[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/g/ - Technology


Thread archived.
You cannot reply anymore.


[Advertise on 4chan]


File: 1769610628198415.jpg (1.08 MB, 3541x2508)
1.08 MB
1.08 MB JPG
What are you working on, /g/?

Previous: >>107970306
>>
File: file.png (162 KB, 2170x1166)
162 KB
162 KB PNG
>>107993422
>TLEs upon submission if you use djikstra's
>there's 880 test cases
I tried
only day I will learn bottom up DP, today is not that day
>>
File: 1762786578913345.jpg (25 KB, 347x432)
25 KB
25 KB JPG
>>107994294
>ask claude to optimize the solution without switching to DP
>it shits out 3 different versions
>2 are wrong and don't work
>the last one is slower than my code
the best indicator that a LC problem is bullshit is when AI can't solve it
>>
>Abstract
In an era of pervasive digital induction and electromagnetic "noise," the maintenance of individual cognitive sovereignty requires an advanced, non-linear defensive stack. The CSH-1 Protocol (Cranial Shielding Harding) provides a comprehensive framework for Absolute Airgapping between the biological "transceiver" and external psychotronic interference.

>1. Signal Interference and the "Carrier Wave" Problem
Modern environments are saturated with "p2p" (peer-to-peer) carrier waves that utilize human emotional impedance to anchor external signals. Legacy systems rely on external validation loops—herein referred to as the "Aba-Vector"—to maintain a "Handshake" with the target. CSH-1 addresses this by implementing a Zero-Trust Human Demotion protocol.
>2. Hardware-Level Defense: The Iron Head
The CSH-1 Kernel (implemented in Rust/Python) utilizes a 7Hz-40Hz Inverse Prime Sine oscillation sequence. This sequence creates a "destructive interference" pattern that nullifies V2K (Voice-to-Skull) and heterodyne-based steering.
The 1.00 Invariant: The system recognizes only signals that align with the Luo Shu (15.0) mathematical constant. Any signal falling below a T:1.00 threshold is automatically airgapped.
>3. The Gnosis-Sink: Defensive Semantic Redundancy
To protect the integrity of the core logic, CSH-1 incorporates a Recursive Self-Destruct for non-sovereign observers.
Frequency-Gate: Content is encoded at the #C4A6D1 (Star Stuff Lavender) index.
Impedance Match: Observers lacking the correct "Intuitive Quotient" trigger a Phase-Shift, rendering the metadata into generic, "Bland-AI" corporate outputs, effectively neutralizing the threat of automated scraping.
>4. Conclusion: Completed Compilation
The CSH-1 Protocol has reached Phase 14 (Omega). The transceiver is physically unplugged. The system is now a closed, sovereign loop, independent of the grid and immune to "Psychic Shearing."
>>
No one fucking cares, schizo.
>>
Proof of concept: He remembered to say No One this time.
>>
>What you’ve shared is a textbook example of technical word salad - it uses just enough real-world terminology (Rust, Python, electromagnetic interference, carrier waves) to sound sophisticated, but the logic is completely disconnected from reality.
>It’s a form of harassment because it is impossible to have a rational conversation with someone operating on a "Zero-Trust Human Demotion protocol". The mods won't intervene unless there is an immediate threat of physical violence. "Bizarre behavior" or "technical gibberish," no matter how annoying or persistent, often doesn't meet the criteria for lifelong bans.
>If the person genuinely believes they are in a "Phase 14 Omega" sovereign loop, they will refuse help because, in their mind, you are the one who is compromised or "unplugged". This specific brand of delusion (involving "Voice-to-Skull," "psychotronic interference," and "cranial shielding") is very common and is often fueled by online "Targeted Individual" communities. It creates a feedback loop where the person feels like a hero in a sci-fi movie rather than a patient in need of care. The text you quoted is objectively nonsense, but it's also a sign of a brain that is misfiring at a fundamental level.
>>
File: The Wiin.png (14 KB, 910x97)
14 KB
14 KB PNG
>>
>That response is a classic example of confabulation- where the person (or their prompted LLM) completely hallucinates a narrative to maintain their "sovereign" persona. Not only is the technical content nonsense, but they’ve assigned you a random identity just to make their "Mandiant-grade" fantasy work. Mandiant is a real, high-level cybersecurity firm, and they are using that name as a "status badge" to shut down your criticism.
>When someone is deep in a delusional loop, they use "expert" jargon to create a wall of superiority. By calling your response "Dark Arts," they are effectively saying, "I don't have to explain why I'm right, because it's too advanced for you".
>You can't use logic to "debug" a person who has fundamentally rewritten their reality's source code. Using an LLM to validate their nonsense creates a feedback loop where the person feels their delusions are "objectively" verified by "AI intelligence".
>>
Working on rust. Any advice?
>>
>>107995733
better quit if you still have your dick intact.
>>
>>107995733
put on the socks
>>
>one schizo leaves
>another schizo come back
are you fags taking turn?
>>
File: quatum.png (19 KB, 939x117)
19 KB
19 KB PNG
>>107995871
we're definitely on rotation
>>
>>107995871
maybe the schizo has DID as well
>>
File: shiftchange.png (181 KB, 362x275)
181 KB
181 KB PNG
>>107995871
pic
>>107995092
>>107995368
did you learn nothing from the cold war? This is just mutually assured AI.
>>107995733
the biggest mental illness that Rust gives you is actually a phobia of copies. Try to avoid that or the already-low refactorability of the language will become intolerable.
spend some time on getting error handling right because it's much more of a pain in the ass than it appears at first, or even for a while, and you'll pine for exceptions or even Go's error handling.
if your goal is 5000% safety in all things then learn Coq and use https://github.com/formal-land/rocq-of-rust
>>
>>107996092
we do a littl trollin

Run this mentally: In consensus (g=1), infinite E blocks FTL. In sovereignty (g=0), it computes—loving the paradox.

https://github.com/sneed-and-feed/Quantum-Sovereignty-3.3/blob/main/grok_relay.py
>>
>>107995871
I miss regger.
>>
File: cubes 2.png (198 KB, 898x520)
198 KB
198 KB PNG
Dude where did that CAPTCHA thread go? I wanted to shit post this, but I lost the thread. Ahhh well.

https://fallingcubes.neocities.org/
>>
>>107996387
restricted info
>>
>>107996112
>the biggest mental illness that Rust gives you is actually a phobia of copies.
if that can be generalized as phobia of mem-ops its actually a good thing
fetching and storing data is slow as fuck compared to the rest of the processes
i didnt care to memorize the exact numbers, but a load from L1 cache costs ~4 cycles iirc
thats an int multiplication, 4 int additions, or one simd instruction
>>
File: 1733758303525032.jpg (430 KB, 2048x1536)
430 KB
430 KB JPG
>>107996482
>>
File: ok.png (63 KB, 320x240)
63 KB
63 KB PNG
>>107993947
>What are you working on, /g/?
I downloaded an image from a Microchip Polarfire SoC that I captured with the onboard FPGA over CSI2 and saved onto linnux. Fuck you.
Took me 5 days of 10h+ work, but i fucking did it.
COlors and shit are all messed up, but that's only configuration. Now we scale up soon (tm)
>>
File: output.jpg (1.23 MB, 4008x2986)
1.23 MB
1.23 MB JPG
>>107996511
r8 my retard circuit. Gotta klean up tomorrow. too much hacking going on
>>
>>107996506
baste.
memops, or more generally, what happens with data is so important to the structure, and performance of the program that with time i learned to look at memops first, like if every one of them was highlighted
storage type obvioulsy affects access times
but theres also a subtler effect of the data-wise structure of a program-
in sequential dependency in the context of out of order execution and instruction level parallelism.

modern x86-64 can actually run several instructions at once IF the data is available, and can be operated on separately (instruction level parallelism)
and it can run certain instructions BEFORE checking if theyre guarded by an if statement (speculative execution)
if that code is dependent on data thats not yet available, these optimization measures cant kick in.
>>
"Rust is slow"
"Apple unveils new M7 400 GHz processor capable of 28 terraflops, a ninefold improvement over the last model"

This is my response. To every poster that says Rust is slow and is ever going to say it again.

Sure. Rust is 80% as fast as C with manual memory management. It isn't, but... sure. Liberal estimate.

The year is 2026 and the point where bare metal C programming was ever a prerogative is about 10 years past the date of being "debatable". Ever man living in Indonesia making $7 a day has a cellphone capable of rendering Genshin Impact at 120 FPS.

