[itchyjunk] has quit [Remote host closed the connection]
<gorgonical>
i am still enraged by i2c
<gorgonical>
the fun never ends
<geist>
that is an appropriate thing to be enraged at
<gorgonical>
pardon my french, but this fucking manual doesn't explain very much at all
zaquest has quit [Remote host closed the connection]
<gorgonical>
and attempting to reverse-engineer what the hell linux does to make it work is a fool's errand
<gorgonical>
1000+ lines of code heavily integrated with pin controls, power management, and the whole i2c subsystem to drive a two-wire protocol from the 1980s
vdamewood has quit [Read error: Connection reset by peer]
mrvn has quit [Ping timeout: 252 seconds]
vdamewood has joined #osdev
Starfoxxes has joined #osdev
dude12312414 has joined #osdev
<geist>
oh wow
<geist>
re: the candle
<gorgonical>
I looked and co-60 decays in all three rays
<gorgonical>
So that's... unsafe
elastic_dog has quit [Ping timeout: 250 seconds]
elastic_dog has joined #osdev
<klange>
forbidden candle
<clever>
ive also seen a forbidden cup
<clever>
the image came with some russian text, asking what it is, because it looks empty, but opening the lid makes the camera go whack
alecjonathon has joined #osdev
smeso has quit [Quit: smeso]
dude12312414 has quit [Quit: THE RAM IS TOO DAMN HIGH]
smeso has joined #osdev
zaquest has joined #osdev
<geist>
the PCID stuff is kinda a fun puzzle it seems
<geist>
it's not super flexible so you have to kinda design around it it seems
Vercas6 has quit [Quit: buh bye]
Vercas6 has joined #osdev
gildasio has quit [Remote host closed the connection]
gildasio has joined #osdev
Vercas6 has quit [Client Quit]
Vercas6 has joined #osdev
Vercas6 has quit [Client Quit]
Vercas6 has joined #osdev
Vercas6 has quit [Client Quit]
Vercas6 has joined #osdev
<geist>
oh yeah also forgot that PCID/INVPCID didn't really get picked up on AMD until Zen 3
<geist>
so you really cant rely on it being ubiquitous
<geist>
OTOH Zen3 has the global TLB shootdown stuff (INVLPGB)
bgs has joined #osdev
orthoplex64 has quit [Read error: Connection reset by peer]
orthoplex64 has joined #osdev
<gorgonical>
holy jesus I got it
<gorgonical>
I had to actually email the driver developer but it works
<gorgonical>
also, what a lad for responding so fast
<geist>
noice!
<gorgonical>
the tl;dr is that I forgot to setup the clocks via the ccu. I was misled into thinking the i2c interface set its own clock
<gorgonical>
but you have to both de-assert the reset pin for it, and then un-gate the clock signal, so there's TWO blocks lol
<geist>
oooh yeah that's common
<gorgonical>
I should know that by now but I haven't done a lot of original, raw work with peripherals across such a complex soc architecture
<geist>
did you have a scope? usually looking at the pins with the scope is a great way to see if it's a problem on yuour end (setting up i2c) or the device not talking to you (i2c working but reset is asserted, etc)
<gorgonical>
I have a logic analyzer. I was getting nothing on the pins before
orthoplex64 has quit [Read error: Connection reset by peer]
orthoplex64 has joined #osdev
scoobydoob has joined #osdev
scoobydoo has quit [Ping timeout: 240 seconds]
scoobydoob is now known as scoobydoo
bgs has quit [Remote host closed the connection]
<geist>
aww was gonna ask heat what their idea was behind generation counters for the PCID delayed stuff
<geist>
i can't think of how it would work. My solution is probably just to track a bitmap per aspace of cpus that have 'dirty' TLB entries
<geist>
and have the cpu dump the PCID when it next loads the CR3, which is probably what intel had in mind when they implemented it
orthoplex64 has quit [Read error: Connection reset by peer]
orthoplex64 has joined #osdev
<geist>
oh interesting... looking at the current AMD docs, which they just released a few weeks ago
<geist>
there is a new cpuid leaf (0x8000.0026) that pretty completely describes topology even better tha intels
<geist>
*and* has places to describe things like Hybrid stuff (ie, big.little)
wand has quit [Ping timeout: 258 seconds]
eroux has quit [Remote host closed the connection]
eroux has joined #osdev
orthoplex64 has quit [Read error: Connection reset by peer]
orthoplex64 has joined #osdev
orthoplex64 has quit [Read error: Connection reset by peer]
gorgonical has quit [Remote host closed the connection]
gog has quit [Quit: byee]
scoobydoo has quit [Ping timeout: 250 seconds]
scoobydoo has joined #osdev
IRChatter has joined #osdev
Vercas6 has quit [Quit: Ping timeout (120 seconds)]
Vercas6 has joined #osdev
orthoplex64 has quit [Read error: Connection reset by peer]
orthoplex64 has joined #osdev
Burgundy has joined #osdev
gog has joined #osdev
KaitoDaumoto has quit [Remote host closed the connection]
GeDaMo has joined #osdev
vinleod has joined #osdev
vdamewood has quit [Read error: Connection reset by peer]
vinleod is now known as vdamewood
Vercas6 has quit [Quit: Ping timeout (120 seconds)]
Vercas6 has joined #osdev
darkstardevx has quit [Ping timeout: 260 seconds]
Vercas6 has quit [Quit: Ping timeout (120 seconds)]
wand has quit [Ping timeout: 258 seconds]
darkstardevx has joined #osdev
darkstardevx has quit [Remote host closed the connection]
darkstardevx has joined #osdev
bauen1 has quit [Ping timeout: 276 seconds]
wand has joined #osdev
gxt has quit [Ping timeout: 258 seconds]
fatal1ty has quit [Ping timeout: 250 seconds]
fatal1ty has joined #osdev
gxt has joined #osdev
orthoplex64 has quit [Read error: Connection reset by peer]
orthoplex64 has joined #osdev
crm has joined #osdev
<klange>
starting to think i should just drop all my libm stubs - really, anything I'm not using myself - and ship the universal standard largely-Sun-derived libm in my package manager for ports...
orthoplex64 has quit [Ping timeout: 240 seconds]
mrvn has joined #osdev
<klange>
going through my package archives and building ports for aarch64, got quake but it uses pow() to build some gamma adjustment tables, I think for lighting? and my `return 0;` did not make that happy (`return x;` was acceptable for this use case, though)
<klange>
I think I at least need to make sure I have binutils+gcc packages for this release, been a while since I poked at cross-compiling them, so that might get interesting...
Burgundy has left #osdev [#osdev]
<mrvn>
klange: it probably divides by pow(x) somewhere and /0 is bad.
