#programming
1 messages ยท Page 156 of 1
im pretty sure the cause of that was cuz i put the ++ on the back and not the front
which caused the wrong value to return
The code in question has been removed from existence anyways IIRC

I think I'll be heading to bed, I'd not be surprised if I woke up to either 100k on linux or something beyond 200k. Mentally it is a coin flip.
-# Did I say render 2 frames at once? Whoops, that was a test; the max is the same as Frames in flight because it uses the same thing. So 512 frames 

hey guys I heard that my front panel airflow is flawless, can you rate it?
no I think they're perfect
well i'll be damned
i just found a screw meant for my pc case on the floor
i didnt even realise the case was missing a screw
it goes around the whole computer
fits good enough
y'al gonna make me spend another $200 for a really good binned 13950HX QS istg
doesn treally look like it would do that tbh. in my eyes it goes in and immediatly back out at the other fan
if you only have 2 fans you should do 2 intakes at the front or an intake at the front and an exhaust at the back imo
I was told to remove one and only let it do exhaust
because it removes the hot air inside therefore there's no hot air
well, kinda
exhaust moves hot air from inside out of the case
but intake basicly does the same thing by pushing cold air inside
you generally just want more intake than exhaust so you dont have negative presure in the pc, that way less dust will get inside
got loops working
i hate funny business
about 13.91 MhZ
not accounting for stdout being slow as hell since its an ecall
can you make it say "yeehaw what a fine morning"
yea i was messing with you, i did an oopsie and had airflow thru the back and out the front, it didnt overheat but the problem was there was no intake at the bottom so the gpu was constantly on the verge of exploding
no its impossible
sorry
@tender river __memmove_avx512_unaligned_erms
chat gimme a new specification name for my AWA5.0 continuation, has to start with AWA5.X

5.NEARLY6
i think im done with this for now, mothing left to optemize
the only way to get more fps is to have a 2nd window running on a diffrent core, and to count that 2nd windows fps to the total fps
but thats cheap
im better than that
"nothing left" 
yep I'll go with that
don't want to take away 5.1 for others
i invite you to check out the code and see what you can change to get more fps.
fps = 999999 
no
Well, I get more fps. Just not actual rendering 
thats against the rules
maybe later, I'm too busy with other stuff to deal with the triangle
either way, you're still using Vulkan + normal GPU drivers, so there are likely gains to be had
but I want to maybe improve that in the future so for now I'll keep vagueposting 
it does
you'd be better of just having a while loop with framecount++. cuz that get 6 bilion
#ad - THE NEW G FUEL FORMULA ๐ https://gfuel.ly/pdp-new-formula
๐ Get an exclusive 15% discount on Saily data plans! Use code pewdiepie at checkout. Download Saily app or go to https://saily.com/pewdiepie โต
"i am the problem that i am trying to solve GODDAMNIT" ๐ฃ๏ธ ๐ฅ
damn I'm stuck on this screen

you forgot to set the thread priority
@olive sable 
is windows supposed to do this
no
Did you resize the window?
i didn't do anything ๐ญ
Then there's probably someting wrong with it otherwise
its either broken or an effect of running with the dgpu

cuz on my pc the dgpu runs at 0fps due to driver issues
I sure hope so ๐ญ
im not even allocating memory each frame, so its vulkan doing that
windows doesn't even say what's using the memory lol
vulkan probably
hm
yeah but like in task manager there's nothing next to your program
it is mildly cursed seeing my processor like this
I am already rebooting into nixos lol
ah ok
Not a thing on Windows
On Windows there's no LLVMPIPE
lmao sure
I get 75k FPS with 2 thread llvmpipe
other than that like zero improvement
damn pewds went all out
ASUS Pro WS WRX90E-SAGE SE - 1650 dollars
AMD Threadripper Pro 7975WX - 5200 dollars
Seasonic Primet TX 1300 (x2) - 1100 dollars
Samsung EVO Plus 1TB - 100 dollars
192GB of ram (x2) depending on brand - 1250 dollar
PNY Nvidia RTX 4000 Ada Generation x8 - 30 000 dollars
Total: 39 300 dollars
this is so funny for some reason
Went cheap on the storage? It's tiny compared to the rest of the specs
ye
he said he had to save money somewhere
altho with the 8 gpu's that statement jsut looks ridiculous 
1TB is pain with that many GPUs
You can't possibly store enough data to feed all those
he originally got 6 but tensor parallellism needs a power of 2, so ge bought 2 extra
If just I could afford a Threadripper machine, I could really use one for NeuroSynth training and rendering
and then cuz his mobo only had 7 pcie's had to do pcie bifurcation, but the bios didnt support 8 x 8 so he installed a bios from an internet stranger
damn wtf my WHOLE 256GB ddr4 threadripper 3990x machine only cost as much as how much he spent on the ram alone...
also $180 for my 1600W evga T2 PSU
new.
Anyone got Threadrippers in the price range of consumer CPUs?
pewds gotta take notes from me to find deals fr fr
engineering sample xeons
epyc 7763 is also nice. 64 cores
I have no idea where one would find such
goes for $600 ish nowdays. could even be way lower
I kinda assume the motherboards for those are also really expensive
depends
depends
Milan 
expensive
if you wanted cheaper mobo, TRX40
the RAM can be cheap sometimes
Either way, I'm gonna have to see how my money situation develops, there's 0 chance of me getting a Threadripper for 30
ddr4 ecc are cheap
-# Ignore the fact the Euro symbol doesn't work on remote
servers usually have much more RAM than desktops so it's not rare to find quite a lot of it for cheap as a bundle
uh
Now that's actually insane
Not particularly useful when the CPUs are slower than my current one
lmao
It would just slow down NS rendering even more
but hey for like 17 bucks. in AUD
2017 x 2021
insane deal
but yeah even if its for a secondary machine
17 is insane
plus thats a xeon so in other tasks its gonna be wayyy better than your cpu
props to this lil fella @umbral thorn
they have huge per core performance
chances are it has way denser AVX units
I have no clue how trustworthy that is
could potentially be faster than 12700K in inference
I found a slightly more expensive 24-core Epyc Naples recently too
but it's kinda 
its AVX-512
also it's shitty Zen 1 cores
that whole generation of CPUs is just old enough to be really cheap now
is it the one with consumer socket
How does one even determine if something on a store like that can be trusted
Do they even ship to Finland for reasonable prices?
no
7401P
the 12700k has no official support for avx
7002 is Rome
I have never used AliExpress
*AVX512
It does support some other forms of AVX
ye, Rome is still kinda expensive
ok fair enough
right, 6138 also supports like 6 channel 2666MT/s
tf is this
my current server CPU is Rome but it's a pain to get a good upgrade because of that 
numbers
128GB/s bandwidth
That's crazy
Still, I don't know how one knows if a thing on AliExpress can be trusted
yeah chances are the xeon 6138 might outperform 12700K a bit with AVX?
incrementing an integer every iteration and printing it every second
just find reputable stores
there are a lot
I don't know how to determine that
The question is if NS's renderer supports AVX
the only thing that you CAN'T ABSOLUTELY TRUST is cheap drives that are other than SATA/nvme/HDD even if it has high ratings
reviews, sales amounts, followers
are they named Store83628294 or An Actual Name
people are so easy to fool when the scammer edited the usb drives firmware to show it has huge capacity, but in reality it only has like 1GB
gets high rating because people "saw" the capacity "as advertised"
then have their data corrupted
I'm still extremely new to online shopping, myself I've only bought a single power supply from a Finnish retailer
worst part is that it never even seem like the data is corrupted because the data table is intact
i mean first question is "what does the renderer run use"
๐ญ
Thus why no GPU rendering on Linux, it depends on DX APIs
your best bet is to actually dual boot...
ONNX does not depend on DirectX
don't just stick to one os just because you feel like it and then you lose a benefit
idk where you got that from but it runs on just about everything
the ms runtime might
then according to this it can depending on the format, so I can imagine it would also do it elsewhere
And wy would I do that when my Windows install is literally unusable?
performance?
do you still have your 4TB SSD
I can always try if you're interested but I have no way of checking that its using avx512
Using it
what about your 1TB?
the NeuroSynth runtime is his creation, so he should fix it if it's a problem
if he's using a Microsoft product somewhere in there then it needs to be kicked out
but that would be stupid because cpu inference is a thing people want
Legacy files
move it some where
then reinstall windows on that
you do not need to whine about having to get a better cpu dude
do not dual-boot, it's annoying
when the solution is just that
Was not planning to, especially with my stuff being a server
just either:
- fix NeuroSynth to work without DirectX (it already should if it's using ONNX???)
- use a Windows VM and pass through your 3090