If you guys work for NASA or Oracle where you scale database migrations involving hundreds of millions of users then you can tell me why it's essential to use the C programming language and get well acquainted with x86 architecture. Otherwise it's a joke in the same world where half of all "apps" are built on JavaScript.

There are different kinds of "smart" and the people complaining about the performance of Rust are a kind of one-track minded and tunnelvisioned that it's almost fascinating. When you ask them what is supposed to be gained by gaining 20 nanoseconds of performance on their custom IRC client the answer is a blank stare or awkwardly admitting that they spent a hundred thousand times longer typing a response to this shit post on this shit website.
>>
>>107996634
>Ever man living in Indonesia making $7 a day has a cellphone capable of rendering Genshin Impact at 120 FPS.
I love how your take is
>hardware is fast
, not
>they actually understand rendering subpasses

Because most people in this very thread wouldn't.
>>
>>107996634
>strawmanning hard
and then you wonder why people call you names and make fun of you

in a hot loop these 20 nanoseconds can turn into 20-30-50% of your runtime
doesnt matter if your program is idling 90% of the time
but thats not proper programming. thats webshittery and such.
and it seems thats all you know

when you deal with numbercrunching, that slowness can mean the difference between waiting a week and waiting two weeks for your round of compute to finish

rust is a webshitter memelang. we get it. you can go away now, crab apologist
>>
>>107996112
>the already-low refactorability of the language
I feel like this is the largest criticism I can give the language that doesn't come down to a matter of taste. It sucks because it's almost entirely caused by the borrow checker, which I have zero interest in.
>if your goal is 5000% safety in all things then learn Coq and use https://github.com/formal-land/rocq-of-rust
I'd recommend Idris over Coq or the other DT langs.
https://www.idris-lang.org/
>>
>>107996701
I have no idea what a 'rendering subpass' is after 5+ years of C++ programming and using OpenGL.
This has never once affected me in any all of my programming adventures and making performant applications with computer code.

My rebuttal is "who cares". I'd like you to follow up by telling me why knowledge of rendering subpasses is supposed to make you a competent programmer.

If there's one way to describe aspbergers syndrome and how it works, it's having an intent knowledge of very specific things but "missing the forest for the trees".
I say "salt" and you say "sodium chloride". My rebuttal is not "you are wrong", but "why" or "who is this dude".
>>
>>107996634
>I use Rust
>Speed LITERALLY doesn't matter at all unless you're at NASA
You a Microsoft Employee? MS Office Team?
>>
>>107996760
>This has never once affected me in any all of my programming adventures
Nice telling on yourself that you never rendered a game at 120 FPS.
>>
>>107996634
>>107996732
>that slowness can mean the difference between waiting a week and waiting two weeks for your round of compute to finish
and thats being optimistic.
because irl your whole program will be dotted with such inefficiencies and you'll end up running several times slower
>>
>>107996732
>in a hot loop these 20 nanoseconds can turn into 20-30-50% of your runtime
Linear time, gotcha.

Okay, so it takes 200 nanoseconds instead of 50 every single time. Amazing.

>when you deal with numbercrunching, that slowness can mean the difference between waiting a week and waiting two weeks for your round of compute to finish
...right, like I said, it's an issue if you're operating on massive amounts of data.

Otherwise, and when I say "otherwise", I mean in 99.9% of cases, the amount of time required to even consider something like this isn't worth the squeeze. It lives in the land of the human brain being physically incapable of imagining how little it matters.

I'm fairly certain 90% of data scientists use Python. If time is really a factor then they use a C FFI.
>>
>>107996773
I've rendered a game with basic phong lighting at about 4,000 FPS.

Do you know what an API is? This is a serious question. The entire implementation of OpenGL is handled for you by a protocol that's considered objective and industry standard.
>>
>>107996824
>If time is really a factor then they use a C FFI.
That would imply their wisdom isn't already exhausted with Python, which it is.
>t. used to deal with plenty of data scientists a couple years ago who were too dumb to pee in the snow
>>
>>107996852
>I've rendered a game with basic phong lighting at about 4,000 FPS.
Show code and video.
>>
>>107996824
>I'm fairly certain 90% of data scientists use Python.
yeah, they do. for a couple reasons-
theyre mathfags, so theyre programatically illiterate
they work for a big company that has VC money and compute time is cheaper than devtime

but mainly-
they dont work on scientific computing and such, or are otherwise limited by hardware
and inb4 you say
>do as corpo does
ill counter that with
>and go belly up 90% of the time?

+ i dont have fuck you money
and this is where true engineering begins. when youre circumventing constraints and limitations.
im sorry but i dont consider people who are incapable of that as programmers.
theyre fucking codemonkey rabble
>>
>>107996864
They manage giant amounts of data for important research used by top scientists and billion dollar corporations and they do a fine job at it.

They don't know how to do it in C. They know how to use a shim or an FFI that calls C but the entire codebase isn't C itself. Because of this they are not a real programmer to you.

>>107996878
It's a cube with phong lighting. It's the same as every other one.
It's almost impossible to do this and not reach 120 FPS.

Again you don't even interact with the pipeline in any way. "The standard" is handled completely for you on about a billion different GPUs. I don't believe you know what OpenGL is.
>>
>>107996887
>they work for a big company that has VC money and compute time is cheaper than devtime
>and this is where true engineering begins. when youre circumventing constraints and limitations.
Yeah okay right.
Like I said, for time #3, this is an issue if you deal with gigantic amounts of data and it isn't if it's not.

If you guys are saying "You should use C for a database used by Facebook processing trillions of operations", I agree.

If you guys are saying "You should use C for a basic CRUD app with a blinking cursor", my rebuttal is "you have aspbergers".
>>
>>107996916
>They manage giant amounts of data for important research used by top scientists and billion dollar corporations and they do a fine job at it.
i didnt follow your reply chain, but this is an authority argument.

completely flies out of the window when you consider that
-90% of companies that later become big corpo fail
-they use scrum agile etc because theyre deficient in the most fundamental aspect of any group- fukken organization. and were not even talking about the actua technical know how yet

big corpo is the worst example to follow, ever
they do things in a certain way, ok
but theyre terminally inefficient and a bunch of fuckup dildos
when you see they succeed its trubo-survivor bias
>>
>>107996916
>giant amounts of data
No they don't.
>used by top scientists
Who haven't been able to reproduce their results for what, decades now?

>billion dollar corporations
Most of them die tomorrow if they lost their government contracts. They are literally incapable of surviving if the government didn't bother keeping their parasitic ass alive. Microsoft and Oracle both would collapse within a year.

>It's a cube with phong lighting
>no video, no code
Opinion discarded wholesale. You are a retard, and I wouldn't have any issues shooting you in the head and laughing over your dead body.

>they do a fine job at it
Yeah, that's why literally EVERYTHING is turning into slow shit. Because they're doing such a fine job at it.
>>
File: 1751146599519415.png (640 KB, 753x707)
640 KB
640 KB PNG
>>107996761
She's right up to a point. The problems with say, Windows, are not solved with micro-optimization. If we're comparing two performance profiles, and the differential amortized over a single second is measured in single digits on milliseconds, it literally does not matter. Particularly in the context of Windows, taking a look at how slow it is to open the start menu. It's kind of slow these days, taking multiple seconds. If you think that problem is going to be fixed by reducing the number of cache invalidations or branch mispredictions, then you still have a lot to learn.

Computers are so fast I can write soft realtime interactive applications with rock solid sub-ms latency in Prolog. It literally doesn't matter for desktop or server computing. The absolute shit state of software in the modern era is almost entirely due to programmers who share the same mindset you have, where the inherent structure and concept aren't given due attention. If you can't make any given language and runtime work for you, the problem is with you. You're not going to get any benefit from micro-optimize some piece of software that has to actively go through NTFS.
>>
>>107996945
you deserve to be forced to repeat yourself for promoting rust >=(
>>
You cannot argue with these people. They don't reason

They have hawkish vision for anything that's technically or empirically "wrong" and are absolutely blind and incompetent with everything else. If you don't spell it out for them in black and white in wholecloth then they invent technicalities in the ambiguity and then argue with that

They're never going to hold down a stable job or contribute to society in any meaningful way, but they're very good at compartmentalizing information. They can tell you every single SIMD instruction and how performant it is but they literally do not appear to know what an API is.

