<klange>
Doing more dumb stuff... previously, menu windows would be placed by the client application calculating where they should go based on the reported x,y of the base window... which worked fine for normal windows, but if you rotated the window the menus would show up in the unrotated location (and not be rotated themselves)
fkrauthan has quit [Remote host closed the connection]
<zid>
what was your solution to ShowWindow jank
<klange>
Added a command to the compositor to move a window relative to an existing window's coordinate system, now this works: https://klange.dev/s/relative_windows.png
<klange>
zid: defer rendering until the first update command is received, works great
<zid>
so basically showwindow but automagic
<zid>
neat
fkrauthan has joined #osdev
<klange>
pretty much, means no api change for clients, except that _one_ client that was incorrectly never sending an update :)
ElectronApps has joined #osdev
<graphitemaster>
klange, Should probably just have a proper transform matrix and then you can parent these to the window, same deal but matrices compose and you can do arbitrary parenting and composition of elements without needing to do special case math / branches / custom commands.
<graphitemaster>
That than handles scale, rotation, and translation all as one.
isaacwoods has quit [Quit: WeeChat 3.2]
gog has quit [Ping timeout: 268 seconds]
ElectronApps has quit [Remote host closed the connection]
ElectronApps has joined #osdev
<klange>
graphitemaster: I am actually sitting here with my editor open poking around at replacing the coordinate mapping system that currently just handles rotation... with the transformation matrices I _do_ already have ;)
<klange>
The animation system / actual blitter uses affine transformation matrices as of a few weeks ago, same as it did back when it was implemented with cairo...
<bslsk05>
github.com: toaruos/compositor.c at master · klange/toaruos · GitHub
sprock has quit [Remote host closed the connection]
sprock has joined #osdev
smeso has quit [Quit: smeso]
flx- has joined #osdev
flx has quit [Ping timeout: 248 seconds]
smeso has joined #osdev
sts-q has quit [Ping timeout: 248 seconds]
mctpyt has quit [Ping timeout: 268 seconds]
sts-q has joined #osdev
ElectronApps has quit [Remote host closed the connection]
ElectronApps has joined #osdev
seanbox has joined #osdev
<seanbox>
hello
<seanbox>
i have encountered an annoying deadlock between my page cache and swap thread
<seanbox>
basically my page cache will lock a mutex on the cache info for a file, it will then look to see if a page out of the file is already present in memory, if so it returns it
<seanbox>
otherwise it allocates a new page to read the page into, reads the data in, and unlocks the cache info
<seanbox>
if there are no free pages available, itll block and wait for the swap thread to evict and spit one out
<seanbox>
(for 5 seconds, then it gives up and reports an out-of-memory error)
<moon-child>
allocate before locking
<moon-child>
(or, rather, keep an allocated page around, refreshing it once you use it up; so, allocate after unlocking)
<seanbox>
hm thats not a bad idea lol
<seanbox>
i didnt finish describing the problem but u seem to have gotten the gist
<moon-child>
another option: allocate on the stack, read into there, unlock, them mallocate and copy out of the stack
<moon-child>
that involves extra copies but entails marginally (1 page) less memory use
<moon-child>
but not bad if you only do that when the alloc fails
<seanbox>
another potential solution i was thinking of was to have it try to allocate the page once, and prohibit blocking that time, so it returns immediately if there are no free pages
<seanbox>
and if it fails, try to get an unused page from that file's cache
<seanbox>
if there are none, then it tries again with blocking allowed that time
<seanbox>
and it being guaranteed that the page cache wont try to evict one of that file's pages, thus avoiding the deadlock
<seanbox>
has the side effect of preferentially evicting a file's own pages before anyone else's which could be weird
<clever>
seanbox: i think linux has a min-free tunable, it will always keep that much ram totally free, so it has some emergency pages to dip into, when it cant do io and complex tasks
<clever>
seanbox: the read-cache is also rigged up, so you can discard pages of it at any time, to reclaim more ram
<seanbox>
i also have must-succeed pages but i was thinking itd be yucky to use that here
nyah has quit [Ping timeout: 248 seconds]
Vercas has quit [Remote host closed the connection]
<seanbox>
when the refcount of a page frame drops to 0 it gets put on one of two global LRU lists of evictable pages
Vercas has joined #osdev
<seanbox>
one is "fast-evict" where the only thing that needs to happen to evict the page is some data structures getting updated
<seanbox>
those get evicted first
<seanbox>
the other is slow-evict where disk I/O has to occur to evict that page frame
<seanbox>
i was thinking i might have it so that threads can synchronously yoink fast-evict pages without waiting for the swap thread to do it
<seanbox>
if they have to
<seanbox>
that would also avoid the deadlock because it would just recursively acquire the cache info mutex which would instantly succeed
<seanbox>
im just probably overlooking some other race conditions or deadlocks that might cause
<clever>
seanbox: have you heard about the ARC in zfs?