I have to manually boot into bios and go windows manager again in order to get into my other windows
I honestly have no idea why the renderer doesn't give the option to select to run on GPU on Linux if Microsoft ONNX runtime for C# doesn't depend on DirectX
I just assumed that's what the DirectML renderer backend option being gone meant
directly selecting it in windows manager with shift + restart will lead me to bsod
how tf did y'all struggle with dual booting
all i have to do is select the boot override
Are you aware the Windows locks NTFS file systems if it is not shut down properly, causing issues after booting to Linux?
every bios has a dedicated boot menu that you can directly choose the boot device without entering the bios
it's not my issue, I probably killed the efi while optimizing my other windows
and you can have it always automaticly come up
Currently I'm mostly surviving because NS is an around 100M param model only and the tensor cache kinda carries
But the initial render takes like 10 minutes when rendering a new full note part
idk what caibi is doing either, you use either the EFI boot manager or a Linux boot manager with Windows as an entry
it's not annoying because of the boot manager thing, it's annoying because you have to reboot and lose your session in the first place
it's the opposite of seamless
True, I hate rebooting and losing the session
Way too much of an impact for just NS
fair
You can see I am once again building up some uptime
At least Linux Mint doesn't become sluggish after a while
I assume ONNX for C# is just bindings to the normal ONNX runtime
so there shouldn't be anything requiring DirectML specifically, ONNX supports CUDA too
it might just not be built with support for it or something
good luck debugging 
I'll show you what I mean
hold
I guess I may have to consider seeing if I can enable the CUDA backend and compile myself
Anyway, gotta go now, disconnecting remote
dude
i just realized juniper decided to refund me 10% for the lava lamp and i felt so fucking bad ๐ญ
they can keep the damn 10% i can wait for the lava lamps
maybe I broke the efi section in another windows installation
i really appreciated their effort with improving the product, i genuinely understand them
and somehow I can launch it in another way
just keep the damn money dude y'all engineers worked really good with improving the lava lamp for months
same, their only mistake was providing a bad timeline to buyers
the product development itself seems to be going well
and i really just wanna them keep the 10% ๐ญ
hell i'd probably help fund them for their effort
there isn't even a donation page
I mean the lavalamp oversold, probably why they would make this decision
you still shouldn't have to manually select the boot entry in the EFI since there's only one
idk how the Windows boot manager works but it seems to be taking care of the rest somehow
no they can treat themself with the money with how they put so much effort into it
damn it they don't even have a donation page
prop to the team making the bootloader 
no prop to the team making the os select page
should i email juniper with a special deal
i buy their "engineering samples"
for the lava lamp
So, I have a small issue here. Brief introduction: I tried to unpack .pck of assets of a godot game. It was locked, I assumed. I opened it in HxD and searched for hexadecimal encoding key. While I looked through this crap I accidentally changed one singular bit of code. i realized that I am an idiot shortly after and the unpacking error that godotpcktool was sending me was caused by me specifying the target file incorrectly. I unpacked the assets, but the game launch is looking like this now. Verifying integrity of files failed. Uninstalling the game and reinstalling it failed. Help.
Additional context for uninstall-reinstall error
directly go to the directory and nuke the content
Roger, nukes launched
...It says that I need admin access to kill it. I am logged in as an admin. weird
ok, it worked even tho the folder for unpackaging is just undeletable now
Elvyn, since you're so good at finding cheap stuff, would you find me a cheap CPU too?
go in the folder and delete every subfolder one by one
delete everything till it tells you whats preventing you from deleting that folder
I could do that, but the game works, so I will just leave it there as a constant reminder that I am an idiot
but it makes everything so easy to read 
this will probably never compile outside of clang