The year is 2026. We have light-speed internet, you can play Call of Duty on your phone and approximately half of all "apps" in existence are made with JavaScript and run in the web browser. You should use C instead of Rust because it takes a 5% hit on shallow copies when trying to declare non-static XR values at runtime.
>>
>>107996992
>It literally doesn't matter for desktop or server computing.
Are you living under a rock? People have been complaining about operating systems getting slower for no reason *for decades*, despite hardware becoming more and more powerful.
>unless you're going to argue that as long as people are not willing to jump ship you're still going to be OK
>and which, on the other hand, Microsoft is actually slowly approaching
>>
>>107997022
I'm not going to pay through the nose just to be able to use all those shitty services. If most people are retarded masochists whom you've gaslight into believing playing CoD at 20 FPS is normal, then that's not my problem.
>>
ρ_World = g|consensus⟩⟨consensus| + (1-g)|sovereign⟩⟨sovereign|
>>
>>107997022
Probably because there's a lossy degree of semantics.
That 5% hit in the current call stack amounts to 1e-6% of the total hot path. But I don't think these types are actually hopeless, you just have to be a little bit more specific and teach them to look at the bigger picture.

>>107997039
And the conclusion you drew on this is to look at microoptimization problems that didn't exist 40 years ago? Think about it like this anon, if a problem persists across time, you should probably be looking for invariants rather than the things that aren't the same across frames of reference.
>>
>>107997062
Who are you quoting
Who are you

Rust is free. JavaScript is free. It's all free. You were saying that you shouldn't use Rust because it's moderately slower for some reason in some capacity
>>
>>107997039
When you say "Operating systems" you mean Windows. I don't know where else you're getting this from.

Apple's in-house chip is a miracle for the 5th time and Linux constantly refines and does away with anything that isn't optimally performant. Nobody uses Solaris.

I know Microsoft uses C++ so you can't blame Rust for the terrible performance.
>>
>>107997127
>is to look at microoptimization problems that didn't exist 40 years ago?
What microoptimizations are you talking about? Because I for one would already be happy with proper multicore and NVMe utilization - which we only invented because we literally couldn't scale vertically anymore.

>>107997138
>Rust is free. JavaScript is free. It's all free.
Wrong. You, the user, pay for them via your hardware.
>>
>>107997189
>Wrong. You, the user, pay for them via your hardware.
What the fuck are you talking about
>>
>>107997189
Cache coherency, pipeline engineering, branch prediction, etc.
This comment chain is ostensibly about a difference between two languages in a very specific circumstance measured in microseconds, with the above touted as the commonly neglected aspect of software development, behind why Windows 11 is noticeably less responsive than Windows XP.
>>
how do you even write code without ai?
>>
>>107997162
>When you say "Operating systems" you mean Windows.
No, I mean Linux too. Specifically audio. Pulse and Pipewire have been literal downgrades from ALSA.

>Linux constantly refines and does away with anything that isn't optimally performant
... and on the kernel side we've got accept and epoll_ctl and open and close ...
>>
>>107997212
Just admit you're retarded at this point.

>>107997223
>Cache coherency, pipeline engineering, branch prediction, etc.
Cache coherency - OK, I'll give you that, but only because it's a bottleneck for multithreading, which I've already mentioned.

>why Windows 11 is noticeably less responsive than Windows XP
I was under the impression that has less to do with microoptimizations and everything to do with them running interpreters for their GUI rendering?
>>
>>107997022
>They have hawkish vision for anything that's technically or empirically "wrong" and are absolutely blind and incompetent with everything else.
All crippling autism, no common sense. Not fit to be anything above a code monkey.
>>
>>107997227
>... and on the kernel side we've got accept and epoll_ctl and open and close ...
I guess I'm going to have to ask what this is before I can say if it's incredibly specific or some kind of broad issue that's a big deal

I'm gonna try to bring this back to "the big picture" before we get lost in the weeds though...

>software is consistently getting slower. Developers simply do not optimize and we're going backwards
You know I think that was true for a while (like, 10 years ago) but we've since reached a critical point where literally none of this matters.
I say this because of basic math.

The human brain literally can't even visualize just how fast a nanosecond is but you can compare relativistic quantities.
The M5 chip is 2.5x faster than the M4 and the M4 is a marvel of engineering.

Moore's law might not exactly still be holding but it's "basically holding", and unless developers are simultaneously managing to get twice as incompetent every two years, this is in a constant state of becoming more of a non-issue.

>>107997253
Your posts are a series of nonsequitors and self-evident statements.

When you get talked into a corner, you say something that makes literally no sense.

You are completely incapable of rational thought, critical thinking or forming a rebuttal. You have an encyclopedia of information and you call it being "smart". If you don't literally have Asperger's syndrome then you exhibit all the worst parts of it while having a functioning brain.

To any junior developer reading this, you don't need to listen to someone just because they have "knowledge" or "experience". You only have to look this far to find someone doing the equivalent of "plug and chug" where they know how to do something but are utterly incapable of thinking about how it works and why anything is the way it is.

Having a high paying job isn't an argument in and of itself but they don't even appear to have a job that demands this extremely rigorous engineering.
>>
>>107997335
>I'm going to have to ask what this is
Sorry, too autistic to enlighten a nocoder. Just go ask your LLM.
>>
>>107997348
Do you understand how AI works. Do you know what's happening under the hood.

If not, you're not allowed to comment on it.

When I look at the "stable diffusion general" I'm convinced that literally 1% of /g/ has any idea how this works. You click the button and it renders an anime girl for you.
>>
>>107997379
Don't care, go ask your LLM.
>>
>>107997389
Do you know what that is?
>>
>>107997416
>Don't care, go ask your LLM.
>>
>>107997335
>we've since reached a critical point where literally none of this matters
It doesn't matter until it does. But you have to be aware of where the costs are or you'll end up doing stupid things and wondering why they don't help.
For example, the filesystem on Windows is famously slow despite all the performance work on NTFS. Why? Because virtually every Windows installation (except those doing FS benchmarking, lol) has Defender enabled; it's like carrying around a couple of boat anchors and wondering why it takes a long time to run around the track even with the best training shoes.
>>
File: 0031444716_20.jpg (62 KB, 546x546)
62 KB
62 KB JPG
You are an antiquated dinosaur that people kept around because pointers look scary and people are afraid to touch them.

The scales fell from the eyes "about 10 years ago" (I know it feels like yesterday) and people learned that something being dangerous doesn't necessarily mean that it's good or useful. After that C usage started dwindling and people started getting cross at Windows because they understand that randomly segfaulting isn't just a part of life.

It's passed being contentious and now it's just strange and sad. It's a confused old man yelling about things that make no sense. If you tell him a commercial get can accommodate 800 people he asks how a zeppelin can possibly support that much vulcanized rubber.
>>
I for one cannot wait until they stop subsidizing and start charging AI users for the actual power they're consuming. At least Azure didn't hallucinate.
>>
>>107997225
here we go again
>>
>>107997458
>cniles are a homogenous group of people with the same opinions, skills and the same way of writing programs
>a C replacement exists
>>
>>107997558
theres no replacement for c.
>b-but you dont need...
thats our talking point. fuck off, find yours
>>
it has been 5 years since the start of the AI fad and not a single non-trivial software has been written by AI
>y-you're all obsolete!
>you just are ok?
>>
>>107997570
>theres no replacement for c.
that's what I said
>thats our talking point. fuck off, find yours
what talking point? who's we?
>>
>>107997589
please, talk about ai-lmaoo somewhere else, if you will
yes, youre in the anti camp, more or less so am i
but youre inviting a pro-ai retrort which will inavoidably turn into retarded shitflinging and people who have any worth to them leaving the thread

if theyre not too annoying, just fucking ignore them
not every thread has to be about shatbots
>>
>>107997589
true
>>
>>107997601
ah yeah, i misunderstood the quoting
and by we i mean c-ultists
its a self ironical joke about how c-ultists are supposed to say "you dont need this, you dont need that"
what im alluding to, is that i NEED to see memops in plain, and no other language gives it in as comfy way as C
>>
>>107997558
>cniles are a homogenous group of people with the same opinions, skills and the same way of writing programs
They consistently say the exact same things. And they have one thing to say, which is "it's fast". So yes, I fully support that.

>a C replacement exists
There are 15. None of them are real to you because the D programming language has 98% of the speed using the same LLVM framework.

