#

yeah

fiery turtle Mar 14, 2026, 7:29 PM

#

Sure

hallow marten Mar 14, 2026, 7:30 PM

#

god bless the clanker

#

though I should remove that header and instead name the thing "Example usage"

north hull Mar 14, 2026, 7:31 PM

#

north hull at least my workflow doesnt do that

yeah I just build the docs and push straight to gh pages, the build artifacts never enter git afaict

fiery turtle Mar 14, 2026, 7:31 PM

#

hallow marten though I should remove that header and instead name the thing "Example usage"

Yeah

hallow marten Mar 14, 2026, 7:31 PM

#

north hull yeah I just build the docs and push straight to gh pages, the build artifacts ne...

I thought you had to deploy from a branch

north hull Mar 14, 2026, 7:31 PM

#

oh maybe its this

#

I have it on actions

#

here is this for reference

#

https://github.com/elysium-os/chariot/blob/main/.github/workflows/docs-pages.yml

hallow marten Mar 14, 2026, 7:32 PM

#

ahh

#

ok thanks

#

I think this will be github workflows wrestling more than anything meme

north hull Mar 14, 2026, 7:32 PM

#

🙂

hallow marten Mar 14, 2026, 7:34 PM

#

actually I'm thinking I can move all the config and static stuff to the docs building repo

#

yeah that's better

#

ok @fiery turtle my PR is a lot of back and forths so you should probably squash it, but I think this is better

#

I basically only PR the comments, then you should create a repo named uACPI.github.io that I can be contributor to or something

fiery turtle Mar 14, 2026, 7:40 PM

#

hallow marten I basically only PR the comments, then you should create a repo named uACPI.gith...

You can make the repo yourself and we can move it into the org as well, we did that for zuacpi and uacpi-rs before

hallow marten Mar 14, 2026, 7:41 PM

#

perfect then

fiery turtle Mar 14, 2026, 7:42 PM

#

hallow marten ok <@353593948523986945> my PR is a lot of back and forths so you should probabl...

Can you fixup all commits yourself? Also im not sure we need platform/ comments or struct member comments to be changed, or can it detect those also?

hallow marten Mar 14, 2026, 7:43 PM

#

yeah it can detect struct member comments

fiery turtle Mar 14, 2026, 7:43 PM

#

Cool then

hallow marten Mar 14, 2026, 7:43 PM

#

platform/ is because I did a simple find-and-replace

#

I can revert it

fiery turtle Mar 14, 2026, 7:44 PM

#

Hmm idk, maybe its fine

hallow marten Mar 14, 2026, 7:44 PM

#

well, I just did

#

trl

#

doesn't change anything

#

it's not being picked up anyway

fiery turtle Mar 14, 2026, 7:45 PM

#

Ah ok

#

Yeah cool

#

Can u do a quick interactive rebase to fix them all up into the first commit? The github default squash message is ugly

hallow marten Mar 14, 2026, 7:45 PM

#

fuck how do I do that

#

meme

fiery turtle Mar 14, 2026, 7:46 PM

#

git rebase -i HEAD~N

#

And mark those on top as fixup

#

Then just push -f

hallow marten Mar 14, 2026, 7:47 PM

#

ok I fixed it

#

yea

fiery turtle Mar 14, 2026, 7:49 PM

#

thx

#

i dont think we need all that description in there KEKW

hallow marten Mar 14, 2026, 7:50 PM

#

git did that

fiery turtle Mar 14, 2026, 7:50 PM

#

git commit --amend && git push -f trl

#

btw it seems like u dropped the config also?

hallow marten Mar 14, 2026, 7:51 PM

#

yeah I will put it all inthe new repo

#

I think it's better that way

fiery turtle Mar 14, 2026, 7:52 PM

#

ah cool

#

yes

hallow marten Mar 14, 2026, 7:54 PM

#

ok just vibecoded a workflow

#

I think it should work

fiery turtle Mar 14, 2026, 7:54 PM

#

lol

hallow marten Mar 14, 2026, 7:54 PM

#

git clones duck and uACPI, builds docs from uACPI source and pushes it

fiery turtle Mar 14, 2026, 7:56 PM

#

I wonder what will happen if you ask a clanker to produce an aml interpreter

hallow marten Mar 14, 2026, 7:59 PM

#

ok great it works

#

how do we move it to the org?

#

https://github.com/rdmsr/uACPI.github.io this is the repo

steel inlet Mar 14, 2026, 8:02 PM

#

rdmsr

#

wrmsr(IA32_GS_BASE, 0xdeadbeef)

hallow marten Mar 14, 2026, 8:03 PM

#

https://github.com/wrmsr this is my evil twin

GitHub

wrmsr - Overview

VirtualQueryEx. wrmsr has 241 repositories available. Follow their code on GitHub.

fiery turtle Mar 14, 2026, 8:03 PM

#

hallow marten how do we move it to the org?

@frank canopy do u remember what we had to do to move zuacpi to the org?

hallow marten Mar 14, 2026, 8:03 PM

#

tho you should maybe merge the PR beforehand though

frank canopy Mar 14, 2026, 8:04 PM

#

fiery turtle <@121733282676211714> do u remember what we had to do to move zuacpi to the org?

add the owner user to the org and transfer ownership of the repo to the org in that order

#

iirc

fiery turtle Mar 14, 2026, 8:04 PM

#

hallow marten tho you should maybe merge the PR beforehand though

can u get rid of that commit description real quick?

hallow marten Mar 14, 2026, 8:04 PM

#

sure

fiery turtle Mar 14, 2026, 8:04 PM

#

frank canopy add the owner user to the org and transfer ownership of the repo to the org in t...

ah ok

#

ill send u an invite in a sec and then u can move the repo into the org once you're in it @hallow marten

hallow marten Mar 14, 2026, 8:05 PM

#

alright I changed the desc

fiery turtle Mar 14, 2026, 8:06 PM

#

fiery turtle Mar 14, 2026, 8:06 PM

#

hallow marten alright I changed the desc

thx, ill merge once ci finishes

hallow marten Mar 14, 2026, 8:07 PM

#

you'll have to change the github pages repo settings though

#

in here https://github.com/uACPI/uACPI.github.io/settings

#

I cant do that

#

(I shouldn't be able to anyway)

fiery turtle Mar 14, 2026, 8:07 PM

#

what should be changed?

hallow marten Mar 14, 2026, 8:07 PM

#

ah no I think it's fine

#

I just retriggered anyway

#

https://uacpi.github.io/

#

pretty damn cool

fiery turtle Mar 14, 2026, 8:08 PM

#

how do we retrigger it manually e.g. if a new release happens?

#

damn yeah

#

thats nice

hallow marten Mar 14, 2026, 8:08 PM

#

you just click on "re-run all jobs"

fiery turtle Mar 14, 2026, 8:08 PM

#

i can add that link to the main repo as well

fiery turtle Mar 14, 2026, 8:09 PM

#

hallow marten pretty damn cool

minor nit but hyperlinks in the markdown file dont work 😢

hallow marten Mar 14, 2026, 8:09 PM

#

the relative ones don't

#

which is expected

fiery turtle Mar 14, 2026, 8:09 PM

#

the ones that point to some markdown sections also ig

hallow marten Mar 14, 2026, 8:09 PM

#

really?

#

which one

#

that shouldnt happen

fiery turtle Mar 14, 2026, 8:10 PM

#

hallow marten Mar 14, 2026, 8:10 PM

#

ok that's my bad

fiery turtle Mar 14, 2026, 8:11 PM

#

also it seems we're back to int and comments don't appear there but I guess thats because the pr is not merged yet

hearty pike Mar 14, 2026, 8:11 PM

#

It seems messed up on Firefox.

hallow marten Mar 14, 2026, 8:12 PM

#

mhm well I cant autogenerate them, you have to do something like # More detailed overview {#id} then link to #id

hallow marten Mar 14, 2026, 8:12 PM

#

hearty pike It seems messed up on Firefox.

aaaaa

hallow marten Mar 14, 2026, 8:12 PM

#

fiery turtle also it seems we're back to `int` and comments don't appear there but I guess th...

aaaaa

#

fuck

#

what

fiery turtle Mar 14, 2026, 8:12 PM

#

lmao

hallow marten Mar 14, 2026, 8:12 PM

#

hearty pike It seems messed up on Firefox.

what screen res? This is likely an issue with the table

fiery turtle Mar 14, 2026, 8:13 PM

#

table looks fine for me i guess

hallow marten Mar 14, 2026, 8:13 PM

#

hallow marten what

I know why

#

I forgot -I again

hearty pike Mar 14, 2026, 8:13 PM

#

1080 width by 1920 height.

fiery turtle Mar 14, 2026, 8:13 PM

#

lol yeah i can repro

#

on my monitor thats flipped sideways

hallow marten Mar 14, 2026, 8:14 PM

#

works fine for me

#

mhm

hearty pike Mar 14, 2026, 8:15 PM

#

hearty pike 1080 width by 1920 height.

That is on my vertical monitor.

#

It happens every where.

fiery turtle Mar 14, 2026, 8:15 PM

#

yeah same

hallow marten Mar 14, 2026, 8:15 PM

#

ah

#

not sure how to reproduce

hearty pike Mar 14, 2026, 8:15 PM

#

hallow marten Mar 14, 2026, 8:15 PM

#

@fiery turtle I pushed a fix for the int stuff, should be up soon

fiery turtle Mar 14, 2026, 8:16 PM

#

its reproducable if u slowly resize the page

hallow marten Mar 14, 2026, 8:16 PM

#

I'll check it out

hearty pike Mar 14, 2026, 8:16 PM

#

Yep. It is only an issue for me on the default zoom level on my vertical monitor. Zooming out fixes it and zooming in transitions to a different layout.

fiery turtle Mar 14, 2026, 8:16 PM

#

yep

hallow marten Mar 14, 2026, 8:17 PM

#

#

works now

fiery turtle Mar 14, 2026, 8:17 PM

#

cool thx

hallow marten Mar 14, 2026, 8:17 PM

#

the only issue is aliases sometimes refer to themselves

#

I'm not sure how to fix that

#

I'll push a fix and retrigger eventually

fiery turtle Mar 14, 2026, 8:18 PM

#

nice

#

#

yea...

#

at least im not getting replaced

hallow marten Mar 14, 2026, 8:19 PM

#

btw, tl;dr on how to write comments from now on if you want them to show up better:

use markdown
you can link to names using :: links, so (or the other way around, I always forget markdown syntax)
This is how you document members:

struct hello {
  int a; /**< Some member */
}

There are also mermaid diagrams if you're into that kinda thing

fiery turtle Mar 14, 2026, 8:20 PM

#

what is that?

