zaquest has quit [Remote host closed the connection]
<heat>
am I the only one that thinks that a compiler that doesn't target -march=x86-64 (the base, generic one) is stupid as all hell and really broken
<heat>
(that doesn't target it by default)
<klys>
it depends, if you mean a pet project compiler, then probably you are the only one who thinks that. if you mean a compiler for production, well, there should be someone. intel is good for project production because it's the industry standard.
<heat>
huh?
<heat>
intel what?
<klys>
I see your point though. someone might build gcc for a target that uses extra features beyond amd64. so, it should default to basic amd64.
elastic_dog has quit [Ping timeout: 272 seconds]
<gog>
intel bad
<klys>
orange pop
gxt has quit [Ping timeout: 268 seconds]
scoobydoo has quit [Ping timeout: 264 seconds]
<moon-child>
heat: seems reasonable
<moon-child>
people who really care about performance will likely -mtune=whatever for their target hardware anyway
<heat>
the problem is
<heat>
that this actively breaks software builds
<heat>
including firmware, kernels, etc
gxt has joined #osdev
scoobydoo has joined #osdev
elastic_dog has joined #osdev
<heat>
apparently some distros that are trying to move into x86-64-v2 are thinking about defaulting their toolchains to that
<heat>
which is fucking bonkers
zaquest has joined #osdev
<heat>
you'll stop being able to reliably package anything
<kazinsal>
wait, hold on, what did I miss? x86-64-v2?
<heat>
yes
<heat>
see the latest x86_64-psABI
<heat>
they added feature levels
<heat>
v2 is like haswell or so
<kazinsal>
interesting
<heat>
we got a patch by a guy at canonical to actively force -march=x86_64 for edk2
<heat>
which makes me scared that canonical and red hat will do this and everything will be fucked and bad
<heat>
now, in practice, most things will still run everywhere because haswell is almost ancient at this point
<heat>
but the concept of the compiler dangerously defaulting to something and the project not controlling it (because who tf sets -march) is scary
<heat>
conceptually, it makes a boatload more sense to default to the generic-est thing possible
<psykose>
afaik even bulldozer works with v2
<heat>
oh woops im wrong
<heat>
v2 is pre-haswell
<heat>
v3 is haswell
<heat>
v4 is avx512 extravaganza
<gamozo>
avx512 <3
<heat>
we should avx512 everything
<heat>
hell
<heat>
avx1024
<heat>
double the bits, double the throttle
<zid>
avx1024, 200MHz <3
<gamozo>
the throttling isn't that bad anymore
<moon-child>
i recently picked up an avx512 chip
<moon-child>
rather cheap on craigslist. Other parts in the mail
<moon-child>
skylake not icelake, but still much better than nothing
<zid>
I always use march
<moon-child>
eh, I prefer april
<zid>
but it's -march=native
<gamozo>
It's absolutely beautiful. The bits don't matter much, it's all about the kmasks
<moon-child>
and compress/expand. and popcnt. and lzcnt
<moon-child>
all the other stuff that was missing before
<zid>
popcnt is part of bmi1 though
<zid>
err, ABM
<moon-child>
zid: vector popcnt
<zid>
ah
<heat>
but now you can popcnt in 512 bits
<moon-child>
and similarly vector lzcnt
<moon-child>
missing before
<heat>
popcunt them all
<moon-child>
(similarly compress/expand was ~available for gpr as pdep/pext; but not for vectors)
<heat>
and all of this is why -v4 should be the default
<heat>
i want to see avx512 instructions everywhere
<moon-child>
oh, and scatter; how could I forget scatter?
<moon-child>
i want to see avx512 cpus everywhere
<zid>
moon-child: send me one, and a mobo, and ddr4
<heat>
ddr5
<heat>
you mean ddr5
<zid>
and a psu
<zid>
and some ddr6
<moon-child>
heat: no avx512 cpu supports ddr5
<heat>
does the i9-12900k not support avx512?
<moon-child>
nop
<heat>
did they ditch it? did they itanium again?
<moon-child>
they ditched it for client cpus
<zid>
does anybody actually use ddr5 on a 12900k
<heat>
why not
<zid>
ddr5 is hard to find still and so are the boards
<zid>
I bet the % isn't very high
<heat>
the steam hardware survey doesn't have ram type
<heat>
shizzle
<heat>
gosh darnit
<zid>
sharnit, gosh dizzle
<zid>
What's ddr5 perf even like, is it a repeat of the ddr4 situation?
<zid>
looks like it isn't? 12900k 'officially' supports 3200 ddr4 and 4800 ddr5
heat has quit [Ping timeout: 272 seconds]
<kazinsal>
yeah, something like that. most workloads are still only a few percent difference between 3200 and 4000 at most
<zid>
Good news is that it's super fucking hard to find ddr4 not in dual channel so ddr3 is still relevent if you have quad :p
<moon-child>
I thought #channels is a function of motherboard (and how many sticks you put in it), not of the sticks themselves?
