dinkelhacker_ has quit [Ping timeout: 256 seconds]
<heat>
optane was an idea with an expiration date
<heat>
good fucking riddance
<nikolar>
why do you hate optane so much heat
<nikolar>
having ssds that basically never die from writing too much doesn't sound too bad to me
<heat>
which version of optane are you talking about?
<nikolar>
the good one
<nikolar>
3d xpoint thing
<heat>
that means nothign
<nikolar>
which one are you talking about then
<heat>
you had optane as an expensive byte addressable pmem thing, you had optane as a regular ssd, you had optane as a drive cache (most common)
<heat>
drive cache had an expiry date, regular ssds are regular ssds, optane as an expensive byte addressable thing was probably a crap idea at the end of the day anyway too (and needs explicit filesystem support to be _really_ good)
<nikolar>
i meant optane ssds
<nikolar>
that you can have like millions of write cycles before they die
hwpplayer1 has joined #osdev
<heat>
thats legit but i'd guess there's not a lot of money involved in that business
<heat>
vs the R&D cost
<heat>
intel is firstly a cpu vendor and secondly a foundry
<heat>
thirdly everything else
<nikolar>
yea
<zid>
they sold the business though right
<zid>
not just deleting the tech
<zid>
heat I came up withj a solution for your timer code elsewhere
<zid>
[14:52] <zid`> keep a big table of how many primes/sec each microarch should be able to do per ghz
<zid>
[14:52] <zid`> time against that
<nikolar>
zid: i think they've sold it to micron
<nikolar>
but i don't think any new ssds came out of that
<zid>
yea probably just cannibalized the process node to make someother thing
eddof13 has joined #osdev
<heat>
i have a horrible thing to tell you
<heat>
you can't measure you frequency without measuring the base freq beforehand
Left_Turn has joined #osdev
goliath has quit [Quit: SIGSEGV]
bauen1 has quit [Ping timeout: 246 seconds]
hwpplayer1 has quit [Quit: I'll be back later]
<immibis>
heat: reminds me of microsoft's naming schemes
bauen1 has joined #osdev
antranigv has joined #osdev
Turn_Left has joined #osdev
Left_Turn has quit [Ping timeout: 244 seconds]
FreeFull has quit []
karenw has joined #osdev
<Ermine>
> intel is secondly a foundry --- separating manufacturing stuff into a subsidiary and going fabless seems like good idea
<GeDaMo>
I think Gelsinger wanted to do that
<nikolar>
they kind of did
<Ermine>
amd did that
<nikolar>
the foundary side is pretty independant and i think tsmc is manufacturing some of intel's stuff too
<nikolar>
Ermine: yea
<GeDaMo>
"The same can't be said of Intel's ailing foundry unit, for which the real reckoning came in early 2024 when Intel officially split off Foundry as a standalone unit and released revised financial reports revealing that the unit had bled $7 billion in 2023." https://www.theregister.com/2024/12/02/intel_gelsinger_leave/
<bslsk05>
www.theregister.com: Intel CEO takes his leave as ambition meets reality • The Register
<Ermine>
well, that's "kind of"
FreeFull has joined #osdev
karenw has quit [Ping timeout: 265 seconds]
SGautam has quit [Quit: Connection closed for inactivity]
SupUser has joined #osdev
craigo has quit [Quit: Leaving]
xenos1984 has quit [Ping timeout: 252 seconds]
xenos1984 has joined #osdev
op has quit [Remote host closed the connection]
obrien has joined #osdev
<SupUser>
la de da di
<SupUser>
we like's to party
<zid>
like is to party? I think you need to brush up on your vengaboys lore
<SupUser>
is can be any word
<zid>
In other news, tunak tunak, tunak tunak, tuna da da da
<SupUser>
w/e
<nikolar>
zid: tunak tunak
<SupUser>
u guys breaking some records though?
obrien has left #osdev [#osdev]
getz has quit [Ping timeout: 252 seconds]
getz has joined #osdev
xenos1984 has quit [Ping timeout: 252 seconds]
xenos1984 has joined #osdev
strategictravele has quit [Quit: strategictravele]
<bslsk05>
lore.kernel.org: Rust in QEMU roadmap - Paolo Bonzini
karenw has joined #osdev
<nikolar>
yuck
<zid>
"it worked great, after we made everything unsafe"
<zid>
"also you need to grab code from a random internet repo"
<zid>
so, same as every other rust proposal ever
<nikolar>
kek
<fedaykin>
qemu is user-space, why not rust?
<zid>
why rust?
<zid>
It's harder to maintain, slower, and you have to disable most of the 'benefits'
<nikolar>
literally why rust
<fedaykin>
memory safe? easy to debug
<nikolar>
it most certainly isn't as easy to debug as c
<zid>
easy to debug is very not a thing C suffers issues with
<zid>
other languages *wish* they could be debugged like C
<nikolar>
indeed
<nikolar>
especially languages like rust or c++
<fedaykin>
debugging c is hard, have you looked into linux kernel memory model, rcu implementation in c, memory barriers and have you tried to debug a kernel issue?
<zid>
yes?