<mrvn>
return x; will be a gamma of 1
<klange>
nah, the end result was I got black frames, but the audio indicated things were working
<mrvn>
oh, then it probably called pow(-x) and multiplied.
<mrvn>
faster
<zid>
return 1.6;
<mrvn>
don't you have exp() to implement pow()?
<klange>
pow(n,0.8)
<sham1>
Wouldn't implementing exp to implement pow also require a log
<mrvn>
sham1: do you ever have exp() but not log()?
<klange>
you'll be delighted to know my exp() is implemented with pow()
<sham1>
Which returns zero
<sham1>
Wow
<klange>
on x86-64, I have a magic FPU sequence that does the y*log2(x) thingy
crm has quit [Read error: Connection reset by peer]
crm has joined #osdev
<sortie>
klange, not a bad idea to drop a custom libm if you're not using it, I am using my a bit for my lil games, but yeah. Is there a standard one to package? Because I'm not sure there is, just a lot of forks and dumps of it. Lots of divergent bug fixes. But I suppose anything is better than return x :)
<sortie>
I was thinking of exploring importing musl's libm since I think they actually went and fixed bugs
<mrvn>
is there a free libm other than the bsd ones?
<sortie>
mrvn, it's basically all forks of the old Sun libm
<mrvn>
sun? my bad. that's what I ment.
<sortie>
But yeah it's a vast genetic tree of mutated libms
<bslsk05>
lwn.net: Would you like signs with those chars? [LWN.net]
<heat>
C is bonkers
<Ermine>
Thanks for subscriber content!
<heat>
np
<heat>
sub to lwn, it's pretty cheap
<heat>
and the content is gr8
<Reinhilde>
C is three frogs in a trenchcoat
<Mutabah>
Today's evil hack: handle SIGSEGV and resume :D
<Reinhilde>
Handle SEGV _via a selfpipe_
<Mutabah>
(handle it the same you handle a vm8086 GPF - emulate the faulting instruction and advance RIP)
<Mutabah>
Why - running a windows driver in linux userland, and it reads fixed kernel addresses
<mrvn>
what makes signed chars actually faster on x86 and unsigned faster on ARM?
<Mutabah>
Wait, speed?
<Mutabah>
Isn't it just legacy?
<heat>
yes
<heat>
it would only matter if your sign extension is slow
<heat>
or your zero extension
<mrvn>
Mutabah: someone picked it for each arch. And the existance of signed or zero extending used to be a factor.
<heat>
which it isn't for any modern arch
<heat>
Mutabah, handling SIGSEGV is great
<heat>
have you gotten to decode instructions yet?
<mrvn>
"it can be signed or not, depending on whatever is most convenient to implement on any given architecture." What makes it convenient one way or the other on modern archs?
<Mutabah>
heat: Just a fixed single instruction sequence for now
<mrvn>
and " Segher Boessenkool pointed out that it constitutes an ABI change, and could hurt performance on systems that naturally want char to be unsigned."
<Mutabah>
(both accesses i've found have been `mov rax, [rax]`
<Reinhilde>
if you found the erroneous pointer in a structure that is itself pointed to by RAM you can free, store the pointer somewhere before you use it, and then pass that up to the SEGV handler if you get a SEGV so that you can destroy the inconsistent state and recover
<heat>
Mutabah, yeah but the funny bit is you can get SIGSEGV for stuff that isn't a seg fault
<mrvn>
Reinhilde: so a SEGV on a.b would turn access to a into a SEGV?
<Reinhilde>
I don't know
<Mutabah>
heat: Oh definitely, it checks that it was that instruction, and one of the known addresses
<heat>
like GPFs are segv, and some instructions (like hlt, etc) are segv
<Mutabah>
otherwise it re-throws
<Reinhilde>
it turns access to a that requires (*(a.b)) to work into an event that crashes that connection
<j`ey>
Mutabah: you mean sigsegvs in the kernel?
<Reinhilde>
sorry, I have bad ideas.
<Mutabah>
j`ey: Nah - userland... pretending to be kernel
<mrvn>
Mutabah: when you mmap() and mprotect() you can implement your own page fault handlers in user space. Some garbage collectors implement their own COW mechanism so they can process the heap in parallel.
<Mutabah>
Heh, that seems like awesome hackery
<mrvn>
Mutabah: so when they get a SEGV on write they make the page writable and mark it as needing to be processed again.
<j`ey>
Mutabah: oh I missed you 'why'
<j`ey>
*missed your 'Why' message
<heat>
why why
<heat>
why would you ever run windows kernel drivers in userspace
<mrvn>
.oO(because, because, because I said so, GO TO YOUR ROOM)
<Mutabah>
heat: LINUX userspace
<j`ey>
Mutabah: is this for fun or work?
<Mutabah>
Which probably gives you an idea why
<Mutabah>
"fun"
<mrvn>
heat: some hardware doesn't have linux drivers
<j`ey>
write one then :P
<mrvn>
no docs
<Mutabah>
j`ey: I considered that...
<j`ey>
RE
<heat>
what hardware
<Mutabah>
And got a large chunk of the way through the RE too
<j`ey>
M1 has no docs, and that runs linux :P
<mrvn>
heat: wireless cards come to mind
<Mutabah>
but then was like "eh - too hard, let's just wrap it instead"
<j`ey>
Mutabah: run windows? :P
<heat>
M1 code was written by INSANE people
<Mutabah>
heat: Capture card
<mrvn>
heat: not sure if it is an issue any more nowadays
<Ermine>
heat: re sub: I cannot due to political reasons. :(
<heat>
NDISWrapper may help
<j`ey>
heat: rude!
<j`ey>
Mutabah: buy a USB capture card?
<heat>
they did what you're doing but in the kernel
<mrvn>
heat: I assumed that is what they where talking about
<heat>
Ermine, hm? why?
<Mutabah>
j`ey: I have to "PCI" ones (both are USB under the hood) already
<heat>
sanctioned country?
