00:01
<
heat >
does the quantum computer check your research grant?
00:01
<
heat >
your 30 day evaluation period of quantumzip has ended, please enter your research grant
00:01
<
zid` >
it calculates the quantum wave function of it
00:01
<
zid` >
and it has to not collapse to 0
00:01
<
zid` >
if it collapses to 0, all the terms of the expansion have a 0 coefficient and disappear
00:02
<
zid` >
and don't get included in the final sum
00:04
leon has quit [Ping timeout: 244 seconds]
00:22
leon has joined #osdev
00:26
Mutabah has quit [Ping timeout: 276 seconds]
00:29
housemate has quit [Ping timeout: 248 seconds]
00:43
fedaykin has quit [Quit: leaving]
00:47
fedaykin has joined #osdev
01:08
<
heat >
in 2025 onyx will take away the entire piechart btw
01:09
<
nikolar >
the year of the linux desktop is the next year (perpetually)
01:13
Mutabah[ucc] has joined #osdev
01:16
<
cloudowind >
metal collars coming
01:17
housemate has joined #osdev
01:25
bliminse has quit [Ping timeout: 244 seconds]
01:27
bliminse has joined #osdev
01:41
housemate has quit [Read error: Connection reset by peer]
02:16
<
klys >
looking around it appears that recent amdgpu hardware has a cbranch instruction. this is perhaps capable of inferring a conditional branch. is anyone aware of efforts to port gcc to amdgpu?
02:19
<
heat >
llvm does $stuff related to amd and nvidia gpus
02:19
<
heat >
definitely no gcc effort because FSF copyright assignment and GPL and companies get spooked
02:22
<
klys >
can clang be used to compile C into amdgpu instructions?
02:25
<
heat >
yesish i think
02:25
<
heat >
there's a backend for amdgpu instructions of sorts, i dont know what that entails
02:25
<
geist >
try it on godbolt
02:25
<
klys >
how can llvm be compiled to enable that
02:25
<
geist >
it's probably in the existing default list of targets
02:28
<
bslsk05 >
godbolt.org: Compiler Explorer
02:28
<
childlikempress >
conditional branches are 'sort of' a thing since forever
02:28
<
childlikempress >
however amd has weak progress guarantees (or at least used to)
02:28
<
geist >
add `-print-targets` to the clang command line to get the list
02:30
ring0_starr has quit [Remote host closed the connection]
02:30
zijjgfs has joined #osdev
02:31
<
geist >
though i'm not sure that's the current AMD stuff, since they switched off GCN to the new stuff last decade
02:32
<
geist >
which isn't in the list
02:32
<
klys >
so back to llvm
02:34
pabs3 has quit [Quit: Don't rest until all the world is paved in moss and greenery.]
02:34
<
heat >
yeah but is that a different ISA though?
02:34
<
bslsk05 >
chipsandcheese.com: Examining AMD’s RDNA 4 Changes in LLVM - by Chester Lam
02:34
<
geist >
looks like it according to wikipedia i think
02:35
<
geist >
well i guess it doesn't explicitly say it, but i'm guessing it does
02:35
<
bslsk05 >
llvm.org: User Guide for AMDGPU Backend — LLVM 20.0.0git documentation
02:38
<
klys >
and even that page covers rdna 1, 2, and 3.5
02:40
<
geist >
oh cool. so yeah there you go
02:40
<
geist >
RDNA is close enough to GCN that it's covered with the same arch
02:40
<
geist >
makes sense, GCN is when they switched away from VLIW stuff
02:43
<
klys >
when I try to specify -mcpu=gfx12-generic or gfx11-generic then it gives me a version error at godbolt
02:44
<
klys >
...is only available on code object version 6 or better
02:45
<
heat >
yep enough internet for today
02:45
<
geist >
NO STAY WITH US
02:45
<
geist >
write russsssst
02:45
<
heat >
CARGOOOOOOOOOOOOOOOOOOOOOOO
02:46
<
heat >
klys, -mcode-object-version=6
02:46
<
heat >
it's on the page i linked
02:47
<
bslsk05 >
godbolt.org: Compiler Explorer
02:47
pabs3 has joined #osdev
02:47
<
bslsk05 >
godbolt.org: Compiler Explorer
02:48
<
klys >
where the objecgt is to write complex enough code to where it cwapitulates to the desire for a conditional branch
02:49
<
klys >
either that or put one in the inline assembly asm(...)