<seanbox>
i have not
<clever>
basically, it keeps a cache of both recently used blocks, and frequently used blocks
<clever>
and it auto-tunes the ratio of how much to dedicate to each
<clever>
it does that, by keeping track of blocks it discarded from the cache
<clever>
so if something WAS in the "recent" set, but was dropped, and then read again
<clever>
then maybe the recent set needs to be bigger
<clever>
by tracking what it had dropped, it knows when it made a mistake, and can change how much ram it dedicated to the 2 types, to not make the mistake again
<clever>
"frequently used blocks" i think, refers to something that your maybe reading every 10 minutes for example
<clever>
while "recent" is much shorter term, and discards more quickly
<clever>
but if there is a sudden short-term burst of activity, bigger then what fits in the recent cache, it knows the misses are stuff it just had, and can make it bigger
shlomif has joined #osdev
seanbox has quit [Remote host closed the connection]
kuler has quit [Remote host closed the connection]
<superleaf1995>
ahhh, the joys of developing for s390x
<shlomif>
superleaf1995: that ibm arch?
<moon-child>
I kinda wish all-ones were a trap representation
<moon-child>
so e.g. maxshort is 32767; minshort would be -32767, and maxushort would be 65534
<moon-child>
that way I wouldn't have to deal with these horrible, awful, asymmetric fixed-point audio formats
<kingoffrance>
how symmetric do you want ? wasnt that called sign and magnitude, with +0 and -0 :)
<moon-child>
no not sign-magnitude
<kingoffrance>
:)
<moon-child>
still two's-complement, but it fudges it if you overflow
<moon-child>
so e.g. 32767+1 = -32767
<moon-child>
i recognise this is horribly impractical and loses a lot of the desirable properties of two's complement
<moon-child>
_but_
<moon-child>
I also really hate the asymmetric fixed-point thing
nj0rd has quit [Quit: WeeChat 3.2]
srjek has quit [Ping timeout: 245 seconds]
bradd has joined #osdev
ElectronApps has quit [Remote host closed the connection]
<bslsk05>
blog.codinghorror.com: COBOL: Everywhere and Nowhere
<zid>
moon-child: yea that'd be pretty useful
<zid>
plus you'd be able to use it as an error signal like EOF
kuler has joined #osdev
<superleaf1995>
well i mean i'm *trying* to make JCL command line for my toy kernel thing :P
<superleaf1995>
and i've even learned the good ol' COBOL
nj0rd has joined #osdev
seanbox has joined #osdev
Izem has joined #osdev
ElectronApps has quit [Ping timeout: 248 seconds]
dude12312414 has quit [Quit: THE RAM IS TOO DAMN HIGH]
Izem has left #osdev [#osdev]
ElectronApps has joined #osdev
kuler has quit [Remote host closed the connection]
kuler has joined #osdev
ElectronApps has quit [Remote host closed the connection]
ElectronApps has joined #osdev
superleaf1995 has quit [Quit: Lost terminal]
kuler has quit [Remote host closed the connection]
kuler has joined #osdev
bradd has quit [Remote host closed the connection]
tacco has joined #osdev
bradd has joined #osdev
bradd has quit [Remote host closed the connection]
mhall has joined #osdev
bradd has joined #osdev
sortie has joined #osdev
Burgundy has joined #osdev
Belxjander has joined #osdev
matt|home has quit [Ping timeout: 248 seconds]
dormito has joined #osdev
GeDaMo has joined #osdev
kwilczynski has quit []
dormito has quit [Ping timeout: 256 seconds]
divine has quit [Ping timeout: 268 seconds]
divine has joined #osdev
sprock has quit [Quit: brb]
sprock has joined #osdev
dormito has joined #osdev
tenshi has joined #osdev
Arthuria has joined #osdev
seanbox has quit []
dennis95 has joined #osdev
sham1 has joined #osdev
Vercas4 has joined #osdev
Vercas has quit [Ping timeout: 244 seconds]
Vercas4 is now known as Vercas
ElectronApps has quit [Read error: Connection reset by peer]
ElectronApps has joined #osdev
mctpyt has joined #osdev
thinkpol has quit [Remote host closed the connection]
thinkpol has joined #osdev
gog has joined #osdev
devcpu has quit [Quit: leaving]
devcpu has joined #osdev
<klange>
I build a partial implementation of a rectangular clipping system for my graphics lib and while it's working it seems to be slower, or on par with, my dumb-as-bricks scanline-only clipping, despite generally re-rendering a lot less of the screen...