<Mutabah>
This isn't really meant to be practical, it's mean to be FUN
<gog>
noooo you can't pass something without a null-terminator to strlen
<heat>
look at that bad boy
<gog>
??????? what
<gog>
gcc returns 0
<gog>
or
<gog>
calls strlen when it really really shouldn't
<gog>
i mean
<gog>
the programmer shouldn't
<gog>
bad heat
<gog>
stop playing in the IDB/UB muck
<gog>
you're going to get hurt
<heat>
#musl did it
<heat>
i'm just amazed at the stupid
<gog>
lmao
<gog>
never use other people's c code
<heat>
no
<heat>
musl doesn't have this
<heat>
they were just talking about this specifically
<heat>
musl is usually unreadably correct
nvmd has joined #osdev
JerryXiao has quit [Quit: Bye]
JerryXiao has joined #osdev
crmur__ is now known as orthoplex64
xenos1984 has quit [Ping timeout: 255 seconds]
xenos1984 has joined #osdev
nyah has joined #osdev
DrPatater has joined #osdev
antranigv has quit [Ping timeout: 252 seconds]
DoubleJ4 has joined #osdev
troseman has quit [Ping timeout: 252 seconds]
justache has joined #osdev
les_ has joined #osdev
childlikempress has joined #osdev
sprock has quit [Ping timeout: 252 seconds]
les has quit [Ping timeout: 252 seconds]
kori has quit [Ping timeout: 252 seconds]
antranigv has joined #osdev
sjs is now known as Guest4901
Guest4901 has quit [Killed (zirconium.libera.chat (Nickname regained by services))]
klange has quit [Ping timeout: 252 seconds]
eau has quit [Ping timeout: 252 seconds]
moon-child has quit [Ping timeout: 252 seconds]
ozarker has quit [Ping timeout: 252 seconds]
Clockface has quit [Ping timeout: 252 seconds]
dormito has quit [Ping timeout: 252 seconds]
arminweigl has quit [Ping timeout: 252 seconds]
Patater has quit [Ping timeout: 252 seconds]
ozarker_ has joined #osdev
arminweigl_ has joined #osdev
sjs has joined #osdev
Ermine has quit [Ping timeout: 252 seconds]
j`ey has quit [Ping timeout: 252 seconds]
merry has quit [Ping timeout: 252 seconds]
DoubleJ has quit [Ping timeout: 252 seconds]
weinholt has quit [Ping timeout: 252 seconds]
stux has quit [Ping timeout: 252 seconds]
jeaye has quit [Ping timeout: 252 seconds]
meisaka has quit [Ping timeout: 252 seconds]
DoubleJ4 is now known as DoubleJ
epony has quit [Ping timeout: 252 seconds]
koolazer has quit [Ping timeout: 252 seconds]
night has quit [Ping timeout: 252 seconds]
pie_ has quit [Ping timeout: 252 seconds]
justHaunted has quit [Ping timeout: 252 seconds]
vancz has quit [Ping timeout: 252 seconds]
valeriusN has quit [Ping timeout: 252 seconds]
arminweigl_ is now known as arminweigl
merry has joined #osdev
amj has quit [Ping timeout: 252 seconds]
cross has quit [Ping timeout: 252 seconds]
vancz has joined #osdev
pie_ has joined #osdev
night has joined #osdev
sprock has joined #osdev
klange has joined #osdev
valerius_ has joined #osdev
dormito has joined #osdev
epony has joined #osdev
j`ey has joined #osdev
eau has joined #osdev
meisaka has joined #osdev
koolazer has joined #osdev
cross has joined #osdev
Ermine has joined #osdev
troseman has joined #osdev
ids1024 has joined #osdev
amj has joined #osdev
jeaye has joined #osdev
justache is now known as justHaunted
Matt|home has quit [Quit: Leaving]
gxt has quit [Ping timeout: 258 seconds]
awita has joined #osdev
gxt has joined #osdev
xenos1984 has quit [Ping timeout: 246 seconds]
ZipCPU has quit [Remote host closed the connection]
ZipCPU has joined #osdev
scoobydoo has quit [Ping timeout: 240 seconds]
scoobydoo has joined #osdev
xenos1984 has joined #osdev
weinholt has joined #osdev
dude12312414 has quit [Quit: THE RAM IS TOO DAMN HIGH]
gorgonical has joined #osdev
ThinkT510 has quit [Quit: WeeChat 3.7]
romzx has quit [Quit: Leaving]
ThinkT510 has joined #osdev
awita has quit [Ping timeout: 255 seconds]
awita has joined #osdev
awita has quit [Ping timeout: 252 seconds]
Matt|home has joined #osdev
<gorgonical>
Is there some common programming interface for i2c controllers? It really feels like my manual here is saying "yeah its just i2c" and there's basically no specification of the state machine, what to do... lol
<gorgonical>
But netbsd even has nearly ten i2c master drivers
<clever>
gorgonical: i think every i2c controller does things differently, and that is what the drivers are meant to hide
<clever>
so you just open /dev/i2c-0 and the api is the same for any controller
<gorgonical>
hmm. from the thin description the manual has given I'm not getting sensible behavior. its stuck on start conditions, then stop conditions get sent. can't send stops, etc.
<gorgonical>
clearly I'm doing something wrong but I'm "pretty sure" I'm doing what the manual says. Maybe I'm just missing some... common knowledge here?
<geist>
yeah the only real common thing is there's a general pattern of exposing a set of registers that you read/write to on i2
<geist>
i2c. so it's usually something like write a register number and then either N bytes of data or read back
<clever>
geist: but thats the smbus protocol? a protocol ontop of i2c
<clever>
i2c can also be used without smbus
<geist>
yeah but my experience is a lot of devices are that. dunno if it's smbus or if that's just a usual solution
<clever>
from memory, i think smbus is where you send out the target device addr, write an 8bit reg#, then read or write the data, and then send a stop code
<clever>
but i have seen at least one i2c device, that never responds to reads, and expects just the addr, and then a single massive (up to 128kb) payload written, without any reg#
* geist
nods
<geist>
the key is if there's some sort of external framing so that it knows where the start of a transaction is or not
<clever>
yeah
<clever>
and after each byte, the remote device has to respond with an ack
<clever>
most MCU datasheets explain the wire level encoding
<geist>
also annoyingly some devices dont call it i2c even though it exactly is
<gorgonical>
that's the case with my board
<geist>
because i think philips owns the copyright of the name
<gorgonical>
they call it twi, two-wire interface
<geist>
yep
<gog>
eye squared cee
* geist
pets gog
* gog
prr
<geist>
it's apple update day!
<gog>
update?
<geist>
both ipad os and macos got updates today
<gog>
never
<gorgonical>
i know philips probably doesn't care, but it really amuses me that they use the exact same protocol and are basically saying "no it's not the same protocol. sue us if you think so"
* Reinhilde
plodes
<gorgonical>
gog are the days growing dark up there yet?
<gog>
yeh
<gog>
sunrise is after i wake up for work now
<gorgonical>
eep
<gog>
:(
<gorgonical>
i imagine that's gotta be one of the harder parts of living that far north
<gorgonical>
i mean we're all computer people but still, having so little sunlight
<heat>
apple? bruh
<gog>
i was getting adequate sunlight for the middle and end of the summer at least
<bslsk05>
www.gaby.com.pl: New kids fish pillow toys|Gaby FishPillows
<zid>
get you a 62cm rainbbow trout
<geist>
i guess that's kinda fundamentally what csh is?