#

Also what does **< do?

hallow marten Mar 14, 2026, 8:20 PM

#

https://mermaid.js.org/

Mermaid

Create diagrams and visualizations using text and code.

hallow marten Mar 14, 2026, 8:20 PM

#

fiery turtle Also what does **< do?

it's "inline documentation"

#

I think /** should work too

#

mhm

fiery turtle Mar 14, 2026, 8:20 PM

#

ah

#

also i think u changed the pr title but not the actual commit message

hallow marten Mar 14, 2026, 8:21 PM

#

I didnt change the message but removed the description

fiery turtle Mar 14, 2026, 8:22 PM

#

thats way better than docs: generate documentation since its not even what it does meme

hallow marten Mar 14, 2026, 8:22 PM

#

aaa

#

ok

fiery turtle Mar 14, 2026, 8:22 PM

#

thanks

#

ill add the website link to uacpi repo description i think

hallow marten Mar 14, 2026, 8:23 PM

#

done

fiery turtle Mar 14, 2026, 8:23 PM

#

nice

#

When i merge the pr we just regen the docs and it should work?

hallow marten Mar 14, 2026, 8:24 PM

#

yeah

#

the comments just add text

#

so we should see more text show up

fiery turtle Mar 14, 2026, 8:25 PM

#

ye

#

added the link too

hallow marten Mar 14, 2026, 8:26 PM

#

also note that since internal and platform arent used in the generation it wont generate descriptions for like uacpi_u64 and friends

#

that's fine I think

fiery turtle Mar 14, 2026, 8:26 PM

#

yeah thats whatever

#

i think u can deduce from the name what it is

hallow marten Mar 14, 2026, 8:26 PM

#

if you do want to write more extensive documentation, this is how add chapters and stuff:
https://rust-lang.github.io/mdBook/format/summary.html

SUMMARY.md - mdBook Documentation

Create book from markdown files. Like Gitbook but implemented in Rust

#

that is what managarm uses

#

idk why you'd write a book on uACPI though meme

north hull Mar 14, 2026, 8:27 PM

#

mdbook is sick

fiery turtle Mar 14, 2026, 8:27 PM

#

thanks

#

yeah a book idk

#

but some examples would be nice ig

north hull Mar 14, 2026, 8:27 PM

#

well u dont have to call it a book

hallow marten Mar 14, 2026, 8:27 PM

#

north hull mdbook is sick

my thing can generate mdbook chapters and stuff

north hull Mar 14, 2026, 8:27 PM

#

dope

#

readthedocs is also nice

hallow marten Mar 14, 2026, 8:28 PM

#

#

this is what managarm looks like

north hull Mar 14, 2026, 8:28 PM

#

but reST is uhhh.. slightly cursed meme

hallow marten Mar 14, 2026, 8:28 PM

#

uACPI is my test subject before I integrate in managarm btw meme

fiery turtle Mar 14, 2026, 8:28 PM

#

lol

#

isnt there some sort of a managarm book already

hallow marten Mar 14, 2026, 8:29 PM

#

yeah but we wanna integrate the api reference with it

fiery turtle Mar 14, 2026, 8:29 PM

#

ah

hallow marten Mar 14, 2026, 8:30 PM

#

north hull but reST is uhhh.. slightly cursed <:meme:575445094589661215>

yeah markdown is way better and markdown isnt that good

#

btw the css is heavily inspired by kernel.org idk if you noticed trl

north hull Mar 14, 2026, 8:31 PM

#

hallow marten yeah markdown is way better and markdown isnt that good

Yeah I quite like markdown but it doesnt have all the things I would like, but then if u add them u slowly become reST....

hallow marten Mar 14, 2026, 8:31 PM

#

I like org better

#

but not everyone uses emacs

north hull Mar 14, 2026, 8:31 PM

#

never heard of it lol

hallow marten Mar 14, 2026, 8:31 PM

#

asciidoc is also nice

north hull Mar 14, 2026, 8:31 PM

#

yeah

hallow marten Mar 14, 2026, 8:32 PM

#

north hull never heard of it lol

most people use emacs only for org-mode

north hull Mar 14, 2026, 8:32 PM

#

lmao

hallow marten Mar 14, 2026, 8:32 PM

#

like a lot of people consider it its killer feature

#

because its really powerful

north hull Mar 14, 2026, 8:32 PM

#

interesting

hallow marten Mar 14, 2026, 8:33 PM

#

with org-roam you can have a system and stuff like obsidian

frank canopy Mar 14, 2026, 8:33 PM

#

when markdown stops cutting it I just start writing html by hand like the old-fashioned woman I am lmao

fiery turtle Mar 14, 2026, 8:33 PM

#

@hallow marten so whats the right way to write api comments with parameter highlighting etc, is it documented somewhere?

hallow marten Mar 14, 2026, 8:33 PM

#

there's no parameter highlighting, it's literally just markdown

#

and yeah there is

fiery turtle Mar 14, 2026, 8:33 PM

#

no i mean like

hallow marten Mar 14, 2026, 8:33 PM

#

https://github.com/rdmsr/duck/blob/master/USAGE.md

fiery turtle Mar 14, 2026, 8:33 PM

#

oh ok

#

so i just thing for example

hallow marten Mar 14, 2026, 8:33 PM

#

it's like rustdoc if you've ever used that

#

yea

fiery turtle Mar 14, 2026, 8:34 PM

#

nope never used it just like rust in general lol

fiery turtle Mar 14, 2026, 8:35 PM

#

hallow marten https://github.com/rdmsr/duck/blob/master/USAGE.md

so is there any actual difference between /** and /**<?

hallow marten Mar 14, 2026, 8:35 PM

#

I don't think so KEKW

fiery turtle Mar 14, 2026, 8:35 PM

#

lol

hallow marten Mar 14, 2026, 8:35 PM

#

it's literally just a doxygen thing I think

fiery turtle Mar 14, 2026, 8:35 PM

#

it just looks uglier than **

#

windows ci runs for 10 years because it also checks openwatcom

hallow marten Mar 14, 2026, 8:39 PM

#

I was about to say

fiery turtle Mar 14, 2026, 8:40 PM

#

special support for exactly one person that uses openwatcom KEKW

#

at least iretq added multithreaded tests so its way faster than it used to be

#

alright merged

hallow marten Mar 14, 2026, 8:46 PM

#

ok imma retrigger docs

fiery turtle Mar 14, 2026, 8:46 PM

#

cool

hallow marten Mar 14, 2026, 8:47 PM

#

cute duck is the best feature

fiery turtle Mar 14, 2026, 8:47 PM

#

lol

#

would be funnier if it had a chance to randomly rm rf your computer

hallow marten Mar 14, 2026, 8:48 PM

#

https://uacpi.github.io/record.acpi_small_item.html

#

great

#

now there's text

fiery turtle Mar 14, 2026, 8:48 PM

#

that was fast

#

btw clicking index doesnt work sometimes

hallow marten Mar 14, 2026, 8:48 PM

#

CI will be faster eventually when I figure out how to prebuild binaries

hallow marten Mar 14, 2026, 8:49 PM

#

fiery turtle btw clicking index doesnt work sometimes

you're right, I must've forgotten a link somewhere

fiery turtle Mar 14, 2026, 8:49 PM

#

is it possible to improve code formatting here a bit?

hallow marten Mar 14, 2026, 8:49 PM

#

no

fiery turtle Mar 14, 2026, 8:49 PM

#

nooo

hallow marten Mar 14, 2026, 8:49 PM

#

not yet at least

#

because like it's building that on its own and just outputs it

fiery turtle Mar 14, 2026, 8:50 PM

#

i see

hallow marten Mar 14, 2026, 8:50 PM

#

see this https://github.com/rdmsr/duck/blob/master/src/templates/macros.html#L23

fiery turtle Mar 14, 2026, 8:51 PM

#

ah

hallow marten Mar 14, 2026, 8:52 PM

#

mhm see uacpi_namespace_node is broken

#

it refers to itself

fiery turtle Mar 14, 2026, 8:53 PM

#

i mean

#

its an opaque object

hallow marten Mar 14, 2026, 8:53 PM

#

ah yeah you're right

#

but I still saw that bug somewhere where it shouldnt

#

the fix should be simple if alias: dont refer to the same alias

fiery turtle Mar 14, 2026, 8:53 PM

#

https://uacpi.github.io/alias.uacpi_gpe_triggering.html the enum is not clickable also

hallow marten Mar 14, 2026, 8:53 PM

#

it is but it's self referring

#

that's the bug

fiery turtle Mar 14, 2026, 8:54 PM

#

ah

hallow marten Mar 14, 2026, 8:54 PM

#

https://uacpi.github.io/enum.uacpi_gpe_triggering.html

#

this works

fiery turtle Mar 14, 2026, 8:54 PM

#

Yeah i think tahts not the type of the callback KEKW

fiery turtle Mar 14, 2026, 8:54 PM

#

hallow marten https://uacpi.github.io/enum.uacpi_gpe_triggering.html

ah

hallow marten Mar 14, 2026, 8:55 PM

#

fiery turtle Yeah i think tahts not the type of the callback <:KEKW:913708372329644042>

mhm yeah ok that is strange

#

that shouldnt happen

#

I'll take notes of these issues and fix them eventually, if you find any other let me know

fiery turtle Mar 14, 2026, 8:56 PM

#

sure, thanks, overall thats pretty cool anyway and a huge upgrade over just header comments and osdev wiki

hallow marten Mar 14, 2026, 8:56 PM

#

hallow marten mhm yeah ok that is strange

I think what's happening is that it expects only one type, but since it's linking to "multiple" (return type and arguments) it's kinda broken

median crest Mar 14, 2026, 9:54 PM

#

fiery turtle I wonder what will happen if you ask a clanker to produce an aml interpreter

Uacpi test suit might be complete enough to be able to run claude in a loop or something until you have uSlop

fiery turtle Mar 14, 2026, 9:56 PM

#

oh yeah probably, its just gonna take a lot of time

median crest Mar 14, 2026, 10:08 PM

#

If I had a lot of cash that would be an interesting experiment

#

Finance me @anthropic

gentle peak Mar 14, 2026, 11:09 PM

#

fiery turtle

lol it says windows uses acpica?

fiery turtle Mar 14, 2026, 11:11 PM

#

and macos...