02:50
<
klys >
in any case where is that syscall
02:50
<
heat >
what syscall
02:51
<
klys >
is my stdlib an empty function
02:51
<
klys >
I put a printf(...) in there
02:51
<
heat >
you're returning early
02:51
<
heat >
gotta delete that line
02:52
<
bslsk05 >
godbolt.org: Compiler Explorer
02:53
<
klys >
s_cbranch_execz on line 51
02:54
<
Mutabah[ucc] >
No return value? that can break codegen (I have no idea of the issue here)
02:55
<
heat >
clang the type of compiler to call a hitman on you if you forget a return
02:55
xenos1984 has quit [Read error: Connection reset by peer]
02:57
<
bslsk05 >
godbolt.org: Compiler Explorer
03:00
heat has quit [Ping timeout: 248 seconds]
03:00
<
CompanionCube >
iirc gcc
*had* a gcn backend at some point, didn't it?
03:01
<
bslsk05 >
gcc.gnu.org: AMD GCN Options (Using the GNU Compiler Collection (GCC))
03:02
<
klys >
you'd need gfx1100 or gfx1103 there
03:04
<
klys >
then look for something like this: 49 s_cbranch_execz .LBB0_4
03:08
edr has quit [Quit: Leaving]
03:10
msv has joined #osdev
03:12
xenos1984 has joined #osdev
03:16
<
geist >
yah gfx1100 is basically current gen 7000 series stuff
03:16
<
geist >
guess that's where that came along huh
03:19
Matt|home has quit [Quit: Matt|home]
03:26
Arthuria has joined #osdev
03:36
<
klys >
> MESA RADV Vulkan Driver Gets AMD’s RDNA 4 “GFX12” GPU Support, Fueled By Valve’s Engineers - 16may2024
03:41
<
klys >
mesa-vulkan-drivers for the deb
04:29
eluks has quit [Remote host closed the connection]
04:30
eluks has joined #osdev
04:41
x64S has quit [Quit: Leaving]
05:24
janemba has joined #osdev
05:48
Arthuria has quit [Ping timeout: 245 seconds]
06:06
Mutabah[ucc] has quit [Ping timeout: 252 seconds]
06:59
goliath has joined #osdev
07:11
heat has joined #osdev
07:55
GeDaMo has joined #osdev
07:56
Mutabah[ucc] has joined #osdev
08:01
pabs3 has quit [Ping timeout: 252 seconds]
08:06
Mutabah has joined #osdev
08:09
ZipCPU_ has joined #osdev
08:10
ZipCPU has quit [Ping timeout: 248 seconds]
08:10
ZipCPU_ is now known as ZipCPU
08:11
Mutabah has quit [Remote host closed the connection]
08:11
Mutabah has joined #osdev
08:27
tjf has quit [Quit: l8r]
08:28
tjf has joined #osdev
08:32
spareproject has joined #osdev
08:47
pabs3 has joined #osdev
09:08
SophiaNya has quit [Remote host closed the connection]
09:08
ptrc has quit [Remote host closed the connection]
09:09
SophiaNya has joined #osdev
09:09
ptrc has joined #osdev
09:40
Etabeta1 has quit [Ping timeout: 252 seconds]
09:40
leon has quit [Ping timeout: 244 seconds]
09:41
leon_ has joined #osdev
09:42
leon_ is now known as leon
09:51
heat_ has joined #osdev
09:52
heat has quit [Read error: Connection reset by peer]
09:59
cloudowind has quit [Ping timeout: 276 seconds]
10:00
cloudowind has joined #osdev
10:23
heat has joined #osdev
10:23
heat_ has quit [Ping timeout: 272 seconds]
10:30
Mutabah[ucc] has quit [Remote host closed the connection]
10:42
vdamewood has joined #osdev
10:58
heat has quit [Read error: Connection reset by peer]
10:59
heat has joined #osdev
11:39
xenos1984 has quit [Ping timeout: 252 seconds]
11:40
heat has quit [Ping timeout: 248 seconds]
11:43
heat has joined #osdev
11:52
<
nikolar >
where can you even find more info on the gpu isas and stuff
11:52
<
zid` >
open directories on nvidia's developer portal was the normal method
11:52
<
zid` >
(not a joke)
11:52
<
GeDaMo >
AMD has some docs, I don't know how comprehensive they are
11:53
<
nikolar >
> geist | makes sense, GCN is when they switched away from VLIW stuff
11:53
<
nikolar >
like i want to know more about this vliw stuff
11:53
<
nikolar >
but i guess the docs are almost unobtainum
11:53
<
heat >
hello nikolar welcome to the year 2000
11:54
<
Ermine >
there are some docs on ati rage 128. Their register reference manual doesn't mention LVDS registers
11:55
<
heat >
gpu docs dont mention things??? shocking
11:56
<
heat >
gotta look at the autogenerated/windows-style 200 line width code in linux then
11:56
<
nikolar >
autogenerated?