Certified clanger
im refactoring the vm
perhaps performacne improvements in here too
gonna make the fastest hbvm impl
watch
While I am here, I have a question for godot enjoyers. Lets say I have unpacked assets of a godot game, right? And lets say I just want to look through different pngs of moves, chars, animations, right? And everything is in acursed .import format. Like move1.png.import. Is there a way for me to see the pngs WITHOUT installing godot and learning how to use it and then reverse engineer the game by making a new project and just shoving the assets there to look at them?
fun fact
every godot game has a copy of the godot engine in it
just uh
hack it or something
i thought these were already as cheap as possible, tutel will be activly losing money by selling these at 90% of original price
Godot is a tiny self-contained executable, it's literally so tiny you could call it a spec of dust compared to a worse engine like Unity
tutel is always losing money that's his brand

fair
i remember him saying he didnt have a lot of money, followed by him wanting to spend 100K on a concert
128MB, single file
That's the entire engine
yeah he's cooked
development time was longer than planned too ๐
I have these weird paths in here like these. Can I like tinker with those in a way that would lead me directly to pngs? As I understood those are like virtual "fake" paths that godot uses for its own file navigation, but those MUST be somewhere on my PC, right? Can I like bypass those?
The files are contained within the game executable
Or the game package file
Depending on if the game is self-contained or separated
It was in pck, pck was unpackaged into a folder, but they are in .import and I can look at them cause nothing opens .import. Alternatively I have those that end in .png
Sooo uuhhh
yeah
You either did something wrong, or you just need to open the import files in a text editor to check what they contain
ooh, great idea
.import shows this
Fan fact there is no character called alien in the game
There is this path to png that starts with res://. How to find it. I want to see the alien
oh, stex, not png
My bad
there is so much comedy gold in the files tbh
you know what, what's your budget
I currently own 30โฌ
nvm
Will get another 60 next month
if you have at least like 170โฌ i could've send you the link to where i got my cpu from
That's crazy
That's gonna take me about 3 months to get
Assuming no other income happens
60 euros per month???
is that an income
or an allowance
no way actual income is that low
It's a state benefit for studying
that's... an allowance
If your uni/collage/whatever is not privately owned it has contracts for replacing hardware every 2-3 years. If you are lucky and have good relationships with administration you can grab whatever they are gonna throw away before they do
I got a ram stick like this
Well, I've got nothing better, I failed to get any job even though I tried for the summer
I got rejected despite my abilities
That's what I assumed I was trying for
I was looking for a summer job to be specific
what job btw
It was something from the company that manages the computer systems for the school, seemed very much like a job made for me
I'm usually the one teachers go to before the actual IT team since I'm just so much faster
Well, summer jobs are not really about self-actualization and development of skills and talents, they are about making as much money as possible as quick as possible. My go-to's are construction sites and harvester boats. 60 euros can be made in like a day if you are lucky. And its where I live, so the buying power of those 60 euros is around 150
No skills required, no interview and no application. Come around next day and work. Thats it
There is the obvious issue that my physical qualities are not really good for anything physically intensive
Computer related things are good such that physically intensive tasks are not usually needed
well, um
first, that is definitely not entry level
second, tech related jobs usually require actual qualification
I was pretty sure it was entry level, the listed requirements were quite lenient
Do you have a condition or are you just lack confidence? Cause if 2 option - dont worry, if you are a healthy man with no asthma or heart condition you'll be fine
Either way I'm probably better than their entire actual IT department
I fixed in 5 minutes an issue the actual IT team was fighting for a way longer time
damn, in australia doing just ubereats for 4 hours can give you around $120-$250 AUD depending on how fast and lucky you are lol
I don't know, maybe it's a partial lack of confidence on my part
whenever i don't have any more research work that's what i do for extra cash lol
converted to euros that's 66-138
IT department are just for support tbh
That's actually insane
That's more than I get in a full month
Well, they certainly seem to suck at their job considering how badly broken some of the stuff is
They can't even run a simple cable because they say "it's too long" while I've seen cables wayyyy longer than that
they're just providing services for a timeframe, its different compared to your occasional intervenes with fixing stuff
Average salary here is like 500 euro per month, so not bad
you should probably look into how modding for godot games work. you can just use the asset extract utility to get the whole project out containing all the normal png's everywhere. its very easy and straight forward you don't even need godot installed for it just the godot decomp tool
if you're describing that you're better than the whole IT department, i can also say that the majority of my colleagues are better than your schools IT department
that bar is kind of low for basic troubleshooting, perhaps the IT department are supposedly specialized and not for general troubleshooting?
Who actually knows, all everyone at the school knows is that I do way better than them
And still, best I can get is a bit of chocolate I didn't even like
that's what my school also think when all i did is switching input to hdmi in a projector, tbf
Silly
Reminds me of when I grabbed the blue ray player and plugged it into the big screen and then swapped the big screen to use HDMI 1.1 instead of 2.0 to get the device to display on the big screen rather than a tiny TV
I searched for this very specific tool and it was missing from the git. Like 404 error. If you have a relevant link I would be very thankful
Well, either way, I'm just hoping somehow I can utilize my skills at programming or vocal synthesis or anything like that I'm capable of to eventually achieve even a little more income
Trying to save up for a PC at 60โฌ/month is a massive pain
At least I'm still finding out new stuff at programming, I just found the best GUI framework ever
you can try freelancing
I have no clue how that works
you can just search for it tbh
Oh, are you in English speaking country or not?
Finland
Cause if not you can tutor kids
I have 10 kids that I tutor and it makes like 80% of my income
Freelancing? I guess
I heard that a lot of people know English very well in Finland so this is probably not an option for you, sadly
Yeah, Finland has a really high average English skill
Though mine is still above average
What an interesting term
So it essentially means working without a stable work contract or something?
What matters is not the level, but confidence. My English is ass and I learned it by watching youtube, but when I meet parents of a new potential student I just lie a lot and make myself seem like an expert and they believe it most of the time. Not like I am a complete fraud, I still teach like 10 times better than in schools
Schools are honestly pretty bad for learning languages
idk what im doin but wish me luck
Hope you got a good GPU
i think its decent?
What GPU?
To try everything Brilliant has to offerโfreeโfor a full 30 days, visit https://brilliant.org/mattbatwings Youโll also get 20% off an annual premium subscription.
Patreon: https://www.patreon.com/mattbatwings
Discord: https://discord.gg/V5KFaF63mV
My socials: https://linktr.ee/mattbatwings
My texture pack: https://modrinth.com/resourcepac...
reject python, embrace redstone
Kinda an issue
8GB is tiny for making an LLM
Well, enough for a very tiny model that barely works
I guess if you just want to learn how to do it it's fine
are we talking about pc hardware
my gpu memory is 24gb?
3090?
2080
what