<gog>
hmm
<geist>
honestly i dont know, i have really never used csh
<zid>
gdb
<zid>
is my interactive C interp
<gog>
csh is not that
<geist>
yah guess not, it just is more inspired by C but not C per se
<zid>
handy to be able to print structs and to run functions
<gog>
its control flow looks more like c
<geist>
one of my raindy day weekend projects is to write a BASIC interpreter, MBASIC style
<gog>
also yes gdb is cool
<heat>
you could probably use gccjit or llvm to do that
<gog>
i'm writing a compiler too
<j`ey>
me too: cp -r clang joeys-c-compiler
<gog>
yes
<gog>
oh there's already one based on clang/llvm
<gog>
cling
<j`ey>
yeah
<zid>
oh jey is a crypted joey?
<kof123>
re: skeletons and clowns, im 99% sure orion was a mummy and then he raises his hand/club as he swings back around "can you loose the band[age]s of orion?" he's still up there, doing his mummy thing whatever timescale
<j`ey>
zid: crypted? but sure, the ` = o
<zid>
The tickoh cipher.
Burgundy has joined #osdev
<kof123>
i dont know, a plush could be devised
<heat>
yeah i was gonna say you missspelt j`ey
<zid>
g`g
<heat>
you know what the reverse of zid is
<heat>
diz nuts
<heat>
thank you for listening to my ted talk
<zid>
My ted talk is about stuffed sea bream
<gog>
i'm sad
<heat>
why
<heat>
if you're sad u should remember it's zid's bday
<heat>
and zid doesn't want you to be sad
<gog>
oh shit happy birthday zid
<heat>
he wants me to be sad
<zid>
You offered to tortment me
<zid>
and my pizza isn't that great
<zid>
normal ass birthday
<heat>
i offer you the allocator you've been asking for
<heat>
for the low low price of eternal suffering
<zid>
can I have the stuffed alligator from gaby.com.pl
<zid>
actually it's a crocodile :(
<zid>
gog: What colour is the moss you're sending me?
<heat>
are we ignoring that today i've seen like 5 people with bdays today
<heat>
what's with this date in specific
<zid>
pigeonhole principle
<zid>
and january is cold
<heat>
exactly
<heat>
who takes their clothes off
<zid>
not me that's for sure
<gog>
it's brown
<zid>
oh the tasty kind, nice
<heat>
qemu can now expose a storage server with nbd, etc
<heat>
so yay mounting qcow images!
<zid>
exposing yourself gets you arrested btw
<zid>
that's my ted talk
<heat>
is that experience talking
<zid>
heat please you have to attend the ted talk
<kof123>
"C is three frogs in a trenchcoat" makes sense, never expose/define char lest c be arrested
<bslsk05>
docs.google.com: PCID notes - Google Docs
<heat>
oooh kewl
<geist>
got a few bugs here and there, but basically wanted to put all the info in one spot, including the AMD bits
knusbaum has quit [Ping timeout: 250 seconds]
<heat>
limiting live processes to 4096 sounds like it would provide an easy DoS method no?
<geist>
yes
<mjg_>
geist: is that your cat in the profile
<mjg_>
looks hilarious
<geist>
heh yeah
<heat>
the 2nd prettiest of cats
<heat>
right behind my catto
<mjg_>
meme material, for real
eck has quit [Quit: PIRCH98:WIN 95/98/WIN NT:1.0 (build 1.0.1.1190)]
eck has joined #osdev
<heat>
you should probably dump what i linked you into your doc
<heat>
i'm relatively sure linux does similar
knusbaum has joined #osdev
bgs has quit [Remote host closed the connection]
<mrvn>
heat: why would you be limited to 4096 processes?
<zid>
what kind of socks do I need to install linux
<heat>
mrvn, number of ASIDs in x86 PCID
<heat>
(see geist's doc)
<clever>
zid: programmer socks obviously!
<mrvn>
heat: that would be 4095 then
<mrvn>
"This feature provides a limited namespace for process identifiers, 12 bits, supporting 4095 simultaneous IDs total."
<mrvn>
and it's per core.
<heat>
4096
<heat>
1 << 12
<zid>
not sure I can fit 4096 socks onto my feet
<mrvn>
I assume one of them is special or unused
<heat>
doesn't matter that it's per core
<zid>
per foot? fuck
<heat>
to install linux you could probably use sysadmin socks
<heat>
but i'm not an expert
<mrvn>
heat: if you don't migrate (or only migrate non conflicting processes) then you can do 4096 * core
<heat>
until a process uses up more than 1 thread
<heat>
having an ASID per core would provide for a confusing time
<heat>
IMO
<geist>
yeah the usual strategy is to do some sort of dynamic allocation vs statically allocating each ASID/PCID as the process is created/torn down
<mrvn>
heat: you can allocate ASIDs for threads on different cores.
<geist>
currently zircon does the latter, because ARM has 16 bit asids
<geist>
can probably tolerate 4096 PCIDs for now, but will need a dynamic solution eventually
<heat>
having any sort of static solution provides an easy DoS
<geist>
but would like to separate the two into two different things. ie the mechanism first, the dynamic part later
<geist>
yes yes i know
<mrvn>
heat: I could live with 4096 processes
<heat>
your multi-user server could live with one user taking up 4095 processes
<heat>
and the others living with not being able to login :)
<sortie>
4096 pids? I forgot how to count that low
<mrvn>
i said "I". For multi-user you would have to set ulimits and then it gets tight.
<sortie>
I run a million process build every night on my OS :D
<geist>
*simultaneously* it's a little bit tighter
<j`ey>
geist: until someone wants zircon on m1 :P
<geist>
j`ey: right, it's already a problem on M1 actually
<geist>
which is why we will need to solve the dynamic asid/pcid thing *anyway*, for both proceses
<heat>
how so?
<sortie>
Simultaneously's gonna be a harder problem.. hmm.. hey at least I have 64-bit pid_t
<heat>
does m1 have less?
<geist>
but i'd like to have the solution in place for both x86 and arm, then sole the dynamic thing
<j`ey>
8-bit ASIds heat
<heat>
woah
<geist>
heat: yes. AFAIK it's the only ARM core that exists that only has 8 bit asids
<geist>
presumably darwin has a dynamic asid solution so the sw folks were like 'sure, 8 is fine'
<mrvn>
What do you do on a 1024 core / 4 hyperthread processor? If you don't handle PCID pre core you have to use up all just for the idle threads.
<heat>
idle threads dont have address spaces
<mrvn>
You special case them?
<heat>
you also don't have 4096 threaded cpus
<geist>
yes, remember it's PCID per *address space*, not per thread
<heat>
they're kernel threads
<heat>
kernel threads use whatever address space they happen to be on
<geist>
so you have to have more than 4096 active processes before it's an issue. clearly that can be done, but you're talking about fairly large scale machines at that point
<heat>
(when ctx switching)
<geist>
256 though is much easier to hit
<geist>
and 65536 is still fairly exotic large scale machine
<geist>
like i've gotten north of 1500 or so processes on my worksation once or twice when doing a build
<mrvn>
geist: You can have a global uint64_t last_asid = 0; Every time you need an ASID you use "++last_asid" and before switching to user space you check if "last_asid - thread.asid < 256"
<heat>
cuz of goma right?