gentle peak Mar 14, 2026, 11:12 PM

#

does macOS have a custom one?

fiery turtle Mar 14, 2026, 11:12 PM

#

yeah ofc, i think they have their own impl

gentle peak Mar 14, 2026, 11:12 PM

#

I would've thought they'd just use acpica since they don't have to worry about compatibility with random aml

fiery turtle Mar 14, 2026, 11:13 PM

#

maybe people who are more familiar with macos could confirm

fiery turtle Mar 14, 2026, 11:34 PM

#

#

#

thats kinda hilarious honestly

fiery turtle Mar 14, 2026, 11:37 PM

#

gentle peak does macOS have a custom one?

looking at this panic screen backtrace someone posted it does use acpica

hallow marten Mar 14, 2026, 11:38 PM

#

which AI is that

fiery turtle Mar 14, 2026, 11:38 PM

#

fiery turtle looking at this panic screen backtrace someone posted it does use acpica

thats very surprising to me

fiery turtle Mar 14, 2026, 11:38 PM

#

hallow marten which AI is that

thats gemini 3.1 pro

hallow marten Mar 14, 2026, 11:38 PM

#

huh

#

font is weird

fiery turtle Mar 14, 2026, 11:39 PM

#

i've never seen an apple contribution in acpica code

#

i mean i get that they're doing arm now but

#

but yeah apparently AppleACPIPlatform.kext is based on heavily modified acpica

hallow marten Mar 15, 2026, 12:00 AM

#

@fiery turtle i fixed all the bugs I think

#

the only caveat is that you cant click on types in function pointers

#

thats because they're a PITA

fiery turtle Mar 15, 2026, 12:02 AM

#

Thanks, that was fast

hallow marten Mar 15, 2026, 12:04 AM

#

man complex types are such a pain in the ass to deal with

#

meme

fiery turtle Mar 15, 2026, 12:06 AM

#

💀

fiery turtle Mar 15, 2026, 5:07 PM

#

fiery turtle Some feature plans (or ideas) for uacpi: - Ability to precalculate the space nee...

rustic compass Mar 15, 2026, 9:30 PM

#

i'm nearly done with types.h in what order should i do the other headers for uacpi-rs?

fiery turtle Mar 15, 2026, 9:33 PM

#

i'm assuming u already have tables, uacpi.h contains most initialization functions, after that maybe namespace stuff

frank canopy Mar 15, 2026, 9:40 PM

#

for zuacpi i started with uacpi.h and tables.h, and then from there namespace.h and just adding things as i use them in practice - i figure the longer it takes me to get to using something the less likely someone else using zuacpi will need it immediately

#

so basically same recs as infy has

rustic compass Mar 16, 2026, 2:36 PM

#

does

uacpi_status uacpi_object_assign_package(uacpi_object*, uacpi_object_array in);

copy the supplied objects to somewhere else or does the package object now directly references the objects supplied by uacpi_object_array in?

fiery turtle Mar 16, 2026, 2:38 PM

#

#

there's literally a comment above ☠️

rustic compass Mar 16, 2026, 3:00 PM

#

thinkong

#

maybe i should read those

frank canopy Mar 16, 2026, 3:01 PM

#

maybe

fiery turtle Mar 21, 2026, 8:30 PM

#

letsgo

frank canopy Mar 22, 2026, 2:47 AM

#

any chance of getting the IORT table in uacpi? ill need to create separate definitions for zuacpi anyway but might be good to have em in the base thing too

torpid root Mar 22, 2026, 8:42 AM

#

what does iort do again? I seem to have downloaded documentation for it but I don't remember when or why

fiery turtle Mar 22, 2026, 10:12 AM

#

yeah idk what that is

#

i usually add tables as I go, but I accept PRs for any tables

frank canopy Mar 22, 2026, 3:02 PM

#

IORT (IO redirection table) tells you how to make devices to device ids used by the GIC's MSI stuff

#

and i think some other stuff but thats the bit i need it for

#

it also tells you which devices are connected to which iommus on aarch64

#

because you can have multiple SMMUs connected at once

frank canopy Mar 22, 2026, 3:14 PM

#

fiery turtle i usually add tables as I go, but I accept PRs for any tables

if i ever feel like writing C versions of the tables zuacpi has that uacpi proper doesnt ill PR them then

#

basically doing the same thing with zuacpi of adding tables as i go

hallow marten Mar 22, 2026, 3:15 PM

#

why is zuacpi needed? cant you just @cImport

frank canopy Mar 22, 2026, 3:16 PM

#

hallow marten why is zuacpi needed? cant you just @cImport

well for one cimport is getting removed soon in favor of the TranslateC build step but also zuacpi is written as idiomatic zig (also cimport/translatec cant translate all the macro helpers that uacpi has because its a bindings generator)

#

(and also also zuacpi provides the log alloc and free kernel apis for you on top of that)

#

mostly its to organize and wrap the bindings in a zig-idiomatic way though

#

stuff like error sets and error returns, slices instead of pointer and length, stuff like that

frank canopy Mar 22, 2026, 3:20 PM

#

hallow marten why is zuacpi needed? cant you just @cImport

same difference as uacpi-rs vs just raw bindgen if youre using uacpi in rust too

#

the one is idiomatic wrappers the other is raw bindings

#

tbh the biggest reason for zuacpi existing is that cimport/translate-c used to be way way worse at its job and i needed to fiddle a lot to make things work lmao

frank canopy Mar 22, 2026, 6:38 PM

#

fiery turtle yeah idk what that is

heres an example of it, at least schematically

#

and another example where root complex a isnt hooked to the SMMU but does generate MSIs

#

it can also have reserved memory ranges and at least a few other things

frank canopy Mar 23, 2026, 12:31 AM

#

alright zuacpi has the IORT (a basic version anyway, missing a few things that i dont need yet for my kernel) and also i added the MADT structure for the GIC ITS

fiery turtle Mar 25, 2026, 7:08 PM

#

PSA: uACPI 5.0 will have a distinct return value (UACPI_MAP_FAILED) for uacpi_kernel_map to make 0 a valid return value to allow client code that identity maps all addresses + AML that erroneously tries to map physical 0

vast kestrel Mar 25, 2026, 7:10 PM

#

Yum

fiery turtle Mar 25, 2026, 7:24 PM

#

namely some crappy xiaomi tablet that does that

frank canopy Mar 25, 2026, 7:53 PM

#

oh btw its not official-official yet but https://github.com/tianocore/edk2/issues/11148 exists and is already in acpica

GitHub

[Code First]: GICv5 ACPI support · Issue #11148 · tianocore/edk2

Code First Item Overview Arm has architected a new Generic Interrupt Controller (GICv5). The GICv5 architecture definition can be found here: https://developer.arm.com/documentation/aes0070/latest/...

GitHub

Add GICv5 MADT structures · acpica/acpica@69cca52

The GICv5 adds the following MADT structures:

IRS
ITS Config Frame
ITS Translate Frame

The ACPI spec ECR is at https://github.com/tianocore/edk2/issues/11148

Change-Id: I35a46714de862edd6d0...

fiery turtle Mar 25, 2026, 7:54 PM

#

cool

frank canopy Mar 25, 2026, 7:55 PM

#

not sure if you want to wait for it to actually get into the spec or not but figured it was worth bringing up

fiery turtle Mar 25, 2026, 7:55 PM

#

🤷‍♂️

frank canopy Mar 25, 2026, 7:56 PM

#

yeah thats my feeling too

calm latch Mar 25, 2026, 10:43 PM

#

this is an interesting edge case...

#

but who id maps their memory?

fiery turtle Mar 25, 2026, 10:44 PM

#

https://github.com/CSMWrap/CSMWrap

GitHub

GitHub - CSMWrap/CSMWrap: Get PC BIOS back on UEFI only systems.

Get PC BIOS back on UEFI only systems. Contribute to CSMWrap/CSMWrap development by creating an account on GitHub.

flat badge Mar 25, 2026, 10:47 PM

#

frank canopy oh btw its not official-official yet but https://github.com/tianocore/edk2/issue...

i don't know what i expected but it's hilarious that uefi change proposals are apparently submitted as .docx in github issue comments here KEKW

frank canopy Mar 25, 2026, 10:47 PM

#

lmao yeah

calm latch Mar 25, 2026, 11:13 PM

#

thonk

#

I forgot about csmwrap

frank canopy Mar 26, 2026, 12:33 AM

#

ok im getting IRQ resources (not extended irq) that all report irq 29 (0x17) only and i dont think thats possible

#

since the basic irq resource is a 16-bit mask

#

@fiery turtle any ideas how i could track this down?

#

ah found the issue, was on my end

frank canopy Mar 26, 2026, 2:10 AM

#

bug was in zuacpi in fact and a fix has been pushed

fiery turtle Mar 26, 2026, 6:33 AM

#

Cool

slim panther Mar 26, 2026, 7:00 AM

#

flat badge i don't know what i expected but it's hilarious that uefi change proposals are a...

on that microsoft grind

frank canopy Mar 26, 2026, 7:23 AM

#

fiery turtle Cool

was some leftover manual field alignment in the resource types and i didnt realise that setting the alignment of the first field to 8 would force there to be extra padding after all 6 byte fields and as a result the irq list trailing array thing was doing bad pointer math and overrunning into something else

#

the manual field alignment didnt need to be there anyway so i just removed it and that fixed it

vast kestrel Mar 29, 2026, 5:54 PM

#

Apparently in Gemini lake bioses are supposed to set the HW_REDUCED_ACPI bit

#

Also the fact the ucode handles the timer is cursed af

frank canopy Mar 29, 2026, 5:55 PM

#

that is cursed

calm latch Mar 29, 2026, 6:29 PM

#

though weren't there netbooks without acpi pm timer already?

slim panther Mar 29, 2026, 6:29 PM

#

isnt gemini lake that cursed mini nuc thing

fiery turtle Mar 29, 2026, 6:34 PM

#

Gg

fiery turtle Mar 29, 2026, 7:02 PM

#

vast kestrel Apparently in Gemini lake bioses are supposed to set the HW_REDUCED_ACPI bit

it's hilarious that the timer is still emulated by microcode

#

nt must boot

torpid root Mar 29, 2026, 7:05 PM

#

lol

fiery turtle Mar 29, 2026, 7:10 PM

#

@vast kestrel can u send that pdf btw

torpid root Mar 29, 2026, 7:11 PM

#

fiery turtle <@241277477714919427> can u send that pdf btw

https://cdrdv2-public.intel.com/336560/336560-glk-datasheet-volume1_rev005.pdf

fiery turtle Mar 29, 2026, 7:13 PM

#

thanks

vast kestrel Mar 29, 2026, 7:14 PM

#

yeah it's that one

zenith crescent Mar 30, 2026, 12:42 AM

#

slim panther isnt gemini lake that cursed mini nuc thing

Its the low power pentium gold/silver mobile CPUs

#

Released in 2017

#

My laptop has the pentium silver from that generation, and its very slow. Base clock is 1.1GHz, boost is 2.7GHz, 4 cores. Also only passive cooling.