<klange>
So now I don't really want to bother porting it to remaining graphics routines... at least until I figure out the slowness.
<klange>
*sigh*
<klange>
I'm not even sure why I started looking at this... it wasn't even really related to what I was doing this morning...
<bslsk05>
fgiesen.wordpress.com: Min, Max under negation and an AABB trick | The ryg blog
<klange>
That's a bounding box...
<GeDaMo>
When you said clipping that popped into my head
<GeDaMo>
I'm not claiming it's relevant :P
<klange>
It probably because - I should throw all of this out and replace it with bounding boxes ;)
<klange>
it probably is*
<GeDaMo>
I remember there was a software implementation of DX9, I wonder if that was ever open sourced
<klange>
The terminal has its own bounding box thing going on to simplify the clips it sends to the server, but I don't do any tricks like that as it composes on the fly.
<klange>
eg. it's a bunch of calls to like "add this cell" and each one pushes the min/max bounds as it goes, interspersed with other stuff, and it doesn't need to keep/juggle the boxes as it only uses them once
<GeDaMo>
It was DX7, it was called Pixomatic and as far as I can see it wasn't opensourced
ElectronApps has quit [Remote host closed the connection]
nur has joined #osdev
nvmd has joined #osdev
heat has joined #osdev
nismbu has quit [Ping timeout: 256 seconds]
tenshi has quit [Quit: WeeChat 3.2]
rwb has quit [Remote host closed the connection]
rwb has joined #osdev
nismbu has joined #osdev
CryptoDavid has joined #osdev
Arthuria has quit [Ping timeout: 245 seconds]
Izem has joined #osdev
<Izem>
If the os is just another program, why does it seem hard to develop with non C like languages?
<heat>
because the OS is not just another program
<heat>
the kernel (which is the thing that needs to be written in a low level language) is the basis of the whole system, on which your programs run (like a language virtual machine or an interpreter)
<heat>
it also has some constraints with respect to memory usage and realtime stuff and threads and whatnot that make it less suitable for GC languages and whatnot
<heat>
still doable, but it's not great
<heat>
you can totally write the rest of the OS in other languages though, windows has a fair share of C# for example
nvmd has quit [Quit: Later, nerds.]
<GeDaMo>
There have been OSs written in all sorts of languages e.g. Lisp
<GeDaMo>
There are some parts which absolutely have to be in efficient machine code like interrupt handlers
<Izem>
I see
<sham1>
Well at least parts of the interrupt handling process. Even if all it does is just that it informs your runtime that "an interrupt has happened, deal with it"
<Izem>
C is a really good choice since it has been considered as a portable assembler right?