<j`ey>
goma would have a local thread for every remote one?
<mrvn>
geist: and obviosuly if the thread is going to wake up on the same core it was on before
Matt|home has quit [Quit: Leaving]
<heat>
j`ey, google projects usually have a "pass -j 1000" for the goma googler instructions
<heat>
probably because they can block for IO quite a bit i guess
<j`ey>
heat: oh I guess it makes sense for it to proxy it with one local thread per remote thread too
awita has joined #osdev
awita has quit [Remote host closed the connection]
<mrvn>
geist: you allocate a PCID per address space so it uses the same on all cores?
<heat>
yes
<mrvn>
seems rather wastefull for single threaded processes
<heat>
no?
<heat>
you still allocate 1
<heat>
as you would
<mrvn>
but from a much smaller pool
<heat>
general tooling around pcid/asid is geared towards this
<heat>
tlbi, etc
<heat>
hrmmm
<heat>
freebsd seems to do it per-cpu
<mrvn>
maybe because the earlier CPUs didn't have INVLPGB?
<heat>
i guess you could just issue multiple tlbi/invlpgb
terrorjack has quit [Ping timeout: 240 seconds]
<mrvn>
If you do IPI shootdown then it doesn't matter if the PCID differ
<heat>
tlbi/invlpgb is a lot faster than IPIs I imagine
<mrvn>
probably as much faster as the whole shootdown code the IPI triggers.
<heat>
and triggering an IPI itself, and context switching
epony has quit [Ping timeout: 252 seconds]
<mrvn>
it probably still does a pipeline flush of the other cpu, at least when the PCID is active
epony has joined #osdev
<mrvn>
and memory barrier
<mrvn>
Must be horribly complex in hardware with all the systems you have to sync.
<mrvn>
Imagine 15 cores doing a shootdown at the same time.
terrorjack has joined #osdev
<mrvn>
I wonder if mikrokernel gain more from ASID than they loose by having to allocate a lot of ASIDs for drivers.
<mrvn>
And would you pin driver threads to ASIDs while user processes get dynamic IDs?
<mrvn>
or just LRU instead of round-robin re-use?
<heat>
you'll always gain more than you have to lose
<heat>
microkernels would gain a lot
<heat>
more ctx switching -> more gainz
<mrvn>
heat: at some point you spend more time with cross core invalidations than you gain.
<heat>
you don't have to do cross core invalidations
<mrvn>
worst case: 4097 processes so every context switch has to steal and invalidate
<mrvn>
you have to cross core invalidate if a process has ever run on another core in the past even if it's not currently running
<heat>
no, you'll just make up a fancy generation counter and work on that
<mrvn>
so the other cores invalidate on wakeup? That could work.
<heat>
yes
<mrvn>
would be horribly bouncing around the caches for my IPC mechanism that maps/unmaps pages for messages.
<mrvn>
(the generation counter and core_active bitmap)
<mrvn>
another thing that is just so much simpler without threads.
<heat>
core_active is always needed
<mrvn>
with threads yes
<geist>
re: INVLPGB it's *extremely* new. Zen 3+ only
<geist>
i dont think even linux has support for it
<bslsk05>
www.netmeister.org: Time is an illusion, Unix time doubly so...
<mjg_>
dude not only there are leap seconds, but sometimes the second is intentionally skipped
<mjg_>
that's funny
<mjg_>
that said decent post
gorgonical has left #osdev [Killed buffer]
gorgonical has joined #osdev
<gorgonical>
friggin rcirc screwed up the entire buffer cause someone used an unmatched quote mark for inches
<heat>
>inches
* heat
laughs in rest-of-the-world
<zid>
Inches should never ever be written down
<gorgonical>
i only measure in angstroms
<mjg_>
you know, poland and other metric system countries used ot have their own fucked units
<mjg_>
i never checked how they managed to switch
<heat>
"hey lets switch"
<heat>
"ok"
<gorgonical>
my favorite/worst part of it is that we in america *do* technically use metric, but with this stupid conversion layer on it
<gorgonical>
according to NIST the inch has a formal definition in terms of meters
<geist>
yah and other things like fahrenheit is locked to celcious or at least is now locked to the same underlying thing
<gorgonical>
which is better, i guess? but america gonna america
<gorgonical>
we'll never get polymer money, better coins, or metric units
<geist>
alas.
<heat>
didn't the US have plans of switching to metric?
<gorgonical>
yes i've heard that in the 80s we technically metrified?
<gorgonical>
like officially
<geist>
yah in the 70s but never fully went through
<mjg_>
actual switch for everyday ues?
<gorgonical>
oh the 70s
<geist>
there are odd things that are, like the coke bottling industry, hence 2 Liter bottles of stuff
<gorgonical>
oh yeah, that's true. 16.9fl oz cans are actually 500ml
<mjg_>
also note wtfs like "month day, year" format
<gorgonical>
etc.
<mjg_>
where the year is often skipped
<mjg_>
09/05
<geist>
oh the way date is written is stupid everywhere, honestly
<mjg_>
what is it mofo
<heat>
don't use them commie units
<gorgonical>
only japan does it reasonably, imo
<mjg_>
geist: yyyy-mm-dd for life
<gorgonical>
agreed
<heat>
GODDANG FREEDOM UNITS
<geist>
exactly. that's the only sane way
<heat>
no
<heat>
dd-mm-yyyy
<heat>
i don't care that it's not easily sorteable
<mjg_>
gorgonical: what are japs doing?