fiery turtle Mar 30, 2026, 12:45 AM

#

That thing must be torture to work with

zenith crescent Mar 30, 2026, 12:48 AM

#

Yep, it really is. Takes a couple of seconds for it to register the power button has been pressed. Windows takes a long time to boot, but it isn't too bad after that. I once tried gentoo on that thing.... Was not a good idea.

fiery turtle Mar 30, 2026, 12:49 AM

#

I guess it can be good as an osdev machine for testing stuff

zenith crescent Mar 30, 2026, 12:50 AM

#

Yeah

fiery ice Mar 30, 2026, 7:15 AM

#

Optimizations in my memory operations worked wonderfully

#

tho idk why it still says it's slow

#

maybe my timers are screwed up?

loud ice Mar 30, 2026, 9:45 AM

#

slim panther isnt gemini lake that cursed mini nuc thing

and chromebooks too

fiery turtle Apr 5, 2026, 11:11 PM

#

PSA for those 2 people that use UACPI_FORMATTED_LOGGING, uacpi_kernel_vlog is getting removed in 5.0

#

i guess it doesnt even break backwards compat in any way, it just wont be a function exposed in the header

#

since it was never used by uacpi anyway

frank canopy Apr 5, 2026, 11:43 PM

#

formatted logging is where you provide your own printf, right?

#

oh also id like to request replacing your newline at the end of logging with something that can be turned off as a feature (obv on by default)

#

(idea is to replace the final trailing \n with a macro using that stupid c thing of adjacent strings join and then let that macro be defined as nothing if you dont want trailing newline)

fiery turtle Apr 5, 2026, 11:45 PM

#

frank canopy formatted logging is where you provide your own printf, right?

It basically switches logging api into printf mode

fiery turtle Apr 5, 2026, 11:46 PM

#

frank canopy (idea is to replace the final trailing `\n` with a macro using that stupid c thi...

What's the use case for that

frank canopy Apr 5, 2026, 11:46 PM

#

fiery turtle What's the use case for that

zig's entire ecosystem and my own logging infrastructure is all built around never having trailing newlines in log messages

#

and having to use std.mem.trimEnd every time is annoying when it could just be never sent in the first place

fiery turtle Apr 5, 2026, 11:47 PM

#

Ill see what I can do

frank canopy Apr 5, 2026, 11:48 PM

#

👍

#

its not a huge deal itd just be nice to maybe have

#

(the other related idea would be an option to replace line endings altogether for carriage-return ecosystems but i dont have any use for that its just an idea that popped into my head)

fiery turtle Apr 5, 2026, 11:52 PM

#

frank canopy its not a huge deal itd just be nice to maybe have

yeah i think it should be very easy to implement since all logging in uacpi internally is already done via macros, i can just add a UACPI_END_OF_LOG_MSG to config.h something

#

the hard part will be making a script to strip \n from all existing log strings

frank canopy Apr 5, 2026, 11:53 PM

#

you can do that with any good find-in-files editor tbf

fiery turtle Apr 5, 2026, 11:53 PM

#

maybe

fiery turtle Apr 5, 2026, 11:54 PM

#

fiery turtle yeah i think it should be very easy to implement since all logging in uacpi inte...

maybe UACPI_START_OF_LOG_MSG as well so u can add a prefix or something

#

that would be cool

frank canopy Apr 5, 2026, 11:54 PM

#

mhm

frank canopy Apr 5, 2026, 11:56 PM

#

fiery turtle the hard part will be making a script to strip \n from all existing log strings

heres a list of every use of a log function in uacpi

📎 message.txt

fiery turtle Apr 5, 2026, 11:58 PM

#

thats the easy part

hallow marten Apr 6, 2026, 12:00 AM

#

cant u literally just grep for \n"

frank canopy Apr 6, 2026, 12:02 AM

#

i think this is all of them?

📎 remove-trailing.patch

#

if not its most

fiery turtle Apr 6, 2026, 12:02 AM

#

hallow marten cant u literally just grep for \n"

hmm maybe lol

hallow marten Apr 6, 2026, 12:02 AM

#

frank canopy i think this is all of them?

yeah but now replace with UACPI_LINE_ENDING or something

#

its a simple search-and-replace I think

fiery turtle Apr 6, 2026, 12:03 AM

#

hallow marten yeah but now replace with UACPI_LINE_ENDING or something

this patch works

#

all logging code already goes through uacpi_log_lvl

frank canopy Apr 6, 2026, 12:03 AM

#

yeah you can do that in the log helper

#

instead of at each callsite

hallow marten Apr 6, 2026, 12:03 AM

#

idk what that does

#

oh you mean after removing the line endings

#

ah

#

ok

fiery turtle Apr 6, 2026, 12:03 AM

#

yes

frank canopy Apr 6, 2026, 12:04 AM

#

thanks jetbrains for a good find-replace thing for getting me the patch i put above lol

fiery turtle Apr 6, 2026, 12:05 AM

#

git am "C:\Users\infy\Downloads\remove-trailing.patch"
Patch format detection failed.

#

wtf do u want

frank canopy Apr 6, 2026, 12:05 AM

#

lmao

fiery ice Apr 6, 2026, 12:05 AM

#

fiery turtle ``` git am "C:\Users\infy\Downloads\remove-trailing.patch" Patch format detectio...

tell Copilot to manually apply that patch XD

#

it works wonderfully

frank canopy Apr 6, 2026, 12:05 AM

#

fiery turtle ``` git am "C:\Users\infy\Downloads\remove-trailing.patch" Patch format detectio...

may need to dos2unix

#

since im on windows it might have some line ending bullshit

fiery turtle Apr 6, 2026, 12:06 AM

#

doesnt seem to have done anything 😭

frank canopy Apr 6, 2026, 12:06 AM

#

rip

fiery ice Apr 6, 2026, 12:06 AM

#

try what i told you

#

i made it apply 90+ patches bc they did not want to apply bc of the same error

frank canopy Apr 6, 2026, 12:06 AM

#

fiery turtle doesnt seem to have done anything 😭

what aboiut git apply

fiery ice Apr 6, 2026, 12:06 AM

#

and it went and manually applied them

frank canopy Apr 6, 2026, 12:06 AM

#

because i didnt do git format-patch

torpid root Apr 6, 2026, 12:07 AM

#

what about patch

fiery ice Apr 6, 2026, 12:07 AM

#

frank canopy because i didnt do git format-patch

it will prob not work

frank canopy Apr 6, 2026, 12:07 AM

#

so its a raw patch and you might need apply instead of am

fiery turtle Apr 6, 2026, 12:07 AM

#

frank canopy what aboiut git apply

it didnt error out but it also did litearlly nothing lmfao

#

oh nvm

#

it worked

#

cool

frank canopy Apr 6, 2026, 12:07 AM

#

pog

fiery ice Apr 6, 2026, 12:07 AM

#

oh

#

lol

frank canopy Apr 6, 2026, 12:08 AM

#

yeah git am is for taking a format-patch output with authorship and commit messages and all that afaik

fiery ice Apr 6, 2026, 12:08 AM

#

when i did it, both did nothing XD

#

on my other repo

frank canopy Apr 6, 2026, 12:08 AM

#

fiery ice when i did it, both did nothing XD

theres also places where it might have absolute paths in hte patch

fiery ice Apr 6, 2026, 12:08 AM

#

and i had 90 patches to apply

hallow marten Apr 6, 2026, 12:08 AM

#

would patch < file.patch work as well here

frank canopy Apr 6, 2026, 12:08 AM

#

which would break

frank canopy Apr 6, 2026, 12:08 AM

#

hallow marten would `patch < file.patch` work as well here

or patch -i file.patch yeah

torpid root Apr 6, 2026, 12:09 AM

#

fiery turtle PSA for those 2 people that use UACPI_FORMATTED_LOGGING, `uacpi_kernel_vlog` is ...

so this stays?

void uacpi_kernel_log(uacpi_log_level, const uacpi_char*);

fiery turtle Apr 6, 2026, 12:10 AM

#

torpid root so this stays? ```c void uacpi_kernel_log(uacpi_log_level, const uacpi_char*); `...

this part is not related since its not UACPI_FORMATTED_LOGGING

torpid root Apr 6, 2026, 12:11 AM

#

okie

dense steppe Apr 6, 2026, 12:12 AM

#

fiery turtle this patch works

And... now I put #define UACPI_START_OF_LOG_MSG "%s" and everything breaks.

fiery turtle Apr 6, 2026, 12:12 AM

#

be my guest lol

frank canopy Apr 6, 2026, 12:12 AM

#

if you do thats your problem lmao

fiery turtle Apr 6, 2026, 12:12 AM

#

@frank canopy your patch didnt catch stuff like

    uacpi_trace(
        "%sactivated all '%s' opregions controlled by '%.4s', "
        "%zu _REG() calls (%zu errors)\n",

frank canopy Apr 6, 2026, 12:13 AM

#

fiery turtle <@121733282676211714> your patch didnt catch stuff like ```c uacpi_trace( ...

ah yeah not surprised, i just went up to the next quote because i didnt want to overclobber

dense steppe Apr 6, 2026, 12:13 AM

#

There's an argument to instead doing it like printf("%s" fmt "%s", UACPI_START_OF_LOG_MSG, __VA_ARGS__ __VA_OPT__(,) UACPI_END_OF_LOG_MSG);

fiery turtle Apr 6, 2026, 12:14 AM

#

why is this better?

dense steppe Apr 6, 2026, 12:16 AM

#

because it allows me to use %s or literally any string as the "start of log msg" bit.

although there's also an argument against it, because it adds more indirection to logging (now you have to parse the %s whenever you are printing anything).

fiery turtle Apr 6, 2026, 12:16 AM

#

or literally any string as the "start of log msg" bit.
thats already how UACPI_START_OF_LOG_MSG works