Likorn has joined #osdev
<psykose>
yes, but ddr4 quad channel motherboards aren't that common i guess
<psykose>
afaik most of the 'workstation' ones are but they cost 78 million dollars
<kazinsal>
yeah for quad ddr4 you're pretty much solely looking at like, TRX40 boards
<geist>
seems like half the time i check a convo after the fact and have somehting to say, heat has already left
terrorjack has joined #osdev
<geist>
was going to point out that x86-64-v2 is somewhat older than haswell. i think -v3 s around haswell
<geist>
-v2 is like, pre AVX but has stuff like cmpxchg16b
srjek|home has quit [Ping timeout: 268 seconds]
Likorn has quit [Quit: WeeChat 3.4.1]
dormito has quit [Ping timeout: 244 seconds]
rindolf has quit [Read error: Connection reset by peer]
rindolf has joined #osdev
eschaton_ is now known as eschaton
xenos1984 has joined #osdev
gildasio has quit [Ping timeout: 268 seconds]
gildasio has joined #osdev
the_lanetly_052 has joined #osdev
theruran has quit [Quit: Connection closed for inactivity]
darkstardevx has quit [Quit: Leaving]
GeDaMo has joined #osdev
arch-angel has joined #osdev
bauen1 has quit [Ping timeout: 272 seconds]
bauen1 has joined #osdev
Burgundy has joined #osdev
dormito has joined #osdev
gxt has quit [Remote host closed the connection]
gxt has joined #osdev
Likorn has joined #osdev
citrons has quit [Ping timeout: 256 seconds]
citrons has joined #osdev
gog has quit [Ping timeout: 272 seconds]
lanodan has joined #osdev
the_lanetly_052_ has joined #osdev
the_lanetly_052 has quit [Ping timeout: 268 seconds]
ethrl has joined #osdev
Oshawott has joined #osdev
archenoth has quit [Ping timeout: 268 seconds]
Likorn has quit [Quit: WeeChat 3.4.1]
Burgundy has left #osdev [#osdev]
gog has joined #osdev
mzxtuelkl has joined #osdev
air has quit [Ping timeout: 246 seconds]
rindolf has quit [Quit: Shlomi Fish ("Rindolf") has left the server. “Chuck Norris was challenged to fight the world, and accepted. He bet on himself, won, and collected the bet money.”]
RAMIII has joined #osdev
blockhead has quit []
<Celelibi>
Hey guys, do you have the name of the OS from a single guy who is like, god-driven or something. I'm sure you'll remember who I'm talking about. ^^
<j`ey>
Terry Davis
<j`ey>
TempleOS
<Celelibi>
Thanks.
<zid>
was, he dead
scoobydoo has quit [Ping timeout: 255 seconds]
<gog>
rest in peace
<zid>
a terry davis zombie would be pretty bad
bauen1 has quit [Remote host closed the connection]
<gog>
yeah nobody wants that
friedy10- has joined #osdev
bauen1 has joined #osdev
blockhead has joined #osdev
bauen1 has quit [Ping timeout: 255 seconds]
ethrl_ has joined #osdev
ethrl has quit [Ping timeout: 246 seconds]
mzxtuelkl_ has joined #osdev
mzxtuelkl has quit [Ping timeout: 240 seconds]
psykose has quit [Remote host closed the connection]
ptrc has quit [Remote host closed the connection]
ptrc has joined #osdev
psykose has joined #osdev
friedy10- is now known as friedy
friedy has left #osdev [#osdev]
friedy has joined #osdev
<friedy>
It's kind of crazy how having a mental illness can make you really creative.
<friedy>
Maybe Mozart was as crazy as Terry Davis.
<GeDaMo>
Maybe they were less likely to fall into fixed patterns
<PapaFrog>
I guess I am doubly cursed.. mentally ill and not creative.
<gog>
i'm in a up-phase but i'm not very creative rn
pretty_dumm_guy has joined #osdev
ethrl_ is now known as ethrl
dude12312414 has joined #osdev
dude12312414 has quit [Client Quit]
theruran has joined #osdev
Goodbye_Vincent has quit [Quit: Ping timeout (120 seconds)]
Goodbye_Vincent has joined #osdev
Goodbye_Vincent has quit [Client Quit]
Goodbye_Vincent has joined #osdev
Goodbye_Vincent has quit [Quit: Ping timeout (120 seconds)]
Goodbye_Vincent has joined #osdev
jafarlihi has joined #osdev
nyah has joined #osdev
<mrvn>
Talking to yourself isn't crazy. Crazy is when you expect a different answer.
<bslsk05>
'Anthem of Europe' by EUofficial (00:02:35)
<vdamewood>
\Test_User: Yep.
<heat>
redundancy is great
<heat>
kinda serves as a checksum
<vdamewood>
Everyone knows that 9:00 AM o'clock inthe morning is a real time.
<mrvn>
UTC still leaves leap seconds
<vdamewood>
but 13: am o'clock at night, not so much.
<heat>
exactly
<heat>
time ECC
<mrvn>
Fun fact: The shape of the earth isn't spherical but such that the time dilation from gravity cancels out the time dilation from the rotation everywhere at sea level.
<mrvn>
vdamewood: what time is 12am?
<heat>
sexy time
SGautam has joined #osdev
knusbaum has quit [Ping timeout: 276 seconds]
knusbaum has joined #osdev
zaquest has quit [Remote host closed the connection]