<geist>
see, exactly, i find that to be much sillier than yyyy first, since the yyyy sorts properly
<gorgonical>
they use the yyyy-mm-dd format, iirc
* mjg_
moves to japan
<geist>
dd-MMM-yyyy also works, or yyyy-MMM-dd
<heat>
yeah but like, it makes no sense
<geist>
ie, use a 3 letter name
<mjg_>
Gregorian all-numeric date: 2022/10/14
<mjg_>
i take it
<heat>
you usually don't care about the year when checking the date
<geist>
but anyway, of all the things that got standardized globally, at least time did
<gorgonical>
it makes it easiest to sort though
<geist>
date less so, but at least everyone uses babylonian time
<geist>
otherwise that would be real shit
<gorgonical>
yyyymmdd can be naturally lexically sorted
<mjg_>
be happy most of the world agrees what year it is
<mjg_>
even north korea ;S
<\Test_User>
usually, there's a few hours of the world disagreeing
<mjg_>
\Test_User: :) fair
<heat>
i count time in 60ths of a second since jan 1, 1971 00:00:00
<geist>
anyway i remember first time i visited UK i was surprised at how mismatch it is re: metric vis imperial
<geist>
even canada still has some imperial hangons
<geist>
i generally assumed that when a country goes metric they fucking go metric period
<mjg_>
makes you wonder what place is the most sane in this regard
<mjg_>
geist: people are notoroius at sticking to the old shit
<heat>
change is bad
<heat>
old things good, new things bad
scoobydoob has joined #osdev
<mjg_>
in poland in the 90s you got all prices in literally thousands of polish zloty
<mjg_>
fucking lolipop 5000 etc
<geist>
yah i wouldn't *mind* switching to centigrade temp, it's just a matter of trying to be different for different sake is not worth it
<gorgonical>
the uk has been conducting a psy-op to make us think they're metric goodboys
<zid>
human scale units are still nice for throwaway coversation, but I'd never do work in them
<mjg_>
so the government replaces the money, 1000 zl became 1 zl
<geist>
but trying to mix km and mi when all the signs and whatnot are in the other is just too silly to be worth it
<mjg_>
standard "good amount of money" wnet from 1 mln to 100
<klange>
There was an attempt to put km/h signs next to mph signs on roads in the US many decades ago
<mjg_>
but people kept referint to it as 1 millin
<geist>
but i'll happily measure things around the house in mm or cm or m. that works fine for me. makes sense
<zid>
mjg_: I need you to hit up gaby.com.pl for me
<geist>
(though now that i think about it i think all my tape measures are inches/feet)
<zid>
go to their office and snag me a present
<mjg_>
zid: lol, hosted by my former employer
<geist>
but omg having to have two sets of wrenches is annoying as heck
<gorgonical>
i use g on my kitchen scale these days
scoobydoo has quit [Ping timeout: 246 seconds]
<mjg_>
be happy the website works
<heat>
mjg_, so the gov just reverse stock splitted their currency?
<heat>
kinda bizarre
scoobydoob is now known as scoobydoo
<klange>
My American Black & Decker tape measure is dual in + cm
<gorgonical>
much less insane than the base-16 oz/lbs
<mjg_>
heat: ?
<heat>
"so the government replaces the money, 1000 zl became 1 zl"
<mjg_>
nothing special about that
<mjg_>
any country with crazy inflation period should do it at some point
<geist>
gorgonical: yeah totally. oz/cups/pints/etc is the worst of the silly. only real advantage i guess is it's all 2s and 4s and whatot, so you get nice division by common things
<geist>
which is presumably why it was so
<mjg_>
otherwise you end up with 10000000 lolumbian dulars for a lolipip
<mjg_>
lolipop
<geist>
ie, 12 being magic because you can divide it by so many things nicely
<gorgonical>
true
<heat>
any other notable cases of it happening?
<geist>
hence babylonian time with all the 12 and 24 and 60s
<gorgonical>
i have to regularly remind myself of cups-pints-quarts-gallons conversions
<geist>
really makes sense at a human scale
<gorgonical>
heat: turkey did this
<mjg_>
heat: i'm pretty sure germany did that to their currency
<heat>
geist, babylonian numbers were also base 60
<geist>
gorgonical: i literally have a little cheat sheet taped to the inside of one of my kitchen cabinets
<mjg_>
heat: they went through a period where deutche mark was literally worthless
<geist>
heat: yep whcih is why i keep calling it babylonian time
<heat>
i know
<geist>
something like you can count to 12 on one hand with finger segments, and then count to 6 on the other with fingers
<geist>
er 5 on the other
<heat>
post ww1 and all, i just don't know what happened to their currency
<geist>
so you can get to 60 pretty nicely
<mjg_>
heat: > he reform replaced the old money with the new Deutsche Mark at the rate of one new per ten old.
<heat>
i'm not a weimar republic buff
<gorgonical>
geist: a measuring system so easy to use you only continually need a reference. lol
<zid>
I think we should remetrify, but using a base 12 system
<mjg_>
all the imperial unit bullshit is so stupid people literally have to exercise converting
<mjg_>
tlak about a man made problem
<zid>
and make the unit about inch sized not centimeter sized
<gorgonical>
we should take after the noble french
<gorgonical>
and decimalize time
<heat>
ah in the weimar republic they just issued new currency
<zid>
decimal sucks
<zid>
10 is an awful base
<heat>
hey guy
<heat>
how do u count in hex
<geist>
well like i said, there *are* advantages in very specific contexts to some of those units. but most of the contexts are like 'you're standing behind a cart in a mideival town selling cow dung in units of 12'
<gorgonical>
selling cow dung lmao
<geist>
ie you want a dozen or a half dozen? or a pint? etc
<geist>
hence why a lot of those weird units came along that were 2 or 4 or whatnot of the previous one
<heat>
9, 10, a, b, c, d , e, f, 1f
<mjg_>
i don't criticize historical use
<heat>
am i doing it right
<mjg_>
maybe it made sense, maybe it did not
<mjg_>
i would hope most of this was born from practicality
<mjg_>
but for decades now it defo does not
<geist>
yah as usual i like to look at the historical context, and then just assume it's inertia from then on out
<geist>
ie, works well enough, and thus hard to change
<geist>
i always liked the idea of metric time, in units of seconds
<mjg_>
similar with time, while pl uses a 24 hour clock, so to speak
<geist>
alas the stupid planet doesn't agree
<geist>
unless you redefined a second here or there
<mjg_>
people are notorious for saying '4' when they mean 16
<mjg_>
etc.
<zid>
metric time sucks because decimal sucks
<gorgonical>
zid you sound american
<zid>
no, americans all use decimal
<gorgonical>
... ?
<heat>
mjg_, but that's fine lol
<zid>
name an american who doesn't
<geist>
if you could nudge seconds by 15% or so you could nicely arrange for a day to be 100ks
<heat>
you can usually understand if its 4PM or 4AM from context
<gorgonical>
klange knows about this with the jpy conversion. salaries like 5.6 million JPY
<geist>
gorgonical: though i think it's totally arbitrary. it's of course officially defined as some number of wobbles of a cesium atom or whatnot
<zid>
I'd need a couple hundred dollars first
<gorgonical>
lol
FreeFull has joined #osdev
<geist>
but that was done after the fact, to make it line up with something
<heat>
your making me wanna get a physics degree
<zid>
caesium-133 I think
<gorgonical>
if we all agree to buy swatch watches we can all switch to .beats
<zid>
curious marc has an atomc clock he likes to play with
<klange>
yen makes a lot of sense if you consider there is no sub-integer unit - just think of it like cents
<klange>
and this made even more sense when the conversion rate was a beauitful 1:100
<zid>
klange: am I not allowed to think of it in terms of bronze, silver and gold coins, which come in small and large, and are worth 10x the previous?