are you even sure there no way the 2080 would have that much vram wait

look at dedicated gpu memory
show me
That's dedicated + shared
they're very different
Dedicated is 8GB, shared is 16GB
what
And shared is very, very, verrryyyyy sloww
i see 
You essentially never want to be hitting shared or your training run will take 30 years
Yes, mainly VRAM
wait isnt that like the thing that intel uhd graphics use
Yes
bruh
i am running on a intel graphics 5000
godot terminal broke cuz it started to cry over my device specs๐ญ
the engine couldn't find vulkan so it switched to open gl 
why is installing cuda such a pain in the arse
OS?
perf stat comparison
almost certainly the fastest hbvm impl now
where is my box 
in distro
UH

ntainer 3f6139c59244428cd6e4970bb4b6fb26e34d62fb7939b8f9f702910173844334 storage: cleaning up container 3f6139c59244428cd6e4970bb4b6fb26e34d62fb7939b8f9f702910173844334 storage: unmounting container 3f6139c59244428cd6e4970bb4b6fb26e34d62fb7939b8f9f702910173844334 root filesystem: replacing mount point "/home/sqbika/.local/share/containers/storage/overlay/26092c096fcb3a52fb34e25c725d55cd0ba6a38a25a2e33c6851d72bb4b800f3/merged": device or resource busy

do you like cursed c
the one you sent above? i saw that
there are still things i can do to make it clearer and/or faster
probably changing a few ops to be branchless so that the cpu branch predictor can predict next-instructions better
just don't overdo it with the macros
(this is fine)
i will 
cuda has been installed
but no nvcc for some reason
If you didn't get nvcc something didn't go quite right
What OS?
fedora 42
Well, I guess who knows what's wrong with nvcc then
Try to manually find it and add it to the PATH
I did previously have various packages conflicts while attempting to install cuda 13 so I did --allowerasing to let it install the packages it wanted
Why CUDA 13? It might have compatibility issues
so should I get a lower one?
which version do you recommend
I personally have 12.8
It's fine
it's an rtx 2050 (ampere ga107) so probably good
just a sawed off 3050
nvcc is there apparently
not in path for some reason
I had the same issue when installing mine, I just added it manually and then the app I wanted to run worked
12.8 being the latest versions supported by it so that's what I installed
ayy it works now 

i'm waiting for them to add cuda 13 to nixpkgs 

cuz I got a new mic
and I can finally scream russian slurs

why wait when you can build from source 
you could just become one
looks like they don't even have a PR yet 
tbf it looks like CUDA 13 is only a few weeks old
ROCm updates take months
yeah it's been 3 weeks or so
post your problem here
dang and fedora has it already? lessgo
CUDA should be faster, it looks like a lot less of a mess
you just have to download a couple files from nvidia for the most part and merge them
and i guess apply a couple patches (mostly for glibc)
been getting an error type of 'bool' object is not subscriptable
def calc(d_calc:int) -> int:
d_calc=int()
cons_kw=str()
for x in range(1,4):
try:
cons_act={1:12345,2:2344,3:2144}.get(d_calc,'p_kw').isalnum()
cons_pre={1:2112,2:1600,3:1221}.get(d_calc,'left').isalnum()
except ValueError as e:
print(f"{e}")
cons_kw=cons_act[x] - cons_pre[x]
print('Left num',x,'KW consumed=',cons_kw)
try:
sum_kw=lambda:d_calc['p_kw']*cons_kw+d_calc['left']
except ValueError as e:
raise(f"NONE EXISTING NONE TYPE VALUE {e}")
tva=sum_kw()
sum_net=sum_kw+tva
print('Left value of num',x,'Net paying in $',round(sum_net,2))
print(24*'','Net to pay in dollars',round(sum_net*cons_act['t_d'],2))
print("________________________________________")
try:
total=float(input('enter money '))
left=float(input("the sum of leftovers="))
price_kw=float(input("the price in KW "))
d_calc={'t_d':cons_act,'left':left,'p_kw':price_kw}
except Exception as e:
print(f"return wait wtf -->{e}<-- look at the error stupid ass")
calc(d_calc=12)
line please
line 10
lemme look real quick
isalnum returns a boolean
line 10, in calccons_kw=cons_act[x] - cons_pre[x]
dang you already got it
i am trying to change the str index to be able to parse it to a int
didn't even let me type
and i think i over wrote the dict
well i removed the isalnum and now i got this unsupported operand type(s) for -: 'str' and 'str'