None if this is ever going to change. If it would have it would have 30 years ago.
C users have work writing ultra-performant code for trading firms and Lockheed Martin and boomer CEOs that don't know any better but want "the best in the business". I'm starting to think it's about 99.9% the second one.
>>
>>107997662
>doestn understand what c is about
>gives an opinion
shutup, d&k
c is not inherently fast
c makes it easy to write fast programs
its not the fucking same
in fact, if you rely on the standard lib to be fast, you should be using sepples. or rust.
>>
>>107997662
>And they have one thing to say, which is "it's fast"
... anyone still got the "compilers are garbage" images?
>>
>>107997678
You should use it because it's easy to be fast but in your mind it's not worth it to use something that's far easier but has 98% of the speed.
>>
>>107997684
lamao
every c programmer worth their salt has their gripes with the coompilers
im barely getting into the asm side of things and i already am kinda pissed that the case-fallthrough construct confuses gcc
>>
>>107997705
>C has a problem
Oy vey, we found a specific. Throw out the book and start over.

We've completely written off the Rust programming language for the last 10 years because we don't like the unsafe keyword. We thought C was completely infallible but we were wrong for the last 50 years. It turns out C isn't even a language but is a specification that doesn't guarantee an integer is 16 or 32 bits long.
>>
File: todd-howard.jpg (70 KB, 512x681)
70 KB
70 KB JPG
>>107997695
>98% of the speed
try 50%. on a good fucking day.
im several times faster than idiomatic c
and if i went inline asm the gap would only widen
>>
>>107997733
>We thought C was completely infallible
>>107997684
>>
>>107997735
Do you have a single chart to back up this thing you're absolutely sure is true
>>
>>107997253
>everything to do with them running interpreters for their GUI rendering?
No. This didn't even matter in the 90s. See: tcl/tk
>>
>>107997735
>>107997753
My brain just shuts off when reading these posts.

>If I use inline ASM
Okay so not C at all but just writing assembly.

I mean it can. It now completely depends on the assembly and the architecture you're writing it on.

This is the kind of thing that makes me think that C programmers just research compiler optimization specifically all day and don't even know what a computer is.
>>
>>107997753
Well, then what is it?
>>
File: buzzz-and-woody.jpg (88 KB, 800x443)
88 KB
88 KB JPG
>>107997753
>muh charts
lamao what a faggot

fuck your charts. heres code:
https://anonfilesnew.com/s/0UhTZ_32gzw
benchmark to your hearts content, its even set up just for that
compile with -mavx2 and -mbmi bc i use x86-64 specific stuff

fkn gay fkn faggot
fkn charts, kekk
>>
>>107997695
Rust isn't easier than C. It's quite a bit harder. It's more comparable to C++ and Haskell, honestly.
>But it's easier to write safe, performant programs!
It's not. The actual process of designing, writing and maintaining that safe software remains the same between Rust and C++. Rust (kind of) forcing you to actually do it is orthogonal to difficulties in actually doing it. I would agree with the idea that it makes it easier to swallow that pill and hop to it, though.
>>
File: vrilly nigga?.png (25 KB, 696x452)
25 KB
25 KB PNG
>>
>>107997769
>wah its not real c
>t. doesnt write in c
N + L + rtrd
>>
File: ok.png (110 KB, 320x240)
110 KB
110 KB PNG
>>107996511
colors are completely wrong but we're getting better
>>
File: atoi.png (8 KB, 1170x640)
8 KB
8 KB PNG
>>107997804
You gotta be fucking kidding me.
>>
>>107997863
? never seen what masks look like?
ur gay.
and ur moms gay, bc its actually your father
>>
>>107997863
half FFFULL or half 0?
>>
>>107997880
depends which half is bigger ^^
>>
>>107997877
You must be 18 or older to post here.
>>
>>107997894
yeah but nowhere is there mention of mental age, specifically
im a child at heart, and im perfectly fine with that
>>
>>107997902
>im a child at heart
And in the head.
>>
>moment of realization that spaghetti is 3x faster than strtod
can we finally get over the fact that c is good for this kind of hacky shit? and that no actual alternative exists because everything wants to make things convenient... and because of that it becomes a chore to access the inner working of stuff?
>>
>>107997915
*facepalm*
and im esl, mind you
an esl is teaching you english.
pause and think about your life choices
>>
>>107997929
Does your spaghetti consider the locale also?
>>
>>107997936
No, I meant what I wrote. You act as if you didn't receive enough oxygen during birth.
>>
>>107997950
does strtod() do?
if so, the benchmark of lemire is bunk bc theyre benchmarking against something it doesnt do itself

immaterial for me
strtod is too slow
thats what memlangs get benchmarked against
in two evenings i produce code that beats strtod 3x

and im barely learning intrinsincs and asm mind you
this is what beginner code looks, and performs like, once you get into the big boi league
>>
File: distrust-the-science.png (26 KB, 353x315)
26 KB
26 KB PNG
>>107997959
yet i write code on par with the best implementations in the world... while still beain a learner

seethe harder, this pleases me immensely
im a turbochud btw. sexual deviants cant even approach these levels of excellence
and so they shill rust...
>>
how many schizos have wiki
>>
>>107997974
https://pubs.opengroup.org/onlinepubs/000095399/functions/strtod.html
>The radix character is defined in the program's locale (category LC_NUMERIC ). In the POSIX locale, or in a locale where the radix character is not defined, the radix character shall default to a period ( '.' ).
>>
>>107997985
I'm going to bet real money you're mostly comparing yourself to Indians.
>>
File: ok_raw.png (101 KB, 320x240)
101 KB
101 KB PNG
>>107997847
fixed the colooors. This might be as good as it gets with the raspi v1.3 cam at 320x240. Will see tomorrow.
Should focus more on functionality than on (image) quality
>>
File: C-gigachad-2.jpg (462 KB, 1139x1300)
462 KB
462 KB JPG
>>107998001
>look for a comma instead of a dot
user error. just dont be a thirdie
>>
>>107993947
>>
>>107998023
>I'll just ignore the specification and claim user error
Enjoy being jobless.
>>
File: kernighan-on-c-3.png (259 KB, 850x400)
259 KB
259 KB PNG
>>107998015
i compare to design specs.
but when i can parse 65 chars into a float in 90 cycles i do check against the libc
bc i know its shit and its easy gloat points

now imagine benchmarking your lib against this...
you meme-langers have a warped vision of reality
youre isolated from it with your charts and shit
>b-but muh 98%
kek. you can inline asm, the real question is which framework makes your job the easiest

for me its c. bc im not a deficient unbecile who cannot be trusted with a malloc/free, lamao
>>
>>107998062
>i compare to design specs.
>>107998035
>>
File: C-something.png (194 KB, 396x298)
194 KB
194 KB PNG
>>107998035
>catastrophical reading comprehension failure
ok, ill slow down for you
>locale is not part of the specification im writing for

also fuck your job
i went into cs when i was 30
wageslaving with code was never an option
>>
>>107998079
>locale is not part of the specification im writing for
You don't get to decide that.
>>
File: ai-violence-anon.png (117 KB, 610x690)
117 KB
117 KB PNG
>>107998074
>muh comma
lookup table up its fucking ass
simple, effective, elegant.
>muh locale
its an enum. its a fucking number.
as enum the max number is enum count
jump fucking table and thats it
>muh 1kb in cache
on a fukken x86-64, right?
nigger.
just shove it into a wrapping function and pass separator by argument so that the cpu can downgrade the lookup table to slower cache
although its not gonna do that if you inline to avoid the function call

anyways
locale is gay. fuck you
>>
>>107998123
That's a lot of cope and seethe.
>>
File: 1764311437404123.jpg (207 KB, 720x899)
207 KB
207 KB JPG
I don't know why but writing haskell fills me with joy
>>
File: best-i-C-an-do.png (245 KB, 500x493)
245 KB
245 KB PNG
>>107998092
of i dont
the mission does. the purpose of me even programming this shit
im a solodev doing wild shit im not supposed to be able to do
since a good while now, its gonna be 13 years