<geist>
yeah i'm down with monetary units that dont subdivide
<geist>
re: 24h time i've been tring to hard stay in 24h mentally for a long time. pretty good at it
<geist>
but there's always some dumb appliance that doesn't support it. like the microwave
<geist>
keeps reminding you of the other time
<zid>
microwaves always have pathetic firmware, idk why
<heat>
they should switch to uefi
<zid>
UEFI is my 2nd favourite football league
<mjg_>
well if i was to write firmware
<heat>
what's your favourite one
<mjg_>
fucking home appliance is not even on the list
<geist>
yah and you have to make sure you keep Sabbath Mode
<zid>
It's a joke about how uefi is similar to eufa
<mjg_>
maybe someone has a kink
<geist>
fuck 24h time
* mjg_
laugs at 00:29
<zid>
I'd be fine with 12h if it wasn't upside down
<moon-child>
12h is stupid
<heat>
zid, if you like eufa more than uefi you've got your priorities wrong
<heat>
uefa*
<zid>
heat: But eufi is way more corrupt
<heat>
uefi best footie league ye
<zid>
uefi
<klange>
japan ostensibly uses 24h time, and most displays will show it, but then in spoken language everyone still uses 12h an am/pm indicator (午前/午後)
<heat>
mjg_, maybe your home applicance runs netbsd
<zid>
klange: Is that an american thing or did they do that before hand?
<mjg_>
klange: that's poland
<zid>
My understanding of archaic japanese time was to do the long/short hours thing depending on season
<zid>
dunno about am/pm
<heat>
here 00:00 and 12:00 have different designations ("meia-noite" = half a night, "meio-dia" = half a day) in spoken language
<zid>
timezones.gz.pt.patch
<heat>
but then you just say "one" or "four" and people interpret it based on the context
<heat>
you would only use the full 24h if you want to be precise (or someone is asking you to clarify)
<zid>
I use 12 time the same way I use imperial
<zid>
"idk about an inch" or "idk about 4"
<zid>
but not to measure things
<mjg_>
heh, wow
<mjg_>
so poland used to be occupied by 3 different countries at the same time
<heat>
good old times
<mjg_>
they all introduced metric system in different years on the occuopied territories
<mjg_>
... and then it stayed
<mjg_>
lol
<heat>
me and the lads about to partition poland for the 3rd time
<zid>
can I have a bit
<heat>
no
<heat>
prussia, russia and austria
<heat>
only if you hcnage your name to zidia
<zid>
just a LITTLE bit
<mjg_>
the 3rd time alerady happened dawg
<heat>
i know
<heat>
we time traveled
<zid>
mjg_: When commonwealth
<zid>
restore glorious PLC
<heat>
this would be, what, the 5th time?
<heat>
first three were those three-way partitions, 4th was molotov-ribbentrop
<zid>
molotov has the best cocktails
<heat>
partitioning poland is my 2nd favourite past time
<heat>
right after partitioning the balkans
<mjg_>
is this something they teach you in school?
<heat>
and right before partitioning africa with arbitrary straight lines
<mjg_>
honest q
<heat>
what is
<mjg_>
the r-m pact for example
<heat>
yes
<zid>
heat: pastime
<zid>
it's pass + time, not past + time
<mjg_>
afair there is almost 0 about the spanish civil war for example
<zid>
it passes the time
<mjg_>
apart rom the fact it happened
<mjg_>
in .pl
<heat>
well, that's more or less how it goes here for r-m
<zid>
mjg_: Did you walk to gaby.com.pl yet?
<mjg_>
zid: i just got back from there, nice lads
<heat>
i'm just a history buff :P
<gorgonical>
giant fish pillows
<mjg_>
heat: then watch out mofo
<zid>
heat: We should do something nice for a change, and un-carve poland imo, remake PLC
<zid>
They can be the russian buffer state for the EU
<mjg_>
heat: i used to be into ww2, then as i started reading different authors i found contradictory info
<zid>
plc was stronk
<heat>
elective monarchies are cringe
<mjg_>
(readinga nd watching different docs as well)
<heat>
make this shitty piece of land into 5 smaller, shittier pieces of land
<CompanionCube>
how would you even balkanize portugal
<zid>
perfect
<heat>
and give algarve to the UK
<zid>
we all live there anywhere
<zid>
anyway*
<mjg_>
carve out half of portugal for the jewish
<heat>
what
<zid>
"In the 1960s, the Algarve became a popular destination for tourists, mainly from the United Kingdom which still is the origin of the largest group of foreigners in the Algarve."
<heat>
OH
<zid>
see, wikipedia agrees
<zid>
If we should just put down a flag, and we'd own the algarve de facto anyway
<heat>
the UK is the highest exporter of bald, fat, red men
<zid>
with string vests
<CompanionCube>
does that mean the uk also obtains various spanish resort lands
<zid>
and tattoos
<heat>
and CRAZY LADS ON A HOLIDAY IN IBERIA
<zid>
INGERLUND
<heat>
me and the lads on a craaaaaaazy trip in magaluf
<zid>
THE FAK DO U MEAN NO CHIPS
<mjg_>
i had a brit coworker, mofo had such a heavy accent i had difficulty making up words
<heat>
100% liverpool
<mjg_>
first few minutes
<zid>
stop making up words mjg_ that's probably why
<mjg_>
jeremy kyle show
<zid>
I'd bring back corporal punishment for jeremy kyle
<zid>
him specific, not his show
<zid>
ly
<heat>
>The show was based on confrontations in which guests attempt to resolve personal problems, often related to family and romantic relationships, sex and addiction.
<heat>
british dr. phill?
<zid>
springer
<zid>
dr. phil is a conman
<zid>
springer is a showman
<zid>
it's very much more the latter than the former
<zid>
but the show is "poor baiting"
<heat>
this jeremy kyle looks like a shitty piers morgan
<zid>
middle-class white guy calling poor people out for leading shitty lives
<zid>
and letting them debase themselves for trivial amounts of cash as an appearance fee
<heat>
mega rich bastard becomes prime minister without being elected
<heat>
say what
<klange>
Prime ministers aren't generally elected.