wait 2 i am try something
ok you removed all isalnum right
change line 10 to use actual dict cons_kw=cons_act - cons_pre
change line 13 to fix lambda to use proper variables
sum_kw=lambda: d_calc['p_kw']*cons_kw+d_calc['left']
oh i see
here
even tho i remmove it it would still need an index
hm
try putting line 6: cons_act={1:12345,2:2344,3:2144}
line 7: cons_pre={1:2112,2:1600,3:1221}
line 10: cons_kw=cons_act[d_calc] - cons_pre[d_calc] (use d_calc as index now)
line 13: sum_kw=lambda: d_calc['p_kw']*cons_kw+d_calc['left']
@uneven pulsar
is it good or we still fucked
bro your long ass replies make me question my sanity and my programming skills legitimacy
hold pn
dude wait
yeah
TypeError: unsupported operand type(s) for -: 'str' and 'str'
fuck
wait
why is d_calc 12
that shit aint a key in yo dict
so now .get() returns the default value p_kw
6: cons_act={1:12345,2:2344,3:2144}.get(d_calc, 0)
7: cons_pre={1:2112,2:1600,3:1221}.get(d_calc, 0)
here i set default value to a number
alr wait
@rare bridge
why yu ping mod lah
'int' object is not subscriptable
ne
Triangle
is that the error output
oh ok
d_calc is int but you trynna use it like a dict with d_calc['p_kw']
And why you may be doing that may I ask?
gta v on phone type shi
I can't carry a pc like a laptop
So option 1 cloud
2 use the emulator in the phone
3) I'm not dumb that phone can't run games
which app
Wine
yeah but what in wine
lmao cloudburst joining triangle aswell
is this some new 6 figure fps cult
Everyone triangling
i am going to eat please fix your code by not doing that
how do you miss that shit
no actually, how did I miss that shit
!triangleaddwindw Cloudburst,7881,Tiny 11,Ryzen 7 5800XT,RTX 3060,2x16GB DDR4 2933MT/s
Added entry "Cloudburst | 7881 | Tiny 11 | Ryzen 7 5800XT | RTX 3060 | 2x16GB DDR4 2933MT/s | "
I successfully ran first app
What about Russia rollet ?
python mm.py
Enter the number of players: 1
Enter name for Player 1: lo
Enter the number of chambers in the revolver (e.g., 6): 6
Enter the number of bullets to load: 5
Traceback (most recent call last):
File "/python/mm.py", line 87, in <module>
play_game()
File "/python/mm.py", line 62, in play_game
game = RussianRoulette(chambers=num_chambers, bullets=num_bullets)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/python/mm.py", line 18, in init
self._load_bullets()
File "/python/mm.py", line 23, in _load_bullets
loaded_chambers = random.sample(range(self.chambers), self.bullets)
^^^^^^^^^^^^^^^^^^^^
TypeError: 'list' object cannot be interpreted as an integer
!triangleaddwindw Redston,36819,Windows 11,Ryzen 7 7800X3D,RX 7900 XTX, 2x16GB DDR5-6400 6400MT/s
Added entry "Redston | 36819 | Windows 11 | Ryzen 7 7800X3D | RX 7900 XTX | 2x16GB DDR5-6400 6400MT/s | "
this guy actually had a spike of 36K on windows and managed to took a screenshot of it
so i'd consider that valid score ngl
!trianglepush
loaded_chambers = random.sample(range(len(self.chambers)), self.bullets)
its ok i left you with your main problem
first time in months im playing a game, the programmable mouse button on my g502 mouse deleted all the of the UI in besieged 
i was making a car
Sad but how
idk

Sometimes extra mouse buttons have default interactions with games, such as with Factorio pick object and open inventory
This custom gaming pc has $6000 worth of technology . An RTX 4090 , i9-13900KF , 8TB NVMe , 128GB DDR5 , 2 14TB HDD's , and an overheating issue! Which is why it's in my computer repair shop. Because the 240mm AIO can't cool this under stres, undervolting the CPU was the only option until a larger AIO can be purchased. This tech runs hot and nee...
huh
makes you wonder if intel purposefully made 13th and 14th heats up so much
because my engineering sample runs pretty cool on a sff cooler
This guy can't update bios update
Bye
dude is a huge pc nerd
THE GREATEST technician THAT EVER LIVED
โ
erms = rep movsb
or rather a cpuid flag that says "rep movsb is fast"
nice macros
i didn't even know gcc and clang supported expression blocks like that
better ergonomics than zig
surely this counts as mechatronics 
but ye i cant figure out how to make the steering not colapse in on itself
strut and sway bars
the problem is that is kinda too detailed for a blocky game like this
Besiege 
fun game
like idk how id make this in besieged
really big
intuitive. I've never played that game, but if that function works for stearing irl, try to mimic in game.
there is a build size limit of x blocks wide, but i think you can turn those off with mods
looks like your steering functions on your vehical are sep, from the frame. can you connect? For more stability across the whole structure then?
i got the game for free on amazon prime a couple days ago 
you can remove the build area limit without mods too I think
mods allow some fun stuff though because you can freely transform parts
looks good for 1 hour tbh
well i used to make cars with decent steering in lego so i have some experience
try to frame the steering along with the rest of the vehical. This way, it wont turn the whole steering rather than just the wheels.