<heat>
yes C is a good choice
<heat>
but so is C++ and rust
<sham1>
Well C is not a portable assembler, but as far as high level languages go, it certainly is lower abstraction level than most and doesn't require as much support as, well, even the aforementioned C++ and Rust to get working
<heat>
you don't need a portable assembler, you need something that has all the abstractions *you* want + the ability to go to the lowest of abstraction levels
<GeDaMo>
You often need to be able to access arbitrary memory addresses and C has that built in with no "protections"
matt|home has joined #osdev
freakazoid343 has joined #osdev
<Izem>
ok, that seems a wide enough berth
<sham1>
Of course one thing you can do, is build a runtime implementation for something in C and then get the benefits
<Izem>
sham1: yeah that seems ok, though I would think you would put off having multiple languages if you could
sortie has quit [Quit: Leaving]
<Izem>
hmm, modula2 is in gcc
mahmutov has joined #osdev
sortie has joined #osdev
shlomif has quit [Ping timeout: 268 seconds]
freakazoid343 has quit [Ping timeout: 245 seconds]
<bslsk05>
FEX-Emu/FEX - A fast usermode x86 and x86-64 emulator for Arm64 (16 forks/282 stargazers/MIT)
<Bitweasil>
I wonder how this performs. Looks like it should be a lot better than qemu userspace.
kspalaiologos has quit [Quit: Leaving]
<geist>
would be surprised frankly
<geist>
QEMU isn't amazing, but it does a pretty good stab at binary translation. you'd need a somewhat more sophisticated implementation to surpass it, and that starts getting into really hard stuff
<geist>
but... from glancing at its readme it does have a jit and whatnot
<geist>
and a thunking scheme for various libs
superleaf1995 has joined #osdev
<Bitweasil>
Yeah, thunking out to ARM seems like a lot of win right there.
<Bitweasil>
I'll have to mess with it.
<Bitweasil>
Anyone want a M1 Mini, should be surplus soon?
<j`ey>
why? :O
Izem has quit [Quit: Lost terminal]
<moon-child>
there are at least a couple of console emulators that have their own fairly sophisticated arm implementations--but that's going the other way!
<geist>
Bitweasil: heh already done with it or you're predicting the M2s?
<Bitweasil>
moon-child, those can probably do pre-compilation stuff... I don't know how much of that is raw CPU horsepower vs GPU either.
<Bitweasil>
geist, trying to de-apple my life. :(
<geist>
ah. got it.
<Bitweasil>
I've been a bit upset with them for a while over the iCloud/China stuff, their continuing general abuse of employees in foxconn ("We're sorry for getting caught, we will try harder to prevent leaks"), a few other things, and... this recent on-device scanning just kicked me over the edge.
<Bitweasil>
I'm aware it's probably fine as is, the crypto is solid ("LOOK AT THE CRYPTO!" scream the whitepapers), and I simply don't like where things are going anymore.
<moon-child>
Bitweasil: I don't think any of them do precompilation; and usually are more cpu- than gpu-dependent
<Bitweasil>
But the focus on the crypto is somewhat at odds with the fact that the neuralhash algorithm is super easy to trip up with minor changes, and the door has been opened to more on-device stuff.
simpl_e has joined #osdev
<Bitweasil>
I thought Apple was going to be OK, but they're clearly not, and have bowed, presumably twice now, to governments in some way or another.
<Bitweasil>
Which, fine, their decisions, but I disagree with that.
<Bitweasil>
Unfortunately, Windows is also out, as Win11 is apparently both removing the offline account option entirely, and requiring fairly new hardware for no discernible reasons.
<Bitweasil>
So, looks like I'm heading down the greybeard path of Linux. Except, I also don't like Intel anymore, see uarch stuff. So, either AMD or ARM it is... I guess.
<moon-child>
yeah, there are lots of demos people have done of making an ml recognition see something that's obviously (to a human) not there
<Bitweasil>
I'm mostly just running out of tech I feel I can ethically use, and what remains is... problematic too, in terms of Chinese supply chains and such.
<Bitweasil>
Apple's CP detection isn't ML, it's just hash based.
<Bitweasil>
Perceptual hashes.
<Bitweasil>
You can mess with it if you've got a recent Mac.
<Bitweasil>
The "sexting detection" is ML, which should be *hysterical* to see what it false flags.