<nikolar>
and rust helps with that how exactly?
<zid>
Rust will help with none of that, but also you won't get debug info
<fedaykin>
well, good for you, but what if a tram hits you tomorrow?
<zid>
what
<nikolar>
wat
<zid>
"How does rust help you debug the linux kernel memory model" "A train could hit you"
<fedaykin>
i'd rather home someone with less experience being able to fix an issue
<zid>
how does it help *anyone*
<fedaykin>
then depend only on highly skilled few
<fedaykin>
*than
<zid>
I cause some spectucular crash inside the rust stdlib by feeding it invalid shit from an unsafe'ly made reference
<zid>
you're now debugging the same speculacular crash I'd have caused in C
<nikolar>
but zid, rust is safe and easy to debug
<zid>
but now it's inside some compiled rust code
<nikolar>
how don't you ge tit
<fedaykin>
if it's user-space, i say go rust
<zid>
again, to what benefit?
<fedaykin>
less bugs
<zid>
We started at you saying it, we asked why
<zid>
we're back to where we started
<zid>
rust code that actually compiles, I would be generous and say, likely has less bugs, yes. The problem is *getting* there.
<nikolar>
rust = less bugs
<nikolar>
it's just obvious
<nikolar>
no further proof needed
<zid>
Without absolutely destroying various things like maintainability, portability, performance etc
<fedaykin>
have you shared your concerns on the mailing list?
xenos1984 has quit [Read error: Connection reset by peer]
<fedaykin>
what did they say? =)
<zid>
it's qemu's perogative to do what they want
<fedaykin>
right, we are moving along thenm
<nikolar>
who's we
<zid>
I think fedaykin must be all of the lead developers on qemu
<fedaykin>
qemu community
<fedaykin>
i am not =)
<zid>
see
<nikolar>
i am a qemu user and i say we don't move along then
<nikolar>
what nowp
<nikolar>
*what now
<fedaykin>
nikolar: let them know then
<zid>
why?
<nikolar>
i just did?
<zid>
It's a proposal
<zid>
also, you are they
<zid>
and we told you
eddof13 has quit [Quit: eddof13]
<fedaykin>
sounds good, let's see more qualified people than myself address your concerns
<zid>
It's up to them what they care about
<zid>
rust isn't broken, it WILL work
<nikolar>
it's just a bad choice
<zid>
it's just not a decision I'd make, because it's going to cause pain to them and to users
<nikolar>
i am sure having build times explode by an order of magnitude is worth it zid
<zid>
I checked the actual qemu community
<zid>
they think it won't go farther than surface level, because of things like tracepoints
GeDaMo has quit [Quit: 0wt 0f v0w3ls.]
<nikolar>
good
<fedaykin>
guys, have you ever tried to fix a data race in rcu code on some weak memory model archs?
<fedaykin>
how much time did it take you to fix it?
eddof13 has joined #osdev
<nikolar>
i wonder how rust fixes that particular issue
<fedaykin>
you don't have data races with rust
ecs has quit [Remote host closed the connection]
<nikolar>
sure buddy
ecs has joined #osdev
eddof13 has quit [Client Quit]
<fedaykin>
i'd rather play video games
<nikolar>
go ahead?
<fedaykin>
you do you and stay cool, im gonna do me and play video games instead debugging c =)
<nikolar>
i've had much easier time debugging c than javascript if you believe that
* Ermine
ticks off rust on his #osdev bingo card
<nikolar>
Ermine: oi
<zid>
dw, nikolar, javascript doesn't h ave types, so all javascript projects come with 1TB of unit tests that reimplement half a type system
<nikolar>
lol
xenos1984 has joined #osdev
m3a has quit [Remote host closed the connection]
eddof13 has joined #osdev
Dead_Bush_Sanpa1 has joined #osdev
eddof13 has quit [Client Quit]
Dead_Bush_Sanpai has quit [Ping timeout: 260 seconds]
Dead_Bush_Sanpai has joined #osdev
Dead_Bush_Sanpa1 has quit [Ping timeout: 245 seconds]
levitating has joined #osdev
<netbsduser>
fedaykin: rcu isn't done for the fun of it!
<heat>
i see we had some interesting rust discussions as always
<netbsduser>
it's done because kernels are full of shared state
<netbsduser>
it's all good and well to boast of rust's phearless concurrency but you need to engage in frightful concurrency in a kernel
<bslsk05>
twitter: <IanCutress> - Optane officially had final nail under his watch, but was dead long before  - Tofino was removed due to lack of demand: no-one wanted P4 switches, they want P4 NICs  - Royal Core is what? Rumor mill wet dream? 🧴  - Rialto Bridge was removed from roadmap due to lack of demand  - [https://twitter.com/Kepler_L2/status/1863612828982088135 <Kepler_L2>
<bslsk05>
> Becomes Intel CEO  > Kills Optane  > Kills Tofino  > Kills Royal Core  > Kills Rialto Bridge  > Fires 20% of employees  > Refuses to elaborate  > Leaves ]
<nikolar>
heat: kek
<heat>
puck, bslsk05 has been super fast now, good shit