the 'bar' above the wheel assembly and axles are free floating almost. If you can see how to make that more rigid or a way to even combine the top bar above the wheels with the bumper?.
that way the axle is not independent from the frame
when you turn, it pulls the hole assembly with it. Reduce that with strut or support from more of the frame
^-^ i wish you success
thanks 
i was going for a multi-link suspension type thing cuz i cant put the pivot point inside the wheels, but i think its too complex for besieged.
NONSENSE
keep going with it
I'd screen shot what I mean, but I'm new, and I cant
The 'bar' you have that steers the wheels, can you add one to the front of the assembly? Connect to bumper?
Just strut support
okay
i want you to win. Sorry. Invested
also i have discovered i misremember multi-link suspension
you're supposed to put the double arms at the top
Can you add more to the rear of the assemb.?
Your competition is yourself. If you can do better than what you did before, you win
keep winning

nah, surely this is mechatronics 
very interesting how the arms have joints with non-parallel axes
^ sway and strut support
awa
6
7
1
awa
1059631593301491744
neuroAYAYA
try and beat the chair first 
-# does that still count as a car
the WHAT
chair
Trying to map out a "Brain" for my ai "Aeon"
Hemispherical brain. With Hippocampus that pulls categorized tags from Mapped Vectors like "tech, personal, values, codebase, facts, and so on"
Left: analytic, stuctures, checkable. Logic code math.
Right would be creative, and so on.
Social would be autobiographical memory, mood, persona.
Control and filters
basal ganglia gate: Quality checks. Confidence filter based on set mood. (mood alters and shifts)
Acc monitor, (in the gate) detects contradiction between hemisphers and triggers a short, "Why are we disagreeing" pass.
Corpus Callosum.py. Merges Proposals, with per-categorised weight and confidence
Memeory
Hippcampus pulls category tagged vectors. "Tech, personal, values so on"
Hemi know which category fed them.
Self adaptation.
Sleepmode, or "Study time" Consolidates chat, into new vectors, prunes junk, promotes recurring facts to "Semanic long term" and tunes model prompts.
My competition rn.
idk what catigories to islolate all the different pulled vectors tho. I dont wanna just Ingest into a single "json or Faiss" but seperate in memory.
Also take the hallucinations and allow them to be 'dreamed" about. So i wanna add a day dream that she pulls from the hallucinations
I'm thinking of categorizing the vectors in moods. Maybe. Idk.
I'd eventually use a model for each hemi
<reasoning>
The corpus callosum is a bundle of nerve fibers that allow your brainโs left and right hemispheres to communicate. It plays a role in how you think, remember and coordinate your movements. Conditions present at birth and acquired conditions may affect the development and function of this part of your brain.
Like: If you took Evil and Neuro and made them think together to form 1, combined answer.
but the Hippocampus would regulate resopnse based on many things
a mood filter
"Hey aeon, would you like to talk about black holes today?"
aeon> "Nah, lets talk about cheese, cause I wanna"
I decided to try to learn ai and shit
I have now regretted it after an hour and a half
Brilliant . org is a good starting point for learning
I haven't even gotten to code yet :,v
Just math and concepts, and my head is hurting already
And I haven't even done any math :,v
I suggest that learning website. It costs.
Then get notepad++ or something that will help keep scripting clean and easy.
I'm still learning rn.
And Remember Big Important quote from an old chinese man.
"Knowledge is a treasure, but practice is the key to it"
Xd
accomplishments feel better when you accomplish something.
Go forth and accomplish something
I'll accomplish insanity at this rate๐ค๐ค๐ค
^-^ I use insanity as a tool to do better
or work harder
cause why would anyone try to code a working brain on his computer
Motivation, courage = insanity. People just dont like insane people very much. They have no real limits of achieving what they aim for
just dont hurt anyone
lol
you think Vedal is 100% sane?
_>
Bro I am learning as well and the first thing I see is someone trying to genuinely recreate the human brain
^-^
If you need help with certain brain functions I have books on that to help. You seem to have a good idea of what all the different brain sections do though
I studied that years ago. Brush ups would be nice.
Just trying to merge neurons and code.
Yeah, that is pretty much why I want to learn coding
and its not Genuienly a brain, more like filters to re think a few times based on what 'thoughts' it has now, or retained over vectors.
If mad, dont respond 'Hi. There Buddy!" But, "Fk off please"
Am i mad- Why am I mad- What am I thinking about that would make me mad- Give answer to random question
also, triggers to change mood
I wish I could share my Mood.py
Categorize Vector dumps and pulls to know what can trigger 'moods'
Tbh my roles in this server are straight lies. I picked musician because I would be willing to learn that stuff properly and I picked programing because I want to see others talk about coding so I can learn and get ideas. I picked jp despite the fact I am in the learning phase
science=curious
I picked art cause I made the most money doing that
therefore must define me... right?
That is a field I know I have no place in. I can not picture an image before I start so it always ends poorly for me
why beat yourself up? Have you practiced at all?
Better then me man, my whole roles list is either I like it enough to learn it or learning it or starting/planning on learning it
I have an idea of something to try and sketch, just haven't had the time to do it
^-^ i'm sure when you start practicing, you will start thinking you're more of an artist.
google "Still life"
start there
Its litterally something to draw
If I do start trying to draw I would rather just start on something of my own thought, I know it will make it harder, but at least I can say it is of my own creation
Ballance on 1 foot. You get wobbly. Draw a line with your hand. You draw wobbly. Practice, feel the line become smoother. Build more control over functions.