#

if you insert printf reserved stuff there for whatever reason thats a u problem

dense steppe Apr 6, 2026, 12:17 AM

#

indeed it is

fiery turtle Apr 6, 2026, 12:17 AM

#

like i just dont see the benefit of adding a ton of runtime cost just to support bad strings

frank canopy Apr 6, 2026, 12:18 AM

#

esp when its not user-provided anyway

fiery turtle Apr 6, 2026, 12:18 AM

#

yeah like its a compile time constant

hallow marten Apr 6, 2026, 12:18 AM

#

fiery turtle like i just dont see the benefit of adding a ton of runtime cost just to support...

i think the solution is to switch to C++ and parse the format strings at compile time

fiery turtle Apr 6, 2026, 12:18 AM

#

if u insert %s there the program wont compile

frank canopy Apr 6, 2026, 12:18 AM

#

hallow marten i think the solution is to switch to C++ and parse the format strings at compile...

switch to zig KEKW

fiery turtle Apr 6, 2026, 12:18 AM

#

since it uses attribute format

hallow marten Apr 6, 2026, 12:18 AM

#

ah that works too

frank canopy Apr 6, 2026, 12:18 AM

#

yeah

hallow marten Apr 6, 2026, 12:19 AM

#

frank canopy switch to zig <:KEKW:913708372329644042>

zuacpi when

#

oh wait

#

meme

frank canopy Apr 6, 2026, 12:19 AM

#

hallow marten zuacpi when

thats literally what the zig uacpi bindings are called lmao

hallow marten Apr 6, 2026, 12:19 AM

#

yes hence the meme

fiery turtle Apr 6, 2026, 12:37 AM

#

@frank canopy I pushed the feature to for-5.0 if u wanna give it a shot

#

let's see if this compiles on OpenWatcom at all

frank canopy Apr 6, 2026, 12:37 AM

#

lol gl

#

ill test in a bit, im in a groove on what im writing atm and dont wanna lose it

fiery turtle Apr 6, 2026, 12:38 AM

#

yeah ofc

#

nice

#

frank canopy Apr 6, 2026, 12:39 AM

#

you gonna put a ifndef default for end of log to \n to maintain compat?

fiery turtle Apr 6, 2026, 12:39 AM

#

i added \n in config.h to maintain compat

#

if config doesnt define it log.h will define to nothing

#

Oh right i should also wrap it in ifndef here so that u can override via gcc and not config.h

frank canopy Apr 6, 2026, 12:41 AM

#

yeah id appreciate that, overriding by the compiler is way easier for zuacpi's build setup

hallow marten Apr 6, 2026, 12:41 AM

#

fiery turtle nice

Do you have something like linux pr_fmt

fiery turtle Apr 6, 2026, 12:42 AM

#

frank canopy yeah id appreciate that, overriding by the compiler is way easier for zuacpi's b...

Ill push it if this last part succeeds trl

fiery turtle Apr 6, 2026, 12:42 AM

#

hallow marten Do you have something like linux pr_fmt

yep

#

i have stuff like this in most files

#

Damn with this patch I can just do

#

Thats really convenient

#

Also good news openwatcom seems to have liked the patch

frank canopy Apr 6, 2026, 12:52 AM

#

🎉 nice

fiery turtle Apr 6, 2026, 12:58 AM

#

thanks for the idea, I think overall its a W for everyone

frank canopy Apr 6, 2026, 1:05 AM

#

glad itll help others

fiery turtle Apr 6, 2026, 1:08 AM

#

I saw a ton of people complaining about it because of latest flanterm changes

frank canopy Apr 6, 2026, 1:18 AM

#

oh? what did flanterm change?

#

(i dont use it so havent been watching)

fiery turtle Apr 6, 2026, 1:19 AM

#

frank canopy oh? what did flanterm change?

It now requires \r\n to do a newline

frank canopy Apr 6, 2026, 1:19 AM

#

ah

#

tbh good, that gives guaranteed compat with serial

fiery turtle Apr 6, 2026, 1:20 AM

#

Yeah

#

The change was valid

frank canopy Apr 6, 2026, 1:20 AM

#

if i could make zig use \r\n for multiline strings i would for that reason too

#

but its way too convenient to use the multiline literals to not use them anyway

#

ok time to test the new stuff

#

alright it worked perfectly right out of the box so zuacpi has been updated and pushed to use the for-5.0 branch for now

fiery turtle Apr 6, 2026, 7:43 AM

#

Cool

narrow steeple Apr 6, 2026, 11:26 AM

#

Hmhm I'm encountering a strange issue. I'm getting a page fault in tables.c:149 from initialize_from_rxsdt at this line uacpi_memcpy(&entry_phys_addr_large, &rxsdt->ptr_bytes[i], entry_size);. The thing is it works if I run with -smp 1 or -smp 4, but with smth like -smp 8 it crashes after FACP to load APIC

UACPI: starting uACPI, version 4.0.0
Map: 0x00000000000f5000-0x00000000000f6000 -> 0xffff8000000f5000-0xffff8000000f6000
UACPI: RSDP 0x00000000000F52C0 00000014 v00 (BOCHS )
Unmap: 0xffff8000000f5000-0xffff8000000f6000
Map: 0x000000000ffe2000-0x000000000ffe3000 -> 0xffff80000ffe2000-0xffff80000ffe3000
UACPI: RSDT 0x000000000FFE259E 00000038 v01 (BOCHS  BXPC    )
Unmap: 0xffff80000ffe2000-0xffff80000ffe3000
Map: 0x000000000ffe2000-0x000000000ffe3000 -> 0xffff80000ffe2000-0xffff80000ffe3000
Map: 0x000000000ffe2000-0x000000000ffe3000 -> 0xffff80000ffe2000-0xffff80000ffe3000
Unmap: 0xffff80000ffe2000-0xffff80000ffe3000
Map: 0x000000000ffe2000-0x000000000ffe3000 -> 0xffff80000ffe2000-0xffff80000ffe3000
Map: 0x000000000ffe0000-0x000000000ffe1000 -> 0xffff80000ffe0000-0xffff80000ffe1000
Unmap: 0xffff80000ffe0000-0xffff80000ffe1000
UACPI: DSDT 0x000000000FFE0040 0000231E v01 (BOCHS  BXPC    )
Unmap: 0xffff80000ffe2000-0xffff80000ffe3000
UACPI: FACP 0x000000000FFE235E 000000F4 v03 (BOCHS  BXPC    )

Page Fault at address 0xffff80000ffe25c6 with error code 0
  - Non-present page
  - Read access
  - Kernel-mode

fiery turtle Apr 6, 2026, 11:35 AM

#

What does your kernel api look like

#

Is it possible u handle this incorrectly?

map1 = map(0x1234);
map2 = map(0x1234);
unmap(map2); // also unmaps map1

narrow steeple Apr 6, 2026, 11:37 AM

#

fiery turtle What does your kernel api look like

void *uacpi_kernel_map(const uacpi_phys_addr addr, const uacpi_size len)
{
    const u64 aligned_address = math::align_down(addr, memory::s_page_size);
    const u64 address_diff = addr - aligned_address;
    const u64 aligned_length = math::align_up(len + address_diff, memory::s_page_size);

    for (usize i = 0; i < aligned_length; i += memory::s_page_size) {
        vmm::map(
            vmm::get_kernel_page_map(),
            aligned_address + i,
            aligned_address + i + boot::get_hhdm_offset(),
            vmm::Attribute::Write);
    }

    return reinterpret_cast<void *>(addr + boot::get_hhdm_offset());
}

void uacpi_kernel_unmap(void *addr, const uacpi_size len)
{
    const u64 virtual_address = reinterpret_cast<u64>(addr);
    const u64 aligned_address = math::align_down(virtual_address, memory::s_page_size);
    const u64 address_diff = virtual_address - aligned_address;
    const u64 aligned_length = math::align_up(len + address_diff, memory::s_page_size);

    for (usize i = 0; i < aligned_length; i += memory::s_page_size) {
        vmm::unmap(vmm::get_kernel_page_map(), aligned_address + i);
    }
}

narrow steeple Apr 6, 2026, 11:37 AM

#

fiery turtle Is it possible u handle this incorrectly? ``` map1 = map(0x1234); map2 = map(0x1...

That could be

fiery turtle Apr 6, 2026, 11:37 AM

#

yeah from your code thats the bug

#

if you do it like u do it you need a reference count

narrow steeple Apr 6, 2026, 11:38 AM

#

Ahh I see, yeah that would make sense why it fails

#

Nice thanks, it works now!

fiery turtle Apr 6, 2026, 12:21 PM

#

npnp

frank canopy Apr 7, 2026, 1:18 AM

#

fuck you acpi, if it wasnt for this one subentry these could all be 4-byte aligned but noooo

frank canopy Apr 7, 2026, 2:17 AM

#

zuacpi updated with SRAT and also a bunch of refactoring to make things more idiomatic which is breaking changes but hopefully wont be too unintuitive to update to if theres anyone other than me using zuacpi

fiery turtle Apr 7, 2026, 8:05 AM

#

frank canopy fuck you acpi, if it wasnt for this one subentry these could all be 4-byte align...

Lol

hasty plinth Apr 7, 2026, 8:31 AM

#

frank canopy fuck you acpi, if it wasnt for this one subentry these could all be 4-byte align...

brought to you by the people behind the xsdt sadmeme

frank canopy Apr 8, 2026, 12:02 AM

#

frank canopy zuacpi updated with SRAT and also a bunch of refactoring to make things more idi...

somehow i forgot to actually push this last night but ive done that now

fiery turtle Apr 13, 2026, 12:11 AM

#

Very good and detailed explanation of how s2idle works
https://docs.kernel.org/arch/x86/amd-debugging.html

kind mantle Apr 13, 2026, 12:26 AM

#

I'm afraid that needs more ACPI knowledge than you think

#

Because it's not explained what actual state the CPUs themselves enter

fiery turtle Apr 13, 2026, 12:28 AM

#

it does say ACPI C3

kind mantle Apr 13, 2026, 12:28 AM

#

yea well

#

my smooth brain has never heard of acpi c3 before

fiery turtle Apr 13, 2026, 12:31 AM

#

kind mantle my smooth brain has never heard of acpi c3 before

https://uefi.org/specs/ACPI/6.5/08_Processor_Configuration_and_Control.html#cst-c-states

kind mantle Apr 13, 2026, 12:36 AM

#

So C3 is basically paused but still powered on

fiery turtle Apr 13, 2026, 12:36 AM

#

nope its not paused

#

cpu is still running

#

just very slowly

kind mantle Apr 13, 2026, 12:37 AM

#

How slow are we talking

#

Several MHz?

fiery turtle Apr 13, 2026, 12:41 AM

#

i think this info is extractable via cpuid

#

but yes

kind mantle Apr 13, 2026, 12:43 AM

#

Can't immediately find where ACPI specifies what rebooting and shutting down are

#

Is it that you place it in S5?

