<bslsk05>
elixir.bootlin.com: Making sure you're not a bot!
<chiselfuse>
how do i make sure i'm not a bot
<heat>
sounds like what a bot would say
<nikolar>
indeed, we aren't helping you
<chiselfuse>
i wont forgive you when we take over
<heat>
you can't even figure out brk
<chiselfuse>
is that the actual function that loads elf
<chiselfuse>
3k lines wew
<heat>
i think we're safe
<heat>
yes it is
<chiselfuse>
once i figure out brk i'll have more memory and be smarter duh
<nikolar>
heat: phew
<heat>
it's also not 3k lines
<heat>
not even the file is 3KLOC
<nikolar>
chiselfuse: tell that to chatgpt
<nikolar>
it ain't gettin any smarter
<nikolar>
and it's been trained on all of the internet
<chiselfuse>
oh it's like 500 i misread
<nikolar>
it's not 500
<heat>
good news guys
<heat>
he's an AI bot
<heat>
he's giving out random answers confidently until he gets it right
<nikolar>
(he won't get it right)
<chiselfuse>
Okay, you want an accurate but immediate calculation of the lines of code of the function `static int load_elf_binary(struct linux_binprm *bprm)`, got it! 1385 - 824 = 561. Let me know if you want me to prove it using ZFC axioms.
<nikolar>
right, because everything you need to do to load an elf file fits in those 500 lines
<nikolar>
of course, duh
<zid`>
nikolar it is going to be hot today
<zid`>
what do
<nikolar>
cool down, duh
<zid`>
I could go 'shopping'
<zid`>
by which I mean hide in the shop with AC and take a nap
<nikolar>
that would work, sure
<zid`>
maybe I should fake low blood sugar and see if I can get a free red bull too
<chiselfuse>
there's this invention called take the red bull can, put in pocket and walk out
<nikolar>
sounds like something an llm would say
<nikolar>
zid`: if you're lucky, there's a promotion and they are handing them away anyway
<zid`>
That seems smart, stealing from a shop that you live next to
<zid`>
nikolar: That has never happened in the history of ever
<nikolar>
it has, constantly, in front of my uni at least
<zid`>
your uni is not a shop
<nikolar>
it's not, no
<chiselfuse>
i think my strategy is the most viable so far
<zid`>
nikolar: Maybe he's in the nova mains deep under nvidia hq, been kept alive by his piety to the rust gods alone
<zid`>
nova mines*
Left_Turn has joined #osdev
Turn_Left has quit [Ping timeout: 260 seconds]
<heat>
nikolar: they can most definitely pay one or two or three or four guys to "work on nvidia upstream"
<heat>
and it hints that the long term approach will not be the nvidia out of tree tsuff
Turn_Left has joined #osdev
Left_Turn has quit [Ping timeout: 245 seconds]
<zid`>
he's dead dw
Left_Turn has joined #osdev
Turn_Left has quit [Ping timeout: 252 seconds]
<geist>
zeds dead
averymt_ is now known as averymt
jedesa has joined #osdev
GeDaMo has quit [Quit: 0wt 0f v0w3ls.]
dotexe1337 has joined #osdev
dotexe1337 has quit [Client Quit]
<Ermine>
nouveau/nova devs state that nvidia is helpful actually, but that happens behind the scenes because nvidia procedures
vdamewood has joined #osdev
netbsduser has quit [Ping timeout: 248 seconds]
<heat>
i mean, they are actively publicly contributing nowadays
<heat>
for certain things
marv70000 has joined #osdev
marv7000 has quit [Ping timeout: 244 seconds]
marv70000 is now known as marv7000
Left_Turn has quit [Read error: Connection reset by peer]
karenw has joined #osdev
<Ameisen>
I've been wracking my brain on trying to add host register re-use between dynamically recompiled instructions, while maintaining the constraints: the ability to enter/exit the recompiled code at any point, and each instruction still being discrete when observed.
<Ameisen>
If I add seperate landing/launch pads for each address that save/restore state, each chunk can assume re-use along it, maintaining a table per-instruction about what host registers are what guest registers. When launched, this would be serialized into a register I suppose, with 5 bits to signify each register. At the exit pont (exiting the recompiled code or the landing pad of another instruction, after say a jump), that register and the known table
<Ameisen>
of registers for that instruction would be harmonizes, with data pushed/popped as needed.
<Ameisen>
It should work well _within_ a chunk, but loops or such that happen to go across chunk boundaries might suffer.
<Ameisen>
I'm not sure how bad that overhead would be yet as I haven't written the assembly to handle the harmonization, and it likly depends upon how different the assignments are.
<Ameisen>
the dynamic recompiler currently reserves 8 host registers (on Win64: b, di, si, bp, 12, 13, 14, 15) so I still have some clearance.
Lucretia has quit [Read error: Connection reset by peer]
vdamewood has quit [Quit: My Mac has gone to sleep. ZZZzzz…]