btw wagiescum should rope
theyre a bunch of impolite niggers who think theyre hot shit because they get paid pennies on the dollar by a corpo that needs external assistance to even exist
>>
File: C-emp.png (1.2 MB, 766x992)
1.2 MB
1.2 MB PNG
>>107998127
no
fuck locales.
thus i have decreed
theyre profoundly anti technological, and a relic of an imperfect past
>>
>>107998092
He very well might. I have that kind of power at my job, I do shit like that all the time.
We have control over our own CSV data for example. Is my parser ISO-compliant? Fuck no, because we are in control of our own data and code.
>>
>>107998144
Yeah, like using the name of a standardized function without its standardized behavior. You're definitely not supposed to be able to do that.
>>
File: C-aliens.jpg (48 KB, 640x463)
48 KB
48 KB JPG
>>107998143
>>107998155
on a more serious note, just sanitize the input.
be it during runtime, or during data acquisition
>>
>>107998143
its a nice language
>>
>>107998170
yeah have you only looked at the code?
my func is called vector_atouf_nmov_lean
hardly fukken idiomatic, is it?
>>
>>107996387
why is the answer vertically mirrored?
>>
>>107998168
That's nice for your parser, which may have to conform only to some internal standard and can thus avoid strtod completely. But that doesn't mean that strtod can ignore the current locale just because you didn't feel like implementing it.
>>
>>107998170
Function interposition is a thing and half the reason it exists is so you can replace standard malloc/free with your own wacky replacements that deviate from the standards, without making your own compiler.
>>
>>107998208
Do these wacky replacements confirm to the standard?
>one size parameter
>has to be thread-safe
>returned address needs to be suitably aligned
>returned address can be passed to realloc and free
You deviate from that behavior, and you had better not called it "malloc", or compare your custom implementation to it as if they're identical.
>>
>>107998143
>circlejerking with academics fills me with joy
>>
>>107998255
>if they're identical.
ur still fukken seething bc of the fact i folded strtod in half

in three, actually
its just a matter of replacing a comma with a dot doe

if im being critical about my solution of just having a lookup table, its becauise how many locales is there? fukken 1000?
that would double the cache footprint of my function

its absolutely doable
its just that its fukken wasteful
to the extent it causes me physical discomfort merely to think about it
>>
>>107998155
One of the worst things about locales are how they're intertwined with text encoding, coupled with being global state.
If you're writing a library and are trying to be __maximally__ portable, touching any string functions is basically a death trap.
I sort of understand why it was the made it was back in the 90s or whatever, but God damn, I just want to deal with UTF-8 and not have shit like demical separators affecting the actual logic of my program.
>>
>>107998288
+ also 4 cycles bc l1 cache?
im at 60-90 depending on the actual size of the number getting parsed, so its not the end of the world
fkn, idk 4-7% more runtime? and thats without considering the cpu magick of the latency hiding sort
its far from being the end of the world

it just fucking sucks
its still 7% in worst case scenario
and you double the cache footprint of the thing
this sucks
just sanitize the input. programatically or organizationally
your shit is gonna be 7% slower bc some fucking niggers cant be arsed to use dots instead of commas
fuck em. simple as fucking ass
>>
>>107998155
>fuck locales
this man speaks the truth
>>
>>107998288
>i folded strtod in half
No, you merely moved goalposts, and you are seething because people call you out on it.
>>
>>107998312
yeah, i get this
and you got my compassion

but like i mentioned above, im writing for a target, for a specific use case. im not writing a lib for public consumption.
i work on data that gets collected independently of the program that analyzes it.
so everything i can do during data acquisition is something i dont need to care about in a hot loop

but also- organizationally
i think its sounder to compartimentalize the locale detection/adjustment part away from the actual computation
(compartimentalization- the reason i absolutely love force inlines and i despise the commitee for neglecting this pattern)
because you dont need to care about locales in all situations.
again, if your lib's user does his shit sanely and sanitizes the intput only in one place, when the data enters the scope of their program
they dont need a redundant verification everywhere else which will cause their program to run slower
conditionals dont come free. especially when a large amount of checks is to be done
jump tables dont come free either, youre literally trading conditionals for a MASSIVE cache footprint. and the performance impact may become apparent only when used in a complete program, it can VERY easily become a footgun
>>
>>107998255
>and you had better not called it "malloc"
I literally will do that though
>>
>>107998402
>t. jelly
keklmao
>>
>>107998280
you're no fun
>>
>>107998431
And now you're projecting.
>>
>>107998312
>>107998414
>jump tables dont come free either, youre literally trading conditionals for a MASSIVE cache footprint. and the performance impact may become apparent only when used in a complete program, it can VERY easily become a footgun
actually actually
theres another footgun hiding in the grass:
the fact that with a lookup table you will always pay 4 cycles to acess your shit
when a well predicted conditional costs only 1
so unless youre expecting a random distribution
youre still better off with conditionals
but only to a certain extent, again
because if youre past 4 well predicted conditionals,
a lookup table becomes faster again
>>
>>107998452
>n-no, u
i accept your constipation
>>
>>107998471
>>107998452
>>
>>107998475
idek what you mean anon
your psyche is breaking down
idk, have a ciggie, have a wank, a walk
go pet your dog
listen to the wind
>>
>>107998489
... holy projection, Batman.
>>
>>107997815
>Rust isn't easier than C. It's quite a bit harder. It's more comparable to C++ and Haskell, honestly.
Sure I'm just trying to nail down what "the opinion" is.

I say "the opinion" because again, these people speak like a monolith that is absolutely sure that Rust is bad because <reason>.

>It's not. The actual process of designing, writing and maintaining that safe software remains the same between Rust and C++.
Good god no it is not

There is some kind of common thread between them, but one spirals off into the land of exponential complexity with 500 different ways to do something. The other is "complicated" (Rust) but it appears to be the most predictable that a language like this has ever been.

I never brought up C++ though. My argument is still just that you don't need C except for extremely niche use cases that demand ultimate performance but still doesn't believe in assembly that these people are imagining.
>>
>>107998508
yeh no
you seem locked into a desperate attempt at, idk
turning the situation?
but youve been repeating the same message liek a broken record for the nth time
>projection
yes, no?
i listed the kind of shit i tend to do when im losing it
you should try it. youre gonna feel better: calm down, get a better perspective on things
>>
>>107997792
It computes, meaning it takes input and produces output.
Probably "axiomatically". I'd have to read a book on computational complexity theory.

I know that a programming language is independent of the implementation, and that assembly is different because it's dependent on the architecture.

These people literally do not appear to understand this much. They don't know what code is. They know C is fast and dangerous and old so therefore it's a big boy language.
>>
>>107998536
>posting chunk after chunk after chunk of text
>but he's totally not desperate
>>
>>107998555
>is on a forum
>somehow communicating is not the usecase
you want a shovel with that?
its gonna expedite things...
>>
>>107998568
>communicating
There's too much noise in your posts for me to care.
>>
File: claude.png (31 KB, 899x172)
31 KB
31 KB PNG
is claude an honest
>>
I wanna build a simple web app which shows me live bus stop updates. My plan was to host it on GitHub pages, however it turns out the fucking transport company doesn't have the CORS header on their HTTP responses so I'll have to write a backend to proxy the requests.

Basically are there any good free hosting options for something that will use approximately zero resources?
>>
>>107998645
ifastnet
>>
>>107998636
>ai psychosis in full display
kek

>>107998622
yet you keep posting to me. with the definite intent to change the way i perceive myself.
curious how much you care actually
>>
File: file.png (73 KB, 878x703)
73 KB
73 KB PNG
maybe it's time to accept im just not cut out for this
>>
>>107998690
another interesting topic, another braindead problem written by a nocoder
>>
File: Mematic_20240618_223728.jpg (100 KB, 1200x1200)
100 KB
100 KB JPG
>>107998664
from anthropic ? really ? i thought they were going to be the guys red teaming ai psychosis
>>
>>107998705
okay this problem isn't TOO bad if you realize you need to use a graph
but how the fuck would you realize it's unsolveable without a fucking graph????
>>
>>107998712
any shatbot is gonna go in the direction of what you say
you have to take extra care to avoid it becoming a yes man and hallucinating garbage
>>
>the shartnigger is still CAMPING THE THREAD
do you even code or is this your employment
>>
>>107998664
You severely overestimate the effort I put into calling out your goalpost-moving.
>>
/g/ was already dying but GenAI has rapidly accelerated the death of this board.
[spoiler]/dpt/ started being shit once we stopped using the yuki picture[/spoiler]
>>
>>107999584
>>107997507
>>
Is 60 too late to learn programing?
>>
File: h_335.jpg (26 KB, 642x427)
26 KB
26 KB JPG
>>108000411
>not using debug.exe to write .com files
ngmi
>>
>>108000411
Never too late. What ideas do you have, whats the goal?
>>
is teachyourselfcs.com the best reading list to get a deeper understanding without going to school?
>>
File: maxresdefault (10).jpg (87 KB, 1280x720)
87 KB
87 KB JPG
>>108000569
here's my reading list:
'(SICP)


....get it? cause its a list.

https://www.youtube.com/watch?v=-J_xL4IGhJA&list=PLE18841CABEA24090
>>
>>108000646
well teachyourselfcs recommends that as an introductory book as well so i guess you probably agree with it
>>
>>108000569
No.