<geist>
this path leads to madness
<geist>
as i'm sure you're aware
<geist>
gotta search for a reasonable comfort level
<Bitweasil>
I am both aware and embracing the madness. :)
<kazinsal>
the CSAM detection is just using the same NCMEC database that most other stuff does
<moon-child>
don't particularly care for macos either way; but I do kinda want to play with an m1
<Bitweasil>
I know, and I fundamentally object to using my device resources to search for it before stuff is uploaded to iCloud, and then they're talking about extending the various protections out, so it's "I have to prove my innocence before uploading," which I'm not really OK with.
<geist>
yah dont have to use their ecosystem
<kazinsal>
the "concern" people have is that the DOJ is going to ask NCMEC to add non-CSAM stuff to their database which is a bizarre concern because NCMEC does not care about anything that isn't CSAM
<moon-child>
however I already have a pretty decent desktop; am more in the market for a laptop atm. And not sure if I more want that or to get a recent intel, because I also wanna play with avx512
<geist>
yah same thing. i'd kinda like to piddle with avx512 but am currently very happy with AMD solutions
<HeTo>
kazinsal: you really think databases like that won't be abused?
<moon-child>
kazinsal: yeah, the running-on-device bit is what skeeves me out. I don't want to _touch_ that database or have anything to do with it
<moon-child>
scanning in icloud is fine
<HeTo>
uhh is the database maintained by a private nonprofit?
<kazinsal>
it sounds like they're just going to run a perceptual hash on device and send that to the database
<Bitweasil>
kazinsal, they've got some whitepapers out that explain it.
<moon-child>
HeTo: government-sponsored nonprofit
<Bitweasil>
The crypto is legitimately neat.
<moon-child>
kazinsal: yes. I object to that
<Bitweasil>
Minimum share requirement, false matches to obscure total match count, etc.
<moon-child>
kazinsal: (rather, they run the hash on-device, compare it against hte db of hashes whcih is also on the device, and send a signature of match-or-not to server)
<Bitweasil>
But I simply don't like the direction, have been uncomfortable with the way things are going, and am opting out, please give me my freedom grope.
<Bitweasil>
moon-child, read the whitepaper.
<Bitweasil>
The crypto is legitimately neat.
<geist>
sure
<geist>
i totally get that
<kazinsal>
HeTo: you really think that state-level malactors are going to go to the effort of poisoning a well-known set of data used to take down pedophiles instead of just, y'know, doing state level malactor shit like disappearing you
<Bitweasil>
Back in a bit.
<kazinsal>
it's another branch of the mossad/not-mossad threat model
<Bitweasil>
kazinsal, I don't like that they now have the trivial option to add more on devie scanning, and the seal on that has been broken. My device is now actively doing things not for my benefit.
<HeTo>
kazinsal: how would they know whom to disappear?
<moon-child>
kazinsal: privacy is less about threat model and more about ideology
<Bitweasil>
Which is different from cloud servers doing things not for my benefit.
<moon-child>
for me at any rate
<kazinsal>
specifically, the threat model wherein the US Marshals stuff you in a trunk, shoot you 30 times through the lid, leave a note behind saying WE TOTALLY DID IT LOL, and then tell your family it was a suicide
<Bitweasil>
brb
* geist
gets back to work for The Man
<HeTo>
kazinsal: we know that the Finnish child porn web site block list was used to block more gay porn web sites than child porn web sites
<geist>
gotta meeting in a bit where i present <redacted> for the council of <redacted> concerning your <redacted>
<kazinsal>
HeTo: if they consider you enough of a threat to inject data they're targeting you for into a CSAM database so they can snap you up they already *know* who you are
<kazinsal>
also people are REALLY bad at shutting the fuck up about the illegal shit they want to do or whatever they're targeting for tbh
<HeTo>
kazinsal: there are some sensitive images that leaked but not widely. they want to track down who are the people who got those images. put hashes for those images in the database, and boom, Apple will tell you who those people are
<kazinsal>
I'm of the opinion that everyone is already constantly pwned, and combined with the objective lack of ethical consumption under capitalism, there is nothing you can do that is morally "good" other than to leave technology behind and wander off into the forest to become a scavenging hermit
GeDaMo has quit [Quit: Leaving.]