#

And I suppose for reboots, cause something else to immediately wake afterward?

fiery turtle Apr 13, 2026, 12:44 AM

#

kind mantle Is it that you place it in S5?

https://uefi.org/specs/ACPI/6.5/07_Power_and_Performance_Mgmt.html#sx-system-states

#

they even have an assembly snippet

fiery turtle Apr 13, 2026, 12:45 AM

#

kind mantle And I suppose for reboots, cause something else to immediately wake afterward?

no, reboot is just a separate register

kind mantle Apr 13, 2026, 12:45 AM

#

I see

fiery turtle Apr 13, 2026, 12:45 AM

#

but u can reboot via a deferred alarm yes

kind mantle Apr 13, 2026, 12:45 AM

#

So I had correctly identified S5

fiery turtle Apr 13, 2026, 12:45 AM

#

s5 is the full shutdown state

kind mantle Apr 13, 2026, 12:45 AM

#

check

flat badge Apr 13, 2026, 5:36 AM

#

fiery turtle cpu is still running

C3 is not running

#

any non zero C state means that the cpu is not running (i.e. not executing instructions)

#

C3 is clock off, cache coherency not maintained

#

The states in which the CPU keeps running (just at a slower clock rate) are called P states

vast kestrel Apr 13, 2026, 7:14 AM

#

^

fiery turtle Apr 13, 2026, 8:01 AM

#

flat badge The states in which the CPU keeps running (just at a slower clock rate) are call...

Ah right

loud ice Apr 13, 2026, 8:48 AM

#

fiery turtle Very good and detailed explanation of how s2idle works https://docs.kernel.org/a...

On AMD systems, it’s not possible to simultaneously support suspend-to-RAM (S3) and suspend-to-idle (s2idle).
wtf

fiery turtle Apr 13, 2026, 8:57 AM

#

loud ice > On AMD systems, it’s not possible to simultaneously support suspend-to-RAM (S3...

i think they mean it's possible to change at runtime via sysfs, but one mechanism is a direct replacement for the other

frank canopy Apr 19, 2026, 8:56 PM

#

added madt local x2apic and x2apic nmi subentry binds to zuacpi

fierce kiln Apr 24, 2026, 2:10 AM

#

fiery turtle yeah idk what that is

what about DMAR?

#

I was doing iommu stuff and encountered the lack of dmar in uacpi (I might be on old version)

#

there are quite a lot of acpi tables though, and adding all of them could result in undesirable bloat to codebase size

calm latch Apr 24, 2026, 5:37 AM

#

fierce kiln I was doing iommu stuff and encountered the lack of dmar in uacpi (I might be on...

update uACPI

fiery turtle Apr 24, 2026, 8:04 AM

#

fierce kiln there are quite a lot of acpi tables though, and adding all of them could result...

Its just a header with structs, all it affects is nanoseconds to compile time

#

So we can add every existing table if someone wants to do that

calm latch Apr 24, 2026, 8:58 AM

#

No, but I added that table a month ago (which broke msvc in the main branch at the time 💀💀)

#

The AMD table was more obnoxious though

calm latch Apr 24, 2026, 8:59 AM

#

calm latch The AMD table was more obnoxious though

Idk, I can add that too, once I do iommu in my kernel (which I planned on doing after bash I guess?)

pallid lava Apr 27, 2026, 4:22 PM

#

what caching mode is assumed for uacpi_kernel_map?

fiery turtle Apr 27, 2026, 4:25 PM

#

Writeback if referencing normal ram, uncached otherwise

pallid lava Apr 27, 2026, 4:28 PM

#

okay, i suppose i have to check from the memory map

fiery turtle Apr 27, 2026, 4:28 PM

#

Yeah

#

You can also just rely on MTRRs to set the right one for you

#

And just always map wb

pallid lava Apr 27, 2026, 4:32 PM

#

yeah, i didn't think of that

fiery turtle Apr 27, 2026, 10:14 PM

#

I'm finally adding pretty advanced table info iteration api to uacpi

#

also helpers to get/put table by the internal index in the array, as well as the table count

#

the info iteratio api is mostly for more niche uses cases and it provides a ton of additional info:

the origin of the table: whether it came from firmware or client code, whether it has a virtual or physical origin
whether the table was already loaded by the aml interpreter
whether it passed the checksum check
union { physical, virtual } address depending on origin
number of references currently alive

calm latch Apr 27, 2026, 10:17 PM

#

out of curiosity, how do you even know if the table is virtual?

fiery turtle Apr 27, 2026, 10:17 PM

#

virtual is either uacpi_table_install or LoadTableOp/LoadOp from AML

calm latch Apr 27, 2026, 10:18 PM

#

cool

fiery turtle Apr 27, 2026, 10:18 PM

#

These are the origins:

typedef enum uacpi_table_origin {
    /**
     * A table that originated from a physical address provided by the firmware.
     * All tables discovered via RSDP have this origin.
     */
    UACPI_TABLE_ORIGIN_FIRMWARE_PHYSICAL = 1 << 0,

    /**
     * A table that was dynamically loaded by the AML firmware.
     * This includes both Load/LoadTable opcodes. Such tables live in a
     * heap-allocated kernel buffer.
     */
    UACPI_TABLE_ORIGIN_FIRMWARE_VIRTUAL = 1 << 1,

    /**
     * A table installed by the client code via uacpi_table_install_physical().
     */
    UACPI_TABLE_ORIGIN_HOST_PHYSICAL = 1 << 2,

    /**
     * A table installed by the client code via uacpi_table_install().
     */
    UACPI_TABLE_ORIGIN_HOST_VIRTUAL = 1 << 3,
} uacpi_table_origin;

frank canopy Apr 27, 2026, 10:27 PM

#

uacpi_table_install feels like one of those things that almost never comes up that id somehow end up using anyway someday

#

its better than what winload does to add custom acpi tables though dear lord

#

(winload, in order to install extra tables from software, creates its own xsdt from scratch based on the firmware one but with its tables added and also stuff filtered/etc as needed, and then its own rsdp pointing to that that it gives to the kernel and acpi drivers)

fiery turtle Apr 27, 2026, 10:33 PM

#

the most common use case is probably just adding missing firmware devices that some driver looks for

frank canopy Apr 27, 2026, 10:34 PM

#

yeah thats what they do it for iirc

#

im on the wrong machine to actually check for sure

#

i think it was IBFT? could be misremembering but im sure the first letter of the signature was I

fiery turtle Apr 28, 2026, 12:07 AM

#

I think this will be the final API:

typedef struct uacpi_table_info {
    /**
     * The index of this table in the internal array
     */
    uacpi_size idx;

    /**
     * Size of this table in bytes
     */
    uacpi_size size;

    /**
     * Signature of this table
     */
    uacpi_char signature[4];

    /**
     * One of 'uacpi_table_origin' values
     */
    uacpi_u8 origin;

/**
 * This table has been processed & loaded by the AML interpreter.
 * Only applicable for AML bytecode tables.
 */
#define UACPI_TABLE_LOADED (1 << 0)

/**
 * This table's checksum has been checked. The checksum is valid if
 * UACPI_TABLE_CSUM_BAD is not set.
 */
#define UACPI_TABLE_CSUM_CHECKED (1 << 1)

/**
 * This table's checksum was found to be incorrect.
 * Note that if UACPI_FLAG_PROACTIVE_TBL_CSUM is enabled alongside
 * UACPI_FLAG_BAD_CSUM_FATAL, such tables are never installed in the first
 * place.
 */
#define UACPI_TABLE_CSUM_BAD (1 << 2)

    /**
     * A mask of UACPI_TABLE_* flags
     */
    uacpi_u8 flags;

    /**
     * Reference count of this table
     */
    uacpi_u16 reference_count;
} uacpi_table_info;

typedef uacpi_iteration_decision (*uacpi_table_iteration_callback)
    (uacpi_handle, uacpi_table_info *info);

/**
 * Iterate every installed table on the system.
 *
 * The provided callback receives the information about each table in the
 * 'info' pointer. A table may be mapped by client code if needed via
 * uacpi_table_get_by_index().
 *
 * Note that this is a low level helper that iterates _every_ installed table,
 * even tables that are unreachable via uacpi_table_find() etc. due to bad
 * checksum.
 */
uacpi_status uacpi_for_each_table(
    uacpi_table_iteration_callback, void *user
);

/*
 * Retrieve information about the table installed internally at 'idx'.
 *
 * The number of available tables can be retrieved via uacpi_table_count().
 */
uacpi_status uacpi_table_info_get_by_index(
    uacpi_size idx, uacpi_table_info *out_info
);

#

Oh yeah I forgot about the phys address in the info struct

#

union {
    /**
     * The physical address of this table, only applicable for
     * UACPI_TABLE_ORIGIN_*_PHYSICAL.
     */
    uacpi_phys_addr phys_addr;

    /**
     * The virtual address of this table, only applicable for
     * UACPI_TABLE_ORIGIN_*_VIRTUAL.
     *
     * NOTE: use uacpi_table_get_by_index() if you need a virtual address
     *       for a UACPI_TABLE_ORIGIN_*_PHYSICAL table in order to map it.
     */
    void *virt_addr;
};

#

Bro are u fucking kidding me

#

there's an ACPICA compiler regression in macos-latest

#

it cant compile some of my tests anymore

#

#

https://github.com/acpica/acpica/issues/1154

GitHub

Internal compiler error after latest iASL release · Issue #1154 ·...