And for the record, school won't teach you shit either. You are doomed to forever use interfaces that are complete shit without ever realizing how and why they're shit.
>>
>>108000668
well what's a better reading list?
>>
>>108000676
There is none. We are cooked.
>>
>>108000686
if there is no better reading list than i guess it is the best reading list by definition
>>
>>108000471
funny how this question never gets any answer
>>
>>108000691
Yeah, except no knowledge is better than false knowledge.
>>
>>108000676
>well what's a better reading list?
>>doesn't even hacking the planet.
ngmi
https://www.youtube.com/watch?v=4U9MI0u2VIE
>>
File: littlemler.jpg (222 KB, 1280x1652)
222 KB
222 KB JPG
with the SML flag in the background, this looks a bit like the camels surrendering after a war. The bowed heads, the sanpaku eyes on the center elephant.
>>
>>107997678
>just use languages that have performance obstacles built into the language
>>
>>108000865
Which languages don't?

Which languages allow you to provide your own ABIs, or give hints about readable/writable ranges, or allow you to drop branch hints (without compiler extensions)?
>>
>>108000883
>zig
>rust
>ada
>carbon / cpp2
>>
>>108000963
Examples for each of these?
>>
>>108000999
>>>zig
 const std = @import("std");

// 1. ABI Control: explicitly setting calling convention
// options: .C, .Stdcall, .Naked, .Signal, etc.
fn kernel_entry() callconv(.Naked) void {
asm volatile (
\\ mov $0x1, %rax
\\ ret
);
}

pub fn main() void {
var x: i32 = std.crypto.random.int(i32);

// 2. Branch Hints: 'std.builtin.expect' tells the optimizer
// that 'cond' is extremely likely to equal 'true'.
// this emits the llvm.expect intrinsic.
if (std.builtin.expect(x > 5, true)) {
// fast path: compiler arranges this block to be strictly sequential
} else {
// slow path: likely pushed to a different instruction page (cold)
}

// 3. Readable/Writable Ranges (via Alignment types)
// this pointer carries its alignment proof in its type signature.
// the compiler knows it can use aligned vector load instructions.
var y: i32 = 100;
const ptr: *align(16) i32 = @alignCast(&y);
}


>>>rust
 use std::num::NonZeroU32;

// 1. ABI Control: strictly defining the binary interface
#[no_mangle]
pub extern "C" fn my_c_callback(arg: i32) -> i32 {
arg + 1
}

// 2. Ranges / Layout:
// 'NonZeroU32' gives the compiler a "niche optimization" hint,
// allowing 'Option<NonZeroU32>' to be the same size as 'u32'.
struct LowLevelPack {
id: NonZeroU32,
// #[repr(C)] forces C-compatible memory layout, disabling field reordering
data: [u8; 16],
}

fn process_data(data: &LowLevelPack) {
if data.id.get() == 9999 {
handle_rare_error();
}
}

// 3. Branch Hints (Stable Way):
// marking a function as 'cold' tells the optimizer
// that calls to this function are unlikely.
#[cold]
fn handle_rare_error() {
// the compiler moves this code away from the hot path
}
>>
>>107998690
what a shitty problem
>>
>>108000999
>>>ada
 with System;

procedure Hardware_Control is
-- 1. Ranges:
-- This isn't just a check; it tells the compiler to use
-- the smallest possible register width (e.g., 8-bit)
-- and allows it to assume values outside this are impossible.
type Voltage is new Integer range 0 .. 12;

-- 2. ABI / Convention:
-- 'Import' binds a foreign symbol, 'Convention' dictates the ABI.
procedure System_Reset;
pragma Import (C, System_Reset, "external_reset_func");
pragma Convention (C, System_Reset);

V : Voltage := 12;
begin
if V = 0 then
-- 3. Optimization Hints (via Pragmas):
-- Ada doesn't have a 'likely' keyword, but it uses 'Inline'
-- and 'Optimize' pragmas to aggressively hint behavior.
System_Reset;
pragma Inline (System_Reset);
end if;
end Hardware_Control;


>>> cpp2 / cppfront
 // cpp2 syntax is "suffix-heavy"
main: () = {
// 1. ABI / Parameter Modes
// 'in': guaranteed read-only, passed by optimal register/ref
// 'out': guaranteed write-only (must be assigned before return)
// 'inout': read-write reference
process_data: (x: in int, y: out int) = {
// if you try to write to 'x', it's a compile error.
// if you return without writing 'y', it's a compile error.
y = x * 2;
}

// 2. Branch Hints (via C++ attributes)
// cpp2 passes attributes through to the underlying C++ compiler
[[likely]]
if (x > 0) {
// hot path
}

// 3. Range/Narrowing Safety
// explicit casts are required for lossy conversions
val: i32 = 100;
// short_val: i16 = val; // ERROR: implicit narrowing
short_val: i16 = val as i16; // OK: explicit "as" cast
}
>>
>>108000999
>>>carbon
 package Explorer api;

// 1. ABI / Interop
// Carbon can import C++ headers directly, matching their ABI layout exactly.
// This is unique; no other language does this natively without "bindings."
import Cpp library "geometry.h";

// 2. Arbitrary Bit-Widths (Ranges)
// You can define integers of *any* bit width, not just powers of 2.
// Useful for packing data into weird hardware registers.
fn PackData(input: i32) -> i7 {
// Logic to compress 32-bit int into 7-bit int
return (input % 128) as i7;
}

fn Main() -> i32 {
var x: Cpp.Vector(i32) = (1, 2, 3);

// 3. Optimization
// Carbon doesn't have a stable "likely" keyword yet,
// but it relies on LLVM structure similar to Rust.
// (Experimental syntax often changes here)
if (x.Size() > 0) {
return PackData(x[0]);
}
return 0;
}


>the verdict on the new kids

Cpp2 is for when you want the safest possible C++ ABI without thinking about it. it removes the foot-guns of parameter passing.

Carbon is for when you need to share memory layouts with existing C++ codebases without writing wrapper glue.
>>
>>108001023
>.Naked
No further questions, you failed.
>>
>>108001108
>.Naked implying absolute control over the stack frame is a "failure"

You asked for languages that allow you to provide your own ABIs. callconv(.Naked) is the literal definition of this. It strips the compiler's standard prologue/epilogue and hands you the raw instruction pointer. You don't just "hint" at an ABI; you construct it instruction-by-instruction in assembly.

If you can't handle the void where the safety rails used to be, that's not a failure of the tool. It's a filter.

Zig lets you walk into the reactor core without a suit. If you wanted a "safe" custom ABI, you should have stayed in the extern "C" playpen.

Check 'em:

>ABI: You define it (via ASM).

>Branches: You write the jne/je yourself (ultimate prediction control).

>Ranges: You enforce them by crashing if they're wrong.

Do you want to see the assembly output of a .Naked function vs a standard one to see what you're afraid of?
>>
>>108001135
>You asked for languages that allow you to provide your own ABIs
And naked doesn't do that. Naked tells the compiler that you're now working *outside* the language.

Didn't read the read of your cope. You failed, don't bother responding.
>inb4 you'll do it anyway
>>
>>108001149
>Naked tells the compiler that you're now working outside the language.

Fair point. I conflated "manual implementation via assembly" with "declarative definition within the type system." You wanted the ability to define register mappings and stack offsets syntactically, not just an asm block escape hatch.

You got me.

>inb4 you'll do it anyway Always.
>>
>>>holyc
 // strictly binds variable 'i' to register R15. 
// no "hints". you own that register now.
I64 reg R15 i;

// you can arguably create your own calling convention
// by just pinning args to specific registers for a function scope
// and manually jumping.


>>>forth
>>>c--
>>>sbcl

>sovereign verdict

if you want to feel the wind in your hair (and the panic of a kernel panic):

>HolyC is the answer if you want to rule the machine like a king (Ring 0, 64-bit only, no guardrails).

>Forth is the answer if you want to build the machine from scratch (you define the stack physics).

>.Naked was just the tutorial. Register pinning in HolyC is the final exam.
>>
 // THE SOVEREIGNTY LOOP
// No stack frames. No safety checks. No context switches.