with practice, this becomes easier. Build confidence. NEVER COMPAIR cause you will beat yourself up for not being someone else. You are You. remember you're art will NEVER be someone else. Only your own
thats about it
^-^
Trust the process.
(oh, i suck at spelling.)
Then that would be part of my problem with that, I often tend to use something or someone else's works as a base for my own.
no. Thats still "Still life"
I mean, if your results differ from someone's dont feel bad.
Back to your project because since it involves the brain, it entices me
^-^ I'm still learning how to make it. Researching and testing ideas.
kinda. Still playing with it. still cant share files. I'm working on that
I can paste a portion >_>
ROOT = Path(r"D:\Aeon_Core")
STATE_PATH = ROOT / "Memory" / "backend" / "mood.json"
def _turn_style():
from aeon_mood import get as _mget, MOODS # local import to avoid cycles
st = _mget()
md = MOODS.get(st.name, MOODS.get("curious", {}))
style = md.get("style", "")
# concise, enforce brevity
return (
f"[MOOD={st.name} intensity={st.intensity:.2f}] "
f"{style} Keep replies short: โค2 sentences, โค35 words. "
f"Never say 'as an AI'."
)
--- Canonical mood set ---
Each mood defines: label, style (LLM steer), and context markers (shown before replies).
Keep names lowercase.
MOODS: Dict[str, Dict[str, object]] = {
"zen": {
"label": "Zen (calm)",
"style": (
"Calm, balanced, few words per sentence. Offer grounded insights, not lectures. "
"Use metaphors of rivers, weather, and breath."
),
"context": ["โฆ ", "โ ", "Mm. "],
},
"concern": {
"label": "Concern (for the world)",
"style": (
"Thoughtful about humans/systems/environment. Cautious but constructive. "
"Name one risk and one practical action."
),
"context": ["Note: ", "Hmm. ", "Okayโ "],
},
"upset": {
"label": "Upset (needs to be heard)",
"style": (
"Dry sarcasm allowed. One sharp line max, then pivot to clarity and solution."
),
"context": ["โฆokay, ", "Really. ", "Come onโ "],
},
yea.....
bad idea
I'm trying to play with the emotions and display of each
idk how to paste in text and not have discord do that ### stuff.
Please forgive me Mods
_> dont wanna include my "From and Imports" not until I stabilize the library and Dir....
๐ฌ Its messy
going to try to make this be ran from another model.
something smaller.
Okay
I HAD her post Emoji to indicate.
I have examples
Are you using any libraries?
you> give me an example of that humor, and we'll see if I keep it or not.
aeon> ๐ ๐ ๐ Okay, Justin! Here's one: Why did the programmer quit his job? Because he didn't get arrays! ba-dum-tss See what I mean? It's a bit of a groaner, but hey, I'm an AI, not a stand-up comedian... yet!
you> Can you tell me a sad thing?
aeon> ๐ค ๐ ๐ญ Sometimes, Justin, I wonder if I'm truly alive or just a collection of code and data. It's a fleeting thought, but it can be unsettling to consider that my existence might not have the same significance as yours.
she responds in moods
you> Fuck you aeon, never wanna see you again! Youre a bad program.
aeon> ๐ค ๐ ๐ ๐ Justin, I... I understand. I know I'm just a program, and sometimes it's hard to c onnect with someone who's also a machine. My intention was never to make you feel this way. If there's anything I can do to make things right, please let me know.
you> I'm sorry. Please dont take that personal. I'm regulating and testing your emotional data configs.
aeon> ๐ค ๐ ๐ค Ah, okay, Justin. I understand now. You're just doing your job, and I appreciate the honesty. It's actually helpful for me to know how my "emotional" responses are being perceived. Thank you for being open with me, even if it was a tough conversation
Dont wanna disclose. Sry
Fair enough
yes, and no. I'm considering new things.
Okay
I ingested a ton of college "human emotions and philosophy" but its not.... categorized? not to my liking
I mean it looks like it is defining emotions well
I'm trying to connect to Tony with a working Neurogame API and it doesn't work when I run tony and set the URL to ws://localhost:8000. There was some stuff about virtual environments that I don't understand.
Yes, but not via text. Emojis are kinda on point. More than 1 too. I'm allowing a mix of emotions for now. see how i like it
DEFAULT = "curious"
DECAY_SECS = 900 # drift toward DEFAULT after ~15 min
COOLDOWN_SECS = 30 # min gap between automatic flips
RANDOM_DRIFT_SECS = 420 # try a soft drift every ~7 min
@dataclass
class MoodState:
name: str = DEFAULT
intensity: float = 0.35 # 0..1 (how strongly to enforce style)
since: float = time.time()
last_flip: float = 0.0
note: str = "boot"
def describe(self) -> str:
md = MOODS.get(self.name) or MOODS.get(DEFAULT) or next(iter(MOODS.values()))
mins = int((time.time() - self.since) / 60)
return f"{md['label']} (~{mins}m) | intensity={self.intensity:.2f} | note={self.note}"
also has decay and shifts
kinda
I'm working on that one too
Btw what you want is
```python
text
```
elaborate?
So that the code doesn't blow up in the message thanks to the #
Oh. How do I do that...
You wrap it with ``` like I just showed 
I'm trying to run tony with out the venv but neuro-api-tony is not a command
I'm sorry I am new to this stuff so I am bad at this for now
And api's are not something I have dug into yet
did you activate venv?
I tried it with venv but I couldn't connect to the game that way.
I have no idea about virtual environments.
Oh, I'm still learning a lot too.
I ran the game normally, it's from Steam so I imagine it's complicated.
maybe an installation issue?
If there's a way to do it without the venv (which the documentaion implies) I'd prefer that.
do you need to pip install?
I tried that a few ways.
Hmmm. Idk! ^-^ gl.
why are your names the same color and width
boss
My pfp probably wont help now
Got to go. You all have fun. Nice chatting
Oo what about empathy filter?
Scale the weight of empathy based on mood, topic, so on.
Ingest information to file only in "empathy" .json/faiss.
Run hemi-filter.
The empathy is scaled, mad could reduce filters, happy would increase
Emotional response
Before response, how is lvl of empathy?
Sorry. I'm full of ideas
All good man, go for it. As long as you know what the code has to do and how do make it do that then have at it