Starfoxxes has joined #osdev
dormito has quit [Ping timeout: 256 seconds]
<Bitweasil>
HeTo, depends on the threshold value used.
<Bitweasil>
kazinsal, I've definitely been considering if it's anything faintly bordering on ethical to use /any/ consumer tech, yes.
<Bitweasil>
:(
<kazinsal>
capitalism is an inherently unethical system that we are all forced into interacting with
<kazinsal>
if you want to be morally positive then do good in your community to offset the issues you have with your iPhone
<kazinsal>
donate to a food bank. volunteer at a local charity. plant and maintain a community garden
ecs has quit [Read error: Connection reset by peer]
<kazinsal>
it is useless and in fact entirely counterproductive to self-flagellate every time you have to interact with a website that has twitter embeds or whatnot. spend the time you would otherwise be wasting on that making an actual tangible difference for real humans in your own community
nshp has quit [Remote host closed the connection]
raggi has quit [Remote host closed the connection]
<Bitweasil>
I block the majority of that stuff anyway.
<Bitweasil>
But, I'm fairly rapidly heading in the direction of using consumer tech professionally, and that's it.
<Bitweasil>
:/
<Bitweasil>
The old open source greybeards had something going 20 years back.
<Bitweasil>
It has far more appeal now.
<kazinsal>
the old open source greybeards were really more interested in just excluding people from their little circus
<kazinsal>
again it's worthless to self-flagellate
<kazinsal>
it solves zero problems and just makes you feel worse
<kazinsal>
at that point you might as well shave your head and become an ascetic monk eating plain gruel three meals a day
<heat>
yo, any particularly strong opinions on surface laptops vs macbook airs?
raggi has joined #osdev
bigassking has joined #osdev
<kazinsal>
the M1 air is a nice little piece of kit and has a nicer keyboard than the standard tier of surfaces but the surface will let you install bizarre shit to bare metal on account of being x86
<heat>
i kind of gave up on the PC build dream for now, GPUs are too expensive so I'm wanting something relatively light and portable
<moon-child>
Bitweasil: even with what the old graybeards were doing there were problems
<moon-child>
with attention
<heat>
something that scares me about the surface is the worst case scenario where something screws me over and I can't install linux
<moon-child>
human mind is not set up for unlimited instant gratification
<heat>
i know mac already has the unix-like thing so that's safer
<Bitweasil>
Correct, though the dopamine loops are convenient for advertisers.
<kazinsal>
I believe Linux doesn't have I2C touchpad drivers for the surface laptop 4 yet
<Bitweasil>
kazinsal, but that doesn't mean to just continue using consumer tech I disagree with. Though the whole monk thing has some appeal at times.
<Bitweasil>
ppst. heat. I hear, uh... you're looking for a gee-pew. Listen, I know a guy... got some gee-pew. Got some thirty-ninety, only a couple Gs... you interested?
<kazinsal>
the surfaces are nice devices but they are very much geared towards people who are happy with the way a laptop arrives
<kazinsal>
and/or enterprise users
<heat>
Bitweasil, :D
<kazinsal>
linux will always be an uphill battle on those kinds of devices because the target market for linux has always been and always been people who are cool with cobbling together hacked up upgrades for 10-year-old thinkpads
<heat>
yeah I was looking at laptops and both macs and surfaces look relatively interesting, good looking, capable, nice battery life
<heat>
they look handy for uni related stuff
<kazinsal>
I like my M1 air. great battery life, excellent screen, keyboard's nice
<Bitweasil>
Hey now. Five year old Thinkpads run Linux just fine.
<Bitweasil>
I would rather have an Air than a Surface.
<heat>
the only thing I really really want is some unix-ish thing and I'm not too enthusiastic about using WSL2 full time
<Bitweasil>
But would really rather not have either at this point. :(
<Bitweasil>
Dell XPS13/15 and Linux?
<Bitweasil>
I think they've got some partnership with Ubuntu so you have legitimate hardware support for everything.