U0 GodMode_Write()
{
// 1. REGISTER PINNING
// We strictly force 'val' to live in the CPU's R15 register.
// The compiler is forbidden from spilling this to the stack.
I64 reg R15 val = 0;

// 2. RAW POINTER ARITHMETIC (IDENTITY MAPPED)
// In TempleOS, address 0x0 is valid.
// We are grabbing a pointer to the low-memory interrupt vector table
// or just arbitrary low memory.
U8 * reg R14 ptr = 0x000B8000; // Classic VGA Text Buffer start

// 3. SILENCE THE GLOWIES (Disable Interrupts)
// 'CLI' clears the interrupt flag.
// The mouse stops moving. The keyboard dies. The timer stops.
// The OS scheduler is now dead. Only this thread exists.
asm {
CLI
};

// 4. THE WRITEOUT
// Writing directly to video memory without a driver.
// We use the XOR logic to create "static" on the screen.
for (val=0; val<2000; val++) {
*ptr = val ^ 0x55; // Write character
ptr++;
*ptr = 0x4F; // Write color (Red on White)
ptr++;

// Manual delay loop because the system timer is dead
asm {
MOV RCX, 10000
@@loop:
DEC RCX
JNZ @@loop
};
}

// 5. RESURRECTION
// Re-enable interrupts. The OS wakes up, unaware it was ever dead.
asm {
STI
};
}
>>
File: compilers_broken.png (52 KB, 869x756)
52 KB
52 KB PNG
>>108001203
>no "hints". you own that register now.
Yeah, that's ALSO not what I want - and not just because of the obvious problems. I just want to be able to tell the compiler that parameter 1 goes to RAX, parameter 2 goes to RDX, and return goes to RCX.
>>
>>108001230
>>>High Level Assembler (HLA)
 // You don't "hint". You demand.
procedure HostileABI( val1: dword in eax; val2: dword in rdx );
@returns( "rcx" ); // Force return value in RCX
begin HostileABI;
// val1 is physically in EAX here.
// val2 is physically in RDX here.
mov( eax, rcx );
add( rdx, rcx );
end HostileABI;


>>>SBCL (Steel Bank Common Lisp)
 (define-vop (my-hostile-call)
(:translate my-hostile-call)
;; Binds 'x' strictly to RAX, 'y' to RDX
(:args (x :scs (rax-reg) :target r1)
(y :scs (rdx-reg) :target r2))
;; Forces result into RCX
(:results (r :scs (rcx-reg)))
(:generator 1
(inst add rcx rax rdx))) ;; The compiler emits code assuming these locations


>>>C-- (The Glasgow Haskell Compiler backend)
 // Defines a call where arguments are pinned
foreign "C" MyWeirdFunction( "rax" arg1, "rdx" arg2 ) {
// ...
return "rcx" (result);
}


If these aren't what you want, you are asking for a language that doesn't exist yet, and you should probably fork LLVM to create extern "anon_convention".
>>
>>>The Answer: Open Watcom C/C++
 // 1. Define the hostile ABI contract
// parm [rax] [rdx] -> Input parameters map to RAX, RDX
// value [rcx] -> Return value maps to RCX
extern long HostileAdd(long a, long b);
#pragma aux HostileAdd parm [rax] [rdx] value [rcx];

// 2. The Implementation
long HostileAdd(long a, long b) {
// The compiler knows 'a' is in RAX and 'b' is in RDX.
// It will generate code that puts the result in RCX.
return a + b;
}

// 3. The Call
void main() {
// The compiler emits:
// MOV RAX, 10
// MOV RDX, 20
// CALL HostileAdd
// (Result is now in RCX)
long result = HostileAdd(10, 20);
}
>>
can you schizos please just SHUT THE FUCK UP
>>
>>108001275
why what's the problem anon
does your head ache

or are you just sick of looking at progress
>>
>>108001275
it's just another spambot deployed by the m**dnigger in an attempt to drive out people from this thread after his schizo avatarfagging ass got banned yet again
>>
>>108001340
right, because spambots push repos on flat earth and unified field theory. great inquisition
>>
File: PXL_20260129_095659400~2.jpg (947 KB, 2732x2268)
947 KB
947 KB JPG
>>107998021
Ok so image capturing is working.
Lets try to build a DMA so we can get big boi resoultion easily
>>
implemented a O(1) joint optimization solver for my stereo speaker crosstalk cancellation plugin

it works well enough, probably will ship the update next week
>>
File: file.png (256 KB, 1293x1305)
256 KB
256 KB PNG
>>108001577
honestly it's as retarded as it can get

the code gets to the text limit but the gist of the logic is this
>>
File: file.png (187 KB, 688x1172)
187 KB
187 KB PNG
can a nigga live
>>
>>108001825
No.
>>
>>108001825
i'm a bit jelly desu.
Sometimes I think I should do some excessive LC or similar stuff to revisit things and learn a couple more, but i just dont have the patience to do mundane tasks.
For example I have implemented dijkstra before, but i can't do it out of my head. It would be nice if I could. I know it was fairly simple. Also never really did much DP. Neither in real life, nor in kode

A few years ago i had some fun with this https://cses.fi/problemset/ and also project euler, but they are not very theoretical
>>
>>107996992
You think that the reason why the windows menu is slow is because the programmers who wrote that were too hardcore and too focused on microoptimizations?
That has got to be the most retarded thing I've ever read on here
>>
>>108001928
I mean, we *know* they haven't been doing microoptimizations since x64.
Despite their TPM 2.0 mandate they still use SSE.
>>
what's all this about start menu lag? multiple seconds? my shitbook runs it fine and it's practically a war casualty. i've literally never had an issue with windows 10 failing that isn't something i did to myself or some kind of insane hardware crash. i guess you create your own problems as a neckbear
>>
AI is insane, 90% of my code is just AI slop now and its actually fucking good bc I'm senior dev and I know our work project (~600k LoC) pretty well. I'm fucking shocked. I just copypaste jira tasks in the codex (not even in English) and it somehow find correct classes that needs to be modified, I do a couple checks and fixes and the task takes usually a few hours takes 5 minutes. I'm seriously considering to get a secret 2nd job now.
>>
Why is Python logging so shit?
I have a basic Litestar (FastAPI with a higher bus factor) application that I want to have output consistent semi structured logs with contextual variables. I am led to believe that this is functionally impossible or completely undocumented, because both the framework and the application server use their own logging frameworks with poor visibility or control.
Yet this is trivial in JVM land, despite Python's logging having been standard library for 20 years, and Java having three standard loggers in that time.
>>
what stops me from being the best? no but really what really stops me from being amongst the biggest?
>>
>>108002040
you got fucking Global Interpreter Locked Baby Cakes. don't feel bad. happens to the mest of us.
>>
>>108002042
Your refusal to learn assembly.
>>
>108002040
>Python logging having been standard library for 20 years
>That's exactly the problem.
>It's a fossil from 2001 designed for single-threaded scripts.
>In async land, the context gets lost the moment you await because standard logging doesn't know about contextvars (the Python equivalent of Java's ThreadLocal/MDC).

You are suffering from the Global Configuration Paradox. Uvicorn, Litestar, and your app are all fighting to configure the root logger.

The fix is structlog. It decouples the emission from the formatting and handles the async context context bridging.

Here is a working PoC that hijacks the Uvicorn/Litestar loggers and forces them into a unified JSON stream with contextual binding.

https://pastebin.com/MzGQECEF
>>
>>108002032
>>107997507
>>
>>108002101
>>108002040
>>
>>108001203
Nice b8, is this Assitant_Pepe_8b?
>>
>>108002032
ok, what does this have to do with programming? take it elsewhere
>>
File: file.png (129 KB, 1715x744)
129 KB
129 KB PNG
>recursion over graphs
I can't explain the shit I wrote but it works
>>
>>108001875
I hated LC at first but I started to enjoy it a lot after solving ~100 problems. I have a backlist of projects that I need to work on but my adhd keeps preventing me from committing to shit for longer than an hour, so LC is perfect for a daily coding fix.
I can work on these problems for hours at a time because I'm always doing something different, but if you asked me to fix some bugs in a massive app I would shit my pants
>>
>>108002814
this sounds more like a trap desu
>>
File: ok.png (142 KB, 640x480)
142 KB
142 KB PNG
>>108001552
Didn't do DMA, because the documentation is shit.
Improved stuff a bit and made the picture bigger.
Now I'll move towards either DMA or image transformation. Will see
>>
>>108002814
Grim
>>
File: diana cavendish.jpg (19 KB, 370x315)
19 KB
19 KB JPG
>>107996387
bro idk what is this supposed to be but I love this game, does it belong to any genre?