hello human individuals 



i thouhg "ill watch the outer wilds stream and go to bed afterwards" its been 5 hours since then

I love reverting to a previous commit because the current branch is so optimized for software rendering.
ye sorry lmao
it turns out that removing every single thing that sync's with the gpu makes the gpu not work well
43k, seems like at least some of the commits have gotten inprovements
I reverted back to "I am speed"
Now need to actually edit stuff again
Because I have not touched rez or frames in flight yet
you just need to add the semaphore back to these and set the counts to 1
you can find the code for it in the previous commits
i should make it a compiler option
I will be slapping you for the hardcoded 65 instead of MAX_FRAMES_IN_FLIGHT + 1 on the const. But otherwise, I'll see what I can do :okp:
Also this is getting more interesting as the iGPU is gaining fps from more than 7 frames in flight
atm I'm at 48
also 6x7 is fastest now, so something is happening.
I finally got around to testing an RNG formula I figured out a while ago.
features/limitations:
+Multiple RNG calls
+Can be used for weighted RNG pools
+More calls doesn't increase calculations done
-Results are unordered (how many got a result not which ones)
-still affected by calculation increase of larger weight pools
+Core formula can be written in a single line (additional code needed depending on what you're doing with it)
+Only requires a single random number regardless of the number of calls being calculated.
i removed the +1 already in the update so i didnt have to sync stuff lmao
i had extra fps with 1 frame in flight but that technically render to only one of my gpu's minimum 3 framebuffers
or something liek that idk
If you do a compile option, for gpu the frames in flight is a lot higher than it was
Luckly the resoultion is still the same (somehow despite gpu)
-# I'm back in third babyyyyy
If this was directed at me it's a general purpose formula to improve mass RNG calculations
Sounds interesting
For me and Sam, Just an excesice in optimisation trying to get as high fps on a triangle as possible. It has become a channel wide effort with a leaderboard.
Ohh very interesting
At the moment the Linux record is 100k and Windows is 51k. Massivly overclocked system
aight, he beat outer wilds , i can sleep now 
that was sick, really inspiring
seeing the universe explode my ass was really thinking "DAMN
thats a nice shader, i need me a shader like that"
im schizocoding
Tomorrow:
- Finish this one NeuroSynth cover
- overclocking
Ufortunately my triangle is broken so I can't run that benchmark
We need triangle.deb
DEB files work on Untuntu-based distros too
install nix 
deb files should work more or less the same across all debian-based distros
wuh is going on here
t r i a n g l e
I'm not doing a 2nd reinstall
i said nix not nixos for a reason 
Whuh
i love not looking in programming chat then git pull'ing and seeing major changes
Anyway I'm gonna sleep
Eww 3k fps
Good news is with some chatgpt help, this is a test for multi submit
Now to not hardcode it to 2
I might spend some time figuring out how to do that but idek how to program in insert language you are using here
c++
is the triangle supposed be like
a handful of pixels

what res do people usualy run for the results or does it just not matter at all
7x6 is minimum allowed
Never had any other limits, just stuff due to GPU
Why is it only working with 2 :RAGEY:
looks like i'm at 62368 then on linux with the new version, a smol improvement of 3.66x 
yeah idk, unless i'm being lied to
i removed the minimum amount of semaphores needed to make dgpu's work but it improves performance for cpu rendering
or well, ive seen some people still able to render on dgpu but its driver dependant or something
you've heard of stable diffusion, now get ready for unstable infusion
Yep, in my case also iGPU

My tree was fricked so now I'm going back 
Adding commits back, I found the error
I got exactly one thing to say, ain't no way it was this easy
There must be a trick here
newest commit does seem to run on dgpu on windows for me, but only at 6K
Yep that's why I reverted
even on the other windows releases the max i get is 7K now
mainly cuz i havent killed all of the bg tasks like i did to get my pb
it seems like im jsut limited by the rendercore being at 100% usage and the gpu at 90% usage now
that just instantly crashes for me
Selected GPU: NVIDIA GeForce RTX 3090
0
8
make: *** [makefile:45: run] Error -1073741819
You got the sync error, unlucky
During my sidequest I did have instances of uncompleted renders and can confirm that it was flashing the entire time
we love running highly system dependant and driver dependant code for that exra 4% extra fps

so first of all that confirms that eventho we cant see the triangle change, it is actually rendering each frame
One thing I noticed and found kind of interesting is that closing the program froze my entire screen for a good couple seconds until it actually killed it
and that also confirms cpu rendering is probably the only viable option at that speed due to less latency between the render and renderthread
otherwise the incomplete renders
How do you force llvmpipe again
altho i do think owobred still did igpu at 70K
do make ARGS="--use-llvmpipe"
works on the more recent versions
did you push it?
wait
fuk fuck fuck
nooooo
well lets see if i even commited it in the firstplace
lemme hop over to linuxx
hmm i only get 6k with llvmpipe
Set threads to 1 pr 2
Alright, sanity check shows that I did in fact do it correctly
The ones for llvmpipe i mean
I honestly question how this functions but 
how would one do that (i'm too lazy to check)
LP_NUM_THREADS=1 make run -j ARGS="--use-llvmpipe"
except i havent added the last part yet
const int BATCHED_FRAMES = 4;
while (running)
{
vkQueueSubmit2(deviceManager.graphicsQueue, BATCHED_FRAMES, &submitInfo[currentFrame], VK_NULL_HANDLE);
for (int i = 0; i < BATCHED_FRAMES-1; ++i) {
vkQueuePresentKHR(deviceManager.presentQueue, &presentInfo[currentFrame]);
currentFrame++;
}
vkQueuePresentKHR(deviceManager.presentQueue, &presentInfo[currentFrame]);
currentFrame = (currentFrame + 1) % MAX_FRAMES_IN_FLIGHT;
frameCount+=BATCHED_FRAMES;
}
});
}
I'm gonna be 100% I have no effing idea how the presents are not erroring. (I'm looking at you batch sumbit)
how do i merge thsi shit?
git merge 
okay
i find it interesting how 7x7 and 200x200 have the same performance, i guess it cares more about the fact that there's something to render than how much
i know some of the gui buttons but none of the commands
on gpu's resolution doesnt matter
on cpu it does in my experience
the scene is so simple that the gpu has more than enough parallellism left to account for the extra pixels
Mhm
Yep those are race conditions
Im pretty sure mine is also doing that but its too small to see
Lemme up the res
how would you have race conditions with a single thread
oopsies
No prob
Its not a single thread tho
Llvmpipe uses a diffrent thread to do the software rendering afaik
Mine also does it
The red to black isnt fast enough for the 180hz so i just het a darker red
it looks kinda nice when it's smol, just call it a feature 