11:56
<
Ermine >
heat: I've realised that those registers exist after looking at xf86-video-r128 headers
11:57
<
heat >
lots of amdgpu is just autogenerated gunk from register descriptions, etc
11:57
<
heat >
which is why it has a quadrillion lines of code
11:57
<
Ermine >
(and that was the moment when I've really got what do you mean by "documentation is lacking")
11:57
<
nikolar >
heat: interesting
11:58
<
nikolar >
and we don't have access to those register descriptions, etc
11:58
<
heat >
wtf do you think this is
11:59
<
heat >
you'll take the linux drm dump and you'll be grateful
12:00
<
heat >
#define SDMA_PKT_COPY_TILED_TILED_ADDR_HI_TILED_ADDR_63_32(x) (((x) & SDMA_PKT_COPY_TILED_TILED_ADDR_HI_tiled_addr_63_32_mask) << SDMA_PKT_COPY_TILED_TILED_ADDR_HI_tiled_addr_63_32_shift)
12:00
<
heat >
this is a real amdgpu line
12:00
<
Ermine >
radeon looks kinda the same
12:01
<
heat >
not really sadge
12:01
<
heat >
this shit is real complicated
12:02
<
heat >
several orders of magnitude more complicated than a fuckin nvme drive or something
12:02
<
heat >
they need to share drivers between platforms, share drivers between different GPU generations, etc
12:03
<
heat >
we should be somewhat grateful that amd doesn't roll their own graphics stack with blackjack and hookers
12:03
<
Ermine >
but... people somehow get engaged with drm development
12:04
<
heat >
they get employed or they're autistically dedicated to fix some weird thing in a weird driver or something
12:04
<
heat >
i would bet that the number of people engaged with drm dev as a hobby rounds down to 0%
12:05
<
Ermine >
well, i guess so
12:05
<
Ermine >
but even then some of them are from collabora
12:05
<
heat >
collabora people usually have the knowhow and contacts
12:06
<
Ermine >
is it possible that collabora has an agreement with amd to get actual docs?
12:07
<
Ermine >
well, I guess so, seems like people in the org that was behind xfree86 had the access under nda
12:07
<
heat >
dunno what amd publishes online
12:07
<
heat >
it's possible they publish $stuff and its incomplete, or hard to grasp without knowledge
12:08
<
Ermine >
it's sadge that hobbyist drm dev is basically a lost cause
12:08
<
heat >
gpus are just really really complicated co-computers
12:08
<
heat >
with their own OS and firmware and co-processors and a shit ton of cores, for which you need a compiler for a special ISA, etc etc
12:09
<
heat >
literally full-stack driver development but you're not messing around with react and node.js this time around
12:09
<
heat >
conversely an NVMe driver is just "haha requests go brrrrrrrrr"
12:10
<
heat >
though, interestingly, i've read that some new fancy NICs natively support eBPF programs
12:11
<
Ermine >
(and no way I can get hired by collabora, because... reasons...)
12:12
<
heat >
just be a good russian boy and move to georgia
12:12
<
heat >
intel does gpu stuff in finland i think, arm does gpu stuff in norway
12:12
<
heat >
amd
*might* do gpu stuff in germany
12:13
<
bslsk05 >
careers.amd.com: Working Student GPU Software Development Engineer in Munich, Germany | Advanced Micro Devices, Inc
12:14
stolen has joined #osdev
12:35
guideX_ has joined #osdev
12:37
guideX has quit [Ping timeout: 248 seconds]
12:37
<
bslsk05 >
github.com: linux/init/init_task.c at master · torvalds/linux · GitHub
12:37
guideX_ is now known as guideX
12:38
<
heat >
this is proof linux processes only have 2MB of memory and everything extra is a lie
12:40
<
nikolar >
640kb ought to be enough for anyone after all
12:40
<
heat >
0.01 does mention 640KB as the limit
12:53
edr has joined #osdev
13:00
bauen1 has quit [Ping timeout: 244 seconds]
13:21
heat has quit [Remote host closed the connection]
13:21
heat has joined #osdev
13:31
brennz has joined #osdev
13:42
MiningMarsh has joined #osdev
13:48
vdamewood has quit [Quit: Life beckons]
14:11
bauen1 has joined #osdev
14:21
hwpplayer1 has joined #osdev
14:21
MiningMarsh has joined #osdev
14:23
xenos1984 has joined #osdev
14:23
hwpplayer1 has quit [Read error: Connection reset by peer]
14:23
hwpplayer1 has joined #osdev
14:39