<heat>
this would never happen in a western, civilized nation
<mjg_>
zid: one episode was decent
<mjg_>
but it was not his doing
<heat>
klange, if your two predecessors were ousted for being shit you should probably get a new election, just sayin
romzx has joined #osdev
<zid>
mjg_: I'm still waiting for you to recant your stance on making up words to confuse your british workmates
<mjg_>
a woman was talking about how there was a fire and her toddler sister got rescued by some firefigher, except she still got some damage from it
<zid>
heat: I saw a projection that they'd get literally 0 seats, the chance of them calling a GE are.. slim
<mjg_>
and they figured, decaded later, it is time to thank him
<zid>
and king charles isn't going to do shit for a long time that boat rocky
<mjg_>
the guy was found, but refused to show up
<CompanionCube>
heat: tbf uk elections themselves are crap because of fptp
<mjg_>
instead wrote a short letter. 10/10 act in my book
<heat>
i would wager money on the tories still winning
<klange>
heat: I assert that the criticism you wish to levy is "mega rich bastard becomes prime minister without _there_ being _an election_"
<zid>
heat: They have 2 years to try
<mjg_>
so what do you think about nigel farage
<CompanionCube>
heat: not now, perhaps in 2025 enough of the stink will sadly be gone
<mjg_>
or whatever the name
<mjg_>
go ukip!
epony has quit [Ping timeout: 252 seconds]
<zid>
mjg_: Go ask a german what they think of hitler
<CompanionCube>
mjg_: ukip is old news
<heat>
CompanionCube, death taxes and labour losing an election
<mjg_>
CompanionCube: what's the laughting stock now?
<heat>
liz
<CompanionCube>
mjg_: currently it's the tories
<mjg_>
i single out nigel because he was popular in pl as a "calling eu out on bullshit" guy
<mjg_>
... and then it turned out he was lying
<CompanionCube>
but the ukip successor in terms of politics is reform uk, who amazingly enough have some non-bullshit policies
<mjg_>
who would have guessed
<heat>
fun fact nigel farage was originally a typo, he was meant to be called nigel garage
<zid>
tories could break into peoples houses and pee on their christmas presents as their platform and they'd still get votes
<heat>
follow me @fundubiousfacts
<mjg_>
heat: DESTROYED by facts & logic
<mjg_>
only videos i watch on youtube
<zid>
mjg_: When are you coming to claim all your benefits over in the UK? newspapers said you're going to.
<mjg_>
do they? i missed the original wave of pl migrants
<CompanionCube>
zid: that's so 2000s
<zid>
I've got several family members married to polish people
<CompanionCube>
the current scaremongering is refugees on boats
<zid>
and we have polish shops and shit now
<mjg_>
zid: oh noe
<mjg_>
s
<zid>
mjg_: Silly polish factory forgot to bend all the pretzels btw
<zid>
you can let them know
<mjg_>
here is a classic anecdote from anyone migrating to the uk/ireland
<mjg_>
the go to a shop, their english is shit so they have difficulty shoping
<mjg_>
but then turns out they are talking to a pole
<moon-child>
'Back in the 1980s, most of the major CPUs in the world were big-endian, while Intel bucked the trend being little-endian. The reason is that some engineer made a simple optimization back when the 8008 processor was designed for terminals, and because of backwards compatibility, the poor decision continues to plague x86 today.'
<mjg_>
problem solved
<moon-child>
fuck off, little endian = best endian
<zid>
heat: I could double buffer, but idk when to swap the buffers
<zid>
timer would probably do though
<heat>
yeah probably get a 60Hz timer or something
<heat>
or whatever the edid says
<heat>
makes me wonder how they achieve vsync with virtio-gpu
<geist>
i think i have an affinity for unloved cpu architectures or whatnot. part of me really wants to play with one of the intel gpus
<geist>
probably the a750, since it seems like a lot more value
<zid>
I'd try an A770 in my desktop for fun
<zid>
deal with it barely working
<geist>
yah basically
<heat>
yeah
<heat>
i think they work fine but have some driver issues
<zid>
exactly as you'd expect
<heat>
yup
<zid>
nvidia spends a lot of time writing application shims
<heat>
mesa does not and will not do that
<heat>
and I assume their windows intel drivers don't either
<heat>
at least, they haven't *really* needed to
<heat>
being lower powered gpus and all
<zid>
I mean, they need to, it's just nobody complains
<heat>
shipping rewritten shaders is a whole new level
<geist>
yah saw a bunch of reviews on gamersnexus and whatnot. the intel drivers are terribad but hopefully they'll fix things
<heat>
you know, it kinda surprises me given that intel is basically pushing linux graphics forward
<heat>
i wonder if the issues reproduce in linux
<zid>
intel's had.. issues with its drivers
<zid>
they had multiple teams working on competing sets internally
<zid>
one was fast and loose, one was good and slow
<zid>
and they never got anything that was good and fast
<zid>
so the tips for making things work were like "navigate to this random intel ftp and try all these random driver builds and see if it works on one of them" :P
lanodan has quit [Ping timeout: 272 seconds]
<geist>
Oh that’s annoying. Just lost internet at my house
<heat>
sgtm
<heat>
geist, how are u still connected
<zid>
internet at someone else's house
<heat>
even geist steals the neighbors' wifi
<geist>
I was gonna say one of the interesting things is the intel cpus seem to be doing better as you start throwing more pixels at it
<geist>
Ie, it must have great memory bandwidth internally
<geist>
Er s/cpu/gpu
<geist>
Oh I just switched my ipad to cellular
<zid>
internet in verizon's house
<zid>
stealing their weird wifi with the silly protocol
<heat>
you mean google fiber
<zid>
google fiber is a terrible breakfast cereal btw, really stringy
<zid>
and it makes your gums bleed
<heat>
:D
<heat>
makes you poop well though
* mjg_
googles fiber
<mjg_>
my fav food scam concerns gluten-free crap
<geist>
ah came back
<geist>
its raining a lot lately so if there are going to start being problems taking out cables, it'll start about now
<zid>
you had power issues too istr
<mats1>
if you live in a crowded area, hashcatting neighbour wifi is a good investment
<heat>
why does the US seem to have the worst infrastructure
<geist>
yeah but if i lost internet it's almost certain that my neighbor will too
<mats1>
poverty
<geist>
oh it's 100% because i live in the middle of the woods
<mjg_>
better interwebz than north korea
<zid>
Yea his internet gets attacked by trees and stuff
<zid>
all my shit is underground
<geist>
i'm basically at the end of 10s of miles of wires strung between poles with trees everywhere
<mats1>
1337
<geist>
so shit is always falling on the wires
<zid>
The only aboveground cable is the pots
<mats1>
that's not as leet
<heat>
i thought getting attacked by trees was a harry potter thing
<heat>
turns out harry potter was in seattle all along
<geist>
oh the lines are totally underground for the last half mile into the neighborhood, so at least if there's something on a wire its out in the main part of the island
<zid>
american trees fire guns at the lines
<geist>
and not just in my yard or so, which would be harder to get fixed
<mats1>
to be fair, harry potter started it
<geist>
all this aside when i lved in seattle for 5 years before i moved a few years ago i lost power exactly 0 times, and internet went down exactly 1
<mats1>
flew a fuckin car into the tree
<geist>
i was very impressed. and the internet down was some sort of upstream thing. fire at a building i think
<zid>
weirdly it's usually americans with the flying cars