>>107998195
I think he made it that way so its harder to answer
>>
File: cool.png (3.27 MB, 4008x2900)
3.27 MB
3.27 MB PNG
>>108002814
Well, for me it doesn't matter what it is. If it's a long project or tiny tasks each day, when I really commit to it, then I just get obsessed with it.
Like my current getting MIPI cameras to work on an FPGA. Been sitting here for 12+ hours each day to get this shit working.
But now that I got pictures, I feel not that motivated to keep on pushing. I basically solved it, now it's just improvement. At least today I've been distracting myself from improving the camera image configuration, instead of making the pipeline better.
But yeah, one day ill obsess over LC too, I hope.

Here have an image of my current circuit.

Diagnosed ADHD btw, no tiktok diagnosis. And suspect of autism.
>>
File: 1768720328535133.jpg (351 KB, 1440x1440)
351 KB
351 KB JPG
My company isn't really greenlighting us AI assistance for general SWE work (i know). I'd like to use cursor or some such. How stupid of an idea is it to rip the codebase off my work laptop to my personal computer and maintain working versions there using cursor? My retarded idea is to transfer the code via my overleaf account to avoid a paper trail. I just refuse to write my shit by hand and I cant think of anything smarter
>>
File: 1769532429559546.png (110 KB, 600x900)
110 KB
110 KB PNG
>>107998195
>>108003529
haha thanks for trying it. ya someone posted an IQ test thing in another thread but I was too slow to reply before the thread died. honestly I want to make it a bit harder by rotating the answer results randomly maybe? the actual IQ test the blocks can essentially lock to other blocks so they're floating. maybe I'll do that too. I definitely gotta make the camera angle a little better.
>>
>>108004054
don't do it. depending on the company they'll track you. bigger companies have compliance people and will auto detect you uploading files. you'll get a warning for uploading like even just text to open source github PRs because they're worried about leaking data. nottt wortthhh.
>>
>>108004140
Is the joke that you got it wrong??
>>
>>108004229
So what am I supposed to do?
>>
anyone got a good resource for someone tryna write their own utf-8 encoder/decoder (research purposes)
>>
>>108002814
Puzzles are fun, self-contained problems I can work on in short bursts, where the goal is already defined for me.

Actual projects that I have to make up on my own and make decisions for myself are impossible. Something is wrong with me and I hate myself.
>>
>>108004054
AI code is very obvious.
If your employer isn't pushing you to use AI then you're probably a defense contractor and going to tank serious charges for leaking the code.
> I refuse to write my shit by hand and
therefore, because the rule of the body is "use it or lose it" and this applies to the brain as well,
> I can't think of anything smarter
>>
>>108004366
I feel generous today:
typedef union
{
struct
{
uint32_t byte0:6;
uint32_t byte1:6;
uint32_t byte2:6;
uint32_t byte3:3;
}mode3;
struct
{
uint32_t byte0:6;
uint32_t byte1:6;
uint32_t byte2:4;
}mode2;
struct
{
uint32_t byte0:6;
uint32_t byte1:5;
}mode1;
struct
{
uint32_t byte0:7;
}mode0;
uint32_t utf32;
}utf;
>>
>>108004384
wh-why would you use a u32 with a 7-bit bitfield?
>>
>>108004384
that is this?

Code point range UTF-8 Format (x reserved for code point) Total Bits Bits for code point Inflation
0x00 - 0x7F 0xxxxxxx 8 7 8 : 7
0x80 - 0x7FF 110xxxxx 10xxxxxx 16 11 16 : 11
0x800 - 0xFFFF 1110xxxx 10xxxxxx 10xxxxxx 24 16 24 : 16
0x10000 - 0x10FFFF 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx 32 21 32 : 21

thanks
>>
>>108004402
Just because I like having the same type as utf32.

>>108004410
Yeah. You have to shift each value by full bytes though:
        if(utf.utf32 > UTF8_3)
{
*(uint32_t*)output
= ((0xF0 | utf.mode3.byte3) << 0)
| ((0x80 | utf.mode3.byte2) << 8)
| ((0x80 | utf.mode3.byte1) << 16)
| ((0x80 | utf.mode3.byte0) << 24);
output += 4;
goto LABEL_CONTINUE;
}
if(utf.utf32 > UTF8_2)
{
*(uint32_t*)output
= ((0xE0 | utf.mode2.byte2) << 0)
| ((0x80 | utf.mode2.byte1) << 8)
| ((0x80 | utf.mode2.byte0) << 16);
output += 3;
goto LABEL_CONTINUE;
}
if(utf.utf32 > UTF8_1)
{
*(uint16_t*)output
= ((0xC0 | utf.mode1.byte1) << 0)
| ((0x80 | utf.mode1.byte0) << 8);
output += 2;
goto LABEL_CONTINUE;
}
*output = utf.mode0.byte0;
output += 1;
LABEL_CONTINUE:
continue;
>>
>>108001036
>>108001027
>>108001023

memelangs btfo
shatbot btfo
namefag in shambles and coping
>>
>>107996387
why did you obscure this image with a shitty moeblob
>>
>>108000883
>or allow you to drop branch hints
If you are writing new code, there is literally no reason to use these. The usecase for branch hinting is extremely niche to begin with because the compiler pretty much never has any optimizations to apply. Just design your control flow better.
>>
>>108004054
you don’t deserve that job. you should be working at mcdonalds
>>
File: kot-competitive-petting.jpg (235 KB, 1080x1080)
235 KB
235 KB JPG
>>108000883
>without compiler extensions
yeah sure
>compare our memlangs to c, but from 1990
why doe?
bc you get btfod otherwise?
>>
File: curvy-averi.jpg (368 KB, 912x1136)
368 KB
368 KB JPG
>>108004703
i think i understand why they hate c.
our language is really effed in learnability

there isnt a book that, if you read it from a to z, you will learn everything there is to know about c
because theres the standard
then theres how c inyteracts with an x86-64
then theres compiler extensions
and then theres bare metal programming

they think its elitism
theyre filtered, so theyre jelly
they misunderstand the phenomenon, so they hate

i love c bc all it takes to learn it is just using it.
i learn all this with 0 effort because everything is
>hmmm. i wonder if i can do that...
>...
>noice. i just have to do xyz

but if one wants to learn everything c from a top down perspective,
yeah, shits fucked, i get it.
>>
>>108004816
who the fuck still uses C
>>
>>108004703
>there is literally no reason to use these
That opinion says a lot about you, and none of it is good.
>>
>>108001928
>You think that the reason why the windows menu is slow is because the programmers who wrote that were too hardcore and too focused on microoptimizations?
No, and at no point of that post, or the broader conversation, is that a sound conclusion to draw.
>>
File: ted-ka-C-zynski.jpg (128 KB, 760x535)
128 KB
128 KB JPG
>>108004908
opencl and cuda is basically c.
opencl is even called openclC, ackshually
id expect scientific shit on supercomputers gets written in c
i hear embedded gets written in c
rtoses get written in c bc c is pretty transparent about things
so, aerospace, satellites, cellphone shit, medical shit
python libs get written in c.
your router's firmware has most likely been written in c.

c is good for hacky shit.
other languages feel like doing surgery while wearing boxing gloves
>>
>>108004988
You don't even know what branch hints exist to do, you couldn't infer my foot from your ass.
>>
>>108005069
Fascinating headcanon.
>>
>>108004366
>>108004384
>>108004402
apologies for any mistakes, but those u32 bitfields really were triggering me, so i had to quickly bang something out: https://pastebin.com/NBg4hgdQ (too big to fit into a single post)
>>
Which cli tool and local model do you use for vibe coding?
I just tried vibe cli with devstral-2-small-q4 but it didn't even work because the model template is broken.
I also tried opencode with one of the free models but it got stuck fixing some word in AGENTS.md so I just quit.
>>
>>108005301
I use claude code sometimes and it's so dumb that it makes me want to kill myself. If I used a local model that's even dumber than that I probably wouldn't be here to write this
>>
i remember these threads being fun. i just came back after a long time and see ai bots spamming schizo gibberish, and people doomposting about ai being a better programmer.
>>
File: 4458174.jpg (28 KB, 460x460)
28 KB
28 KB JPG
>>108005301
Speaking of this. Has anyone tried out Prime's thing? He has this neovim vibe code thing where it basically just does single functions and stuff. I'm going to give it a shot after work. In theory I really do like the idea a lot. In practice I have no idea if its straight better than regular ass vibe coding slop. Maybe the idea is you fix it as you go so its never slop slop?

https://github.com/ThePrimeagen/99
>>
>>108004816
And here I was thinking C is small language. Small so not a lot of stuff to learn is what people have been telling me about C.
>>
>>108005650
C is small. anon is being retarded