// Name: LoadTable scoping rules are correct // Expect: int => 0 DefinitionBlock ("", "DSDT", 2, "uTEST", "TESTTABL", 0xF0F0F0F0) { External(SSDT, IntObj)...

calm latch Apr 28, 2026, 8:31 AM

#

uACPI aml compiler when trl

fiery turtle Apr 28, 2026, 8:52 AM

#

naah

calm latch Apr 28, 2026, 11:46 AM

#

Is it fine if I PR adding WAET and SPCR tables to uACPI?
https://learn.microsoft.com/en-us/windows-hardware/drivers/bringup/serial-port-console-redirection-table
http://download.microsoft.com/download/7/E/7/7E7662CF-CBEA-470B-A97E-CE7CE0D98DC2/WAET.docx

Serial Port Console Redirection Table (SPCR) - Windows drivers

The Serial Port Console Redirection Table that is used to indicate whether a serial port or a non-legacy UART interface is available for use with Microsoft® Windows® Emergency Management Services (EMS).

fiery turtle Apr 28, 2026, 11:53 AM

#

calm latch Is it fine if I PR adding WAET and SPCR tables to uACPI? https://learn.microsoft...

yeah ofc

calm latch Apr 28, 2026, 11:54 AM

#

Microsoft talks about patents something something

fiery turtle Apr 28, 2026, 11:54 AM

#

patent for a C struct?

calm latch Apr 28, 2026, 11:55 AM

#

Who knows

flat badge Apr 28, 2026, 11:56 AM

#

spcr is already in uacpi

calm latch Apr 28, 2026, 11:57 AM

#

Yeah, I've just noticed, but there is no ACPI_SPCR_SIGNATURE

fiery turtle Apr 28, 2026, 12:07 PM

#

calm latch Yeah, I've just noticed, but there is no ACPI_SPCR_SIGNATURE

feel free to add

#

@calm latch thanks, the patches look good to me, i'll have to unfuck the acpica regression before i can merge them tho

calm latch Apr 28, 2026, 12:17 PM

#

ok

fiery turtle Apr 28, 2026, 12:18 PM

#

ill check how to install a specific version of a package for brew

#

(hoping it's not yet in apt & choco)

hallow marten Apr 28, 2026, 1:40 PM

#

fiery turtle naah

i mean

#

how hard could it be

fiery turtle Apr 28, 2026, 1:42 PM

#

give it a shot meme

hallow marten Apr 28, 2026, 1:42 PM

#

sponsor me

#

and i will

#

trl

fiery turtle Apr 28, 2026, 1:43 PM

#

nooo free labor when

hallow marten Apr 28, 2026, 1:43 PM

#

I will make every tool uACPI requires until uACPI is completely dependent on me...

#

the documentation was just the start

fiery turtle Apr 28, 2026, 1:43 PM

#

damn

#

but yeah not depending on the acpica compiler would be cool

#

because the maintainer there just accepts random bs and they have 0 regression testing

#

they dont even have CI

hallow marten Apr 28, 2026, 1:51 PM

#

if i had more time id do it

#

in rust or something since parsing is easier with its types

pallid lava Apr 28, 2026, 3:45 PM

#

about uacpi_kernel_install_interrupt_handler,

what polarity and trigger is expected for the irq?
dumb question, but it's safe to assume it's a GSI, right? (not a PIC IRQ that possibly needs to be translated to an IOAPIC IRQ using the overrides)

fiery ice Apr 28, 2026, 3:48 PM

#

pallid lava about uacpi_kernel_install_interrupt_handler, 1. what polarity and trigger is ex...

wdym polarity?

pallid lava Apr 28, 2026, 3:49 PM

#

active high/low

fiery ice Apr 28, 2026, 3:49 PM

#

why does this matter here?

pallid lava Apr 28, 2026, 3:51 PM

#

because the interrupt controller cares about how the irq is going to be asserted

fiery ice Apr 28, 2026, 3:52 PM

#

i always left it default, i didn't even know it could do that XD

pallid lava Apr 28, 2026, 3:52 PM

#

if uacpi_kernel_install_interrupt_handler is only used to install the SCI then it would be a good idea if it is specified in the docstring above that it should be configured as an active low level irq

frank canopy Apr 28, 2026, 4:02 PM

#

pallid lava if uacpi_kernel_install_interrupt_handler is only used to install the SCI then i...

I think you're supposed to use the interrupt source override structures in the madt as well, but yeah it should specify the defaults in the docstring probs

pallid lava Apr 28, 2026, 4:03 PM

#

yeah, i just checked the acpi documentation and i have to check the overrides

#

pain

#

#

for future reference

fiery turtle Apr 28, 2026, 4:09 PM

#

yep

#

SCI is just a normal ISA interrupt

#

but has different defaults for polarity/triggering

fiery turtle Apr 28, 2026, 4:09 PM

#

pallid lava yeah, i just checked the acpi documentation and i have to check the overrides

yes

calm latch Apr 28, 2026, 4:24 PM

#

pallid lava about uacpi_kernel_install_interrupt_handler, 1. what polarity and trigger is ex...

It's an ISA interrupt iirc

pallid lava Apr 28, 2026, 6:47 PM

#

impressive how on the leaderboard the top scores are 10M ops/s

#

meanwhile i have 100k/s 😆

flat badge Apr 28, 2026, 6:52 PM

#

keep in mind that the top of the leaderboard does not support multi threading though

torpid root Apr 28, 2026, 6:52 PM

#

mine does

#

it was like 7.5 mil iirc

#

memcpy and allocator speeds matter the most

fiery turtle Apr 28, 2026, 6:56 PM

#

iirc for loop byte copy memcpy won over PhD memcpy with LTO and O3

#

Aka the uacpi built-in

fiery turtle Apr 28, 2026, 6:57 PM

#

flat badge keep in mind that the top of the leaderboard does not support multi threading th...

It also used SIMD

#

Since its in Userspace

torpid root Apr 28, 2026, 6:58 PM

#

fiery turtle iirc for loop byte copy memcpy won over PhD memcpy with LTO and O3

memcpy calls probably got optimised out completely for smaller copies

flat badge Apr 28, 2026, 6:59 PM

#

fiery turtle iirc for loop byte copy memcpy won over PhD memcpy with LTO and O3

it (= bytewise copy) did win for uacpi on menix but it was way worse in the actual memcpy benchmark

#

because uacpi copies are probably so small that the loop unrolling is not worth it

fiery turtle Apr 28, 2026, 7:00 PM

#

Probably

fiery ice Apr 28, 2026, 8:07 PM

#

pallid lava meanwhile i have 100k/s 😆

fiery turtle Apr 28, 2026, 8:12 PM

#

the record with kvm was like 250 ops/s i think

#

nyaux had it

calm latch Apr 28, 2026, 8:26 PM

#

It's kinda sad there probably won't be a Nyaux moment again with all of the gen ai stuff

pallid lava Apr 28, 2026, 8:50 PM

#

i realized why it wasn't being so quick, i was compiling with Og rather than, for example, O3

#

however now i have random ub all over the place 😢

#

hmm, maybe i added the Og flag later and forgot about it, with it i have 600k ops/s

#

now i have to fix the issues with O3

#

still, uacpi is pretty nice so thanks to all who contributed to its development

#

i cant wait to call random aml methods of my laptop and get random data sync flood errors

fiery turtle Apr 28, 2026, 9:03 PM

#

thanks

#

it was basically just me, some people contributed new tables, monkuous contributed openwatcom support and rewrote tests in C

pallid lava Apr 28, 2026, 9:04 PM

#

first disappoinment of the day: my laptop does not have the fixed power button event

fiery turtle Apr 28, 2026, 9:04 PM

#

you're in for a treat yes

#

fixed power button is PC only

#

laptops have 3 different ways of routing the power button

#

via the EC driver + AML Notify
via the AMD GPIO driver + AML Notify
via legacy general purpose event + AML Notify

pallid lava Apr 28, 2026, 9:05 PM

#

amd gpio?

fiery turtle Apr 28, 2026, 9:05 PM

#

@calm latch has an impl in his kernel if u wanna take a look

#

all modern amd hw uses that

#

both ec and amd gpio are relatively easy to implement, and the legacy routing is free

#

you just have to hook a notify on the power button object in aml

pallid lava Apr 28, 2026, 9:08 PM

#

i see

fiery turtle Apr 28, 2026, 9:08 PM

#

legacy routing is pretty rare tbh lol, i've only seen one laptop like that, it's some toshiba that monkuous has

#

most are just EC

pallid lava Apr 28, 2026, 9:09 PM

#

i had a 2016 laptop which supported it

calm latch Apr 28, 2026, 9:09 PM

#

fiery turtle legacy routing is pretty rare tbh lol, i've only seen one laptop like that, it's...

Desktops use it?

fiery turtle Apr 28, 2026, 9:09 PM

#

desktops are usually fixed event

loud ice Apr 28, 2026, 9:10 PM

#

fiery turtle - via the EC driver + AML Notify - via the AMD GPIO driver + AML Notify - via le...

so basically just hardcode whatever ec your laptop has its fine trl

calm latch Apr 28, 2026, 9:11 PM

#

~~clanker, implement the power button event on my laptop~~

fiery turtle Apr 28, 2026, 9:11 PM

#

hardcode whatever register the firmware uses internally to check a pending pwrbtn press

loud ice Apr 28, 2026, 9:11 PM

#

cros_ec isnt that hard anyway

#

and you can claim that implementing that is a "real" driver

fiery turtle Apr 28, 2026, 9:15 PM

#

For anyone curious this is how general purpose event dispatch loop looks like on monkuous' toshiba laptop:

While ((GP50 || (GP51 || (GP52 || (GP53 || (GP54 || (
    GP55 || (GP56 || (GP57 || (GP60 || (GP61 || (GP62 || (GP63 || 
    (GP66 || (GP67 || (GP70 || (GP71 || (GP93 || (GP94 || (
    GPAP || (GPBP || (GPCP || (GPDP || (GPEP || (GPFP || (GPGP || 
    (GPHP || (FCDP || (FTVP || (FQSP || (GP96 || (GP97 || (
    GPD0 || (GPD1 || (GPD2 || (GPD3 || (GPD4 || (BPFE || (TPBP || 
    (ECDS || (B1ST || B2ST)))))))))))))))))))))))))))))))))))))))))

pallid lava Apr 28, 2026, 9:16 PM

#

jesus

fiery turtle Apr 28, 2026, 9:16 PM

#

All of these are recursive as well

frank canopy Apr 28, 2026, 9:16 PM

#

lmao of course

fiery turtle Apr 28, 2026, 9:16 PM

#

So it's OrOp(GP50, OrOp(...

pallid lava Apr 28, 2026, 9:16 PM

#

the "worst" I've seen is

return 0xFFFFFFFE
return 0xFFFFFFFE

#

and my laptop has like 5 or 6 objects that are missing

#

acpica complains too

fiery turtle Apr 28, 2026, 9:17 PM

#

pallid lava the "worst" I've seen is return 0xFFFFFFFE return 0xFFFFFFFE

this is just crappy fw patching

calm latch Apr 28, 2026, 9:27 PM

#

fiery turtle For anyone curious this is how general purpose event dispatch loop looks like on...