<moon-child>
I think they ship some xpses with linux
<kazinsal>
it only has two USB-C/Thunderbolt ports on the one side and a headphone jack on the other, but you can get apple-OKed expando-things for like thirty bucks that turn the two USB-C ports into a USB-C/thunderbolt, two USB 3.1 type As, a microSD reader, and a GbE port
<Bitweasil>
Stuff Windows in a VM and run Linux on the iron.
<Bitweasil>
moon-child, I thought they were doing that, at least they used to have the developer edition ones that were shipped with Linux. Not that I'd trust a Dell Linux install...
bigassking has quit [Remote host closed the connection]
<heat>
Bitweasil: looks like a decent option but I can't find one here
<Bitweasil>
Easy enough to just get it with the bare metal and do my own install.
<Bitweasil>
Oh, hm.
<Bitweasil>
:/
<moon-child>
well, of course not. I wouldn't trust any oem-installed os. But for the people who want 'ootb experience'...
<kazinsal>
correction: geared towards people who are only comfortable with cobbling together hacked up upgrades for 10 year old thinkpads
<Bitweasil>
Ah, yes, the oot-bee.
dennis95 has quit [Remote host closed the connection]
dennis95 has joined #osdev
<heat>
something that really bothers me is how little disk space the laptops have these days
<heat>
where are my TBs :((
<heat>
stupid SSDs
<Bitweasil>
External storage is cheap and fast enough now.
<heat>
you mean like external SSDs?
<Bitweasil>
Yeah.
dutch has quit [Quit: WeeChat 3.2]
<Bitweasil>
USB3 or Thunderbolt or [whatever].
<Bitweasil>
They're legitimately quick now.
<heat>
i have mixed feelings about those, mostly out of fear
<Bitweasil>
You can *get* large SSDs in laptops.
<Bitweasil>
You just pay through the nose.
* Griwes
looks at the two 2tb ssds in his laptop
<Bitweasil>
You like big storage and you cannot lie!
dutch has joined #osdev
<kazinsal>
I had a 1 TB M.2 SSD on my desk for like four months that I just kept forgetting to install
<zid>
dang
<zid>
nice life
<kazinsal>
every time I'd run out of space on my SATA SSD it'd be like "shit I really should install that M.2. ooh, funny cat pictures"
heat has quit [Remote host closed the connection]
zid has quit []
zid has joined #osdev
dennis95 has quit [Quit: Leaving]
dragestil_ has quit [Ping timeout: 268 seconds]
dormito has joined #osdev
dragestil_ has joined #osdev
srjek has joined #osdev
ahalaney has quit [Remote host closed the connection]
dutch has quit [Quit: WeeChat 3.2]
dutch has joined #osdev
anon16 has quit [Read error: Connection reset by peer]
zaquest has joined #osdev
zid has quit [Remote host closed the connection]
zid has joined #osdev
anon16 has joined #osdev
sortie has quit [Quit: Leaving]
Izem has joined #osdev
Burgundy has quit [Ping timeout: 248 seconds]
scaleww has joined #osdev
superleaf1995 has quit [Quit: Lost terminal]
<gorgonic->
I love the smell of thread synchronization problems in the morning
<gorgonic->
Except it's evening and I hate them actually
gorgonic- is now known as gorgonical
<gorgonical>
I have HPCG running with loads of openmp parallelism. Hangs on the pine64 unless I have pointless printks in the interrupt handling code to slow things down
<Izem>
I'm curious if you had a choice to use something other than openmp
<gorgonical>
not really
<gorgonical>
Doing single-node performance evaluations pretty much requires openmp unless you're gonna run it serial. These hpc applications are all this way it seems
<Izem>
I was just thinking there's a few others, but I suppose you can't expect much help/support if you use those?
kulernil has joined #osdev
Vercas1 has joined #osdev
hgoel[m] has quit [Ping timeout: 245 seconds]
Vercas has quit [Ping timeout: 244 seconds]
Vercas1 is now known as Vercas
lockna has quit [Ping timeout: 268 seconds]
happy-dude has quit [Ping timeout: 268 seconds]
kuler has quit [Remote host closed the connection]