MiningMarsh has joined #osdev
14:41
asarandi has quit [Ping timeout: 252 seconds]
14:41
svm has joined #osdev
14:44
msv has quit [Ping timeout: 265 seconds]
14:45
msv has joined #osdev
14:49
svm has quit [Ping timeout: 272 seconds]
15:03
MiningMarsh has joined #osdev
15:06
Left_Turn has joined #osdev
15:10
_ngn has quit [Remote host closed the connection]
15:11
_ngn has joined #osdev
15:22
asarandi has joined #osdev
15:28
stolen has quit [Quit: Connection closed for inactivity]
15:44
GreaseMonkey has quit [Ping timeout: 260 seconds]
15:51
hwpplayer1 has quit [Remote host closed the connection]
15:51
bauen1 has quit [Ping timeout: 248 seconds]
15:53
bauen1 has joined #osdev
16:10
dude12312414 has joined #osdev
16:10
dude12312414 has quit [Remote host closed the connection]
16:22
stolen has joined #osdev
16:40
hwpplayer1 has joined #osdev
16:54
spareproject has quit [Remote host closed the connection]
17:11
Turn_Left has joined #osdev
17:15
Left_Turn has quit [Ping timeout: 265 seconds]
17:25
zijjgfs has quit [Ping timeout: 276 seconds]
17:26
Arthuria has joined #osdev
17:41
goliath has quit [Quit: SIGSEGV]
17:44
Arthuria has quit [Ping timeout: 260 seconds]
18:02
sbalmos has quit [Quit: WeeChat 4.4.3]
18:03
[Kalisto] has quit [Quit: No' vemo']
18:06
[Kalisto] has joined #osdev
18:11
sbalmos has joined #osdev
18:35
<
geist >
nikolar: i think if you run that godbolt stuff with arch=r600 you might get the previous gen
18:35
<
geist >
looks like llvm has support for pre-GCN
18:38
hwpplayer1 has quit [Ping timeout: 244 seconds]
18:41
goliath has joined #osdev
18:54
chiselfuse has quit [Remote host closed the connection]
18:55
chiselfuse has joined #osdev
19:00
<
heat >
just updated my openssl
19:00
<
heat >
im no lnger vulnerable to 500 CVEs
19:00
<
heat >
(just the 400 in the kernel)
19:01
<
Ermine >
(you get another 100 CVEs over time)
19:05
zijjgfs has joined #osdev
19:06
<
heat >
thats how software works innit
19:25
vdamewood has joined #osdev
19:29
hwpplayer1 has joined #osdev
19:32
<
Ermine >
you get higher rates with openssl
19:33
<
heat >
no actually you get higher rates with linux
19:33
<
bslsk05 >
lists.openwall.net: linux-cve-announce mailing list
19:36
<
AmyMalik >
oh no, my computer grew case morphology
19:44
npc has joined #osdev
19:48
<
Ermine >
you know what i'll say when we're talking about kernel security
19:48
<
zid` >
what's a kernel
20:01
stolen has quit [Quit: Connection closed for inactivity]
20:09
jjuran has quit [Ping timeout: 244 seconds]
20:11
jjuran has joined #osdev
20:13
GreaseMonkey has joined #osdev
20:17
goliath has quit [Quit: SIGSEGV]
20:19
benlyn has joined #osdev
20:30
benlyn has quit [Ping timeout: 244 seconds]
20:40
benlyn has joined #osdev
20:41
karenw has joined #osdev
20:43
cloudowind has quit [Ping timeout: 252 seconds]
20:50
cloudowind has joined #osdev
21:04
npc has quit [Remote host closed the connection]
21:10
GeDaMo has quit [Quit: 0wt 0f v0w3ls.]
21:16
hwpplayer1 has quit [Ping timeout: 272 seconds]
21:28
brennz has quit [Remote host closed the connection]
21:33
<
nikolar >
geist: oh cool thanks
21:57
hwpplayer1 has joined #osdev
22:24
bauen1 has quit [Ping timeout: 260 seconds]
22:31
bauen1 has joined #osdev
22:37
Yoofie646 has quit [Read error: Connection reset by peer]
22:39
bauen1 has quit [Ping timeout: 252 seconds]
22:41
Yoofie6464 has joined #osdev
22:44
netbsduser has joined #osdev
22:45
<
cloudowind >
how lucky i feel , got this helix 2 from ebay , its got intel m5y71 8g 256gb and ive paid only $50 including delivery , came with pro keybord attached and so clean and the both batteries still have %90 life , goodays osdev people
23:01
<
nikolar >
the heck is a helix 2
23:01
mahk has quit [Ping timeout: 252 seconds]
23:01
<
nikolar >
ah makes sense
23:06
hwpplayer1 has quit [Remote host closed the connection]
23:15
bauen1 has joined #osdev
23:18
remexre has quit [Remote host closed the connection]
23:18
karenw has quit [Ping timeout: 248 seconds]
23:20
craigo has joined #osdev
23:36
Turn_Left has quit [Read error: Connection reset by peer]
23:36
netbsduser has quit [Ping timeout: 276 seconds]