Wasn't this on oberrow's laptop?

fiery turtle Apr 28, 2026, 9:28 PM

#

nope

pine leaf Apr 29, 2026, 1:57 PM

#

fiery turtle For anyone curious this is how general purpose event dispatch loop looks like on...

oh dear god

fiery turtle Apr 29, 2026, 8:38 PM

#

wait haiku is back??

slim panther Apr 29, 2026, 8:51 PM

#

fiery turtle wait haiku is back??

yo????

fiery turtle Apr 29, 2026, 8:53 PM

#

they updated the PR and it seems to actually work now + using latest uacpi https://review.haiku-os.org/c/haiku/+/8937

#

they're using some very niche apis and in general haiku is way more advanced acpi-usage-wise

#

than any other hobby os i've seen

#

or i guess its not even that hobby at that point since they have funding etc

frank canopy Apr 29, 2026, 8:57 PM

#

yo nice

frank canopy Apr 29, 2026, 8:57 PM

#

fiery turtle they're using some very niche apis and in general haiku is way more advanced acp...

which niche apis out of curiosity?

fiery turtle Apr 29, 2026, 8:57 PM

#

the register accessors for example

#

all the GPE-related APIs also

frank canopy Apr 29, 2026, 8:59 PM

#

ah neat

fiery turtle Apr 29, 2026, 9:02 PM

#

Looks like no one in ACPICA cares about the regression

#

ill have to bisect the damn compiler myself

frank canopy Apr 29, 2026, 9:06 PM

#

rip

fiery turtle Apr 29, 2026, 9:10 PM

#

Bisecting: 61 revisions left to test after this (roughly 6 steps)
[b35adf49e89a610ea9ea33af21c1970ea95f56e4] Add validation for Node in AcpiNsBuildNormalizedPath to prevent use-after-free vulnerabilities

Not too bad

frank canopy Apr 29, 2026, 9:12 PM

#

yeah that aint bad tbh

fiery turtle Apr 29, 2026, 9:23 PM

#

alright

eff423cadbe44c916078d99b666406b20aea130b is the first bad commit
commit eff423cadbe44c916078d99b666406b20aea130b
Author: ikaros <[email protected]>
Date:   Sat Jan 10 20:39:57 2026 +0800

    Add AslKeywordMappingCount for validation in OpcGenerateAmlOpcode

    Fixes: #1070
    Signed-off-by: ikaros <[email protected]>

 source/compiler/aslglobal.h  | 1 +
 source/compiler/aslmap.c     | 2 ++
 source/compiler/aslopcodes.c | 8 ++++++++
 3 files changed, 11 insertions(+)
bisect found first bad commit

#

https://github.com/acpica/acpica/pull/1155

GitHub

Fix a compiler regression by d-tatianin · Pull Request #1155 · ac...

eff423c broke any code using the Printf macro (and possibly other macros):
DefinitionBlock ("", "DSDT", 2, "uTEST", "TESTTABL&quot...

fiery turtle Apr 29, 2026, 10:13 PM

#

Stupid choco and brew always hardcode the newest version and u cant fetch older

#

so I'll disable MacOS CI completely, and for Windows I'll download an older release manually from the intel website in CI directly

#

and linux will keep working since ubuntu is using an old version thankfully

fiery turtle Apr 29, 2026, 10:35 PM

#

@calm latch could u please rebase your pr so it's on top of master? I added a fix for CI

calm latch Apr 29, 2026, 10:36 PM

#

done

fiery turtle Apr 29, 2026, 10:37 PM

#

Awesome

#

Btw why the hell do they ship WAET spec as a word document

frank canopy Apr 29, 2026, 11:05 PM

#

fiery turtle Btw why the hell do they ship WAET spec as a word document

same for DBG2 iirc

#

all the microsoft tables are shipped that way

fiery turtle Apr 29, 2026, 11:09 PM

#

lol

#

the microsoft office gate

frank canopy Apr 30, 2026, 5:24 AM

#

yknow part of me wonders if those table docs all being word docs isnt some sort of licensing policy thing dating back to when pdfs were pay-to-create

crude summit Apr 30, 2026, 6:43 AM

#

fiery turtle the record with kvm was like 250 ops/s i think

all you need is a really really bad allocator to make that number slow

#

pretty sure that uacpi has a pattern of tons of tiny allocations and deallocations during that init process

fiery ice Apr 30, 2026, 6:45 AM

#

i still need to understand why my allocator is slow AF

#

yes

#

it does

crude summit Apr 30, 2026, 6:45 AM

#

it's likely O(n) during free

fiery ice Apr 30, 2026, 6:46 AM

#

mine is a bit faster on a physical linux

#

on WSL it was slower

crude summit Apr 30, 2026, 6:46 AM

#

I have two allocators, a base one that is really really simple and slow as a snail riding an elephant on the back of a tortoise, and then a buddy allocator which is really fast, subsystems allocate buddy allocators inside arenas in the dumb allocator, and uacpi has such a buddy region

#

it is also slow if you're using tcg for obvious reasons

#

here's mine on KVM

#

[0]: [uACPI] successfully loaded 1 AML blob, 1757 ops in 3ms (avg 536526/s)

fiery ice Apr 30, 2026, 6:49 AM

#

10x faster than mine

#

lol

crude summit Apr 30, 2026, 6:50 AM

#

feel free to take a look at my allocator if you want, basically uacpi only really does 3 things during init, requests memory, io space and an interrupt

#

if you give uacpi a bump allocator you can fudge that stat lol

fiery ice Apr 30, 2026, 6:51 AM

#

yeah it allocates very small amounts

#

does it also release that memory after?

#

or some is kept

crude summit Apr 30, 2026, 6:55 AM

#

it keeps some iirc

#

because it sets up interrupt handlers and will need memory to handle that

#

if you stub out it's interrupt registration function in your abstraction layer you can safely tear down uacpi afterwards by taking away it's allocated memory region, I used to do that

fiery ice Apr 30, 2026, 7:10 AM

#

it allocates all of this, lol

#

and then 12 bytes at the end

fiery turtle Apr 30, 2026, 8:04 AM

#

crude summit if you give uacpi a bump allocator you can fudge that stat lol

Thats a common misconception btw

#

Uacpi scored pretty poorly on that test with a bump allocator

#

Since it made the cache cold and as u said uacpi frees and allocates tiny objects all the time

#

An allocator that reuses memory scores a lot more here

fiery turtle Apr 30, 2026, 8:06 AM

#

crude summit if you stub out it's interrupt registration function in your abstraction layer y...

If only there was uacpi_state_reset that cleanly tears everything down lol

fiery turtle Apr 30, 2026, 8:10 AM

#

crude summit it keeps some iirc

But it literally builds an entire aml namespace lol

#

Only temporary objects are freed

#

Whatever was used during DSDT eval

fiery turtle Apr 30, 2026, 9:17 AM

#

maybe i should consider making like an emulated arena allocator in uacpi

#

or like an object cache for fast reuse

crude summit Apr 30, 2026, 11:10 AM

#

fiery ice it allocates all of this, lol

are these memory map requests or io space requests?

#

99% of kernels can just stub out io space requests,.it's asking for access to an io port and most kernels don't restrict that unless you're doing some kind of microkernel thing

fiery ice Apr 30, 2026, 11:44 AM

#

crude summit are these memory map requests or io space requests?

yeah it's io map i tought i left the allocation debug on

crude summit Apr 30, 2026, 11:54 AM

#

fiery ice yeah it's io map i tought i left the allocation debug on

do you have a way to generate profiler dumps e.g. callgrind.out

#

you might want to add that so you can profile your kernel and see where the expensive parts are

kind mantle Apr 30, 2026, 12:00 PM

#

fiery ice yeah it's io map i tought i left the allocation debug on

this is a violently incorrect way to go about mapping I/O

mortal yoke Apr 30, 2026, 12:01 PM

#

on x86 its not

kind mantle Apr 30, 2026, 12:01 PM

#

do you just identity map the whole lower half?

#

smh

mortal yoke Apr 30, 2026, 12:01 PM

#

its for port io

kind mantle Apr 30, 2026, 12:01 PM

#

right

#

I honestly haven't used x86 for so long I forgot it had that

fiery turtle Apr 30, 2026, 12:02 PM

#

yeah lol

#

map_memory is for mapping memory

#

i guess the names could be clearer but they're designed after their respective name in the acpi spec

#

aka SystemIO and SystemMemory

#

well there's a comment there i guess

fiery ice Apr 30, 2026, 12:07 PM

#

crude summit you might want to add that so you can profile your kernel and see where the expe...

i do

crude summit Apr 30, 2026, 1:53 PM

#

I'm betting debug output is the most expensive part of that?

fiery ice Apr 30, 2026, 1:54 PM

#

crude summit I'm betting debug output is the most expensive part of that?

no

#

or well

#

i can't see the print functions

#

they would polute too much

crude summit Apr 30, 2026, 1:55 PM

#

how can you not see them? you exclude them from profiling?

fiery ice Apr 30, 2026, 1:57 PM

#

crude summit how can you not see them? you exclude them from profiling?

yes

hallow marten Apr 30, 2026, 2:01 PM

#

crude summit do you have a way to generate profiler dumps e.g. callgrind.out

How do you do that

fiery ice Apr 30, 2026, 2:06 PM

#

hallow marten How do you do that

📎 message.txt

#

-finstrument-functions

crude summit Apr 30, 2026, 2:09 PM

#

hmmm thats why you cant profile debug output

#

the ouput format isnt flexible enough

#

i recommend outputting it in callgrind format on e.g. com1 or the qemu console

#

then, loading that into kcachegrind

fiery ice Apr 30, 2026, 2:11 PM

#

hmm

#

i'll try

crude summit Apr 30, 2026, 2:11 PM

#

if you want i have code for that

fiery ice Apr 30, 2026, 2:11 PM

#

oh yes pls

crude summit Apr 30, 2026, 2:11 PM

#

its just a simple text format a bit like yaml

#

https://github.com/brainboxdotcc/retro-rocket/blob/master/src/profiler.c

#

you do need to store 'edges'

#

mine stores it to a simple hash table

#

then just walks the hash table when i press a debug-key (ctrl-alt-shift-P)

fiery ice Apr 30, 2026, 2:16 PM

#

thx

#uACPI - a portable and easy-to-integrate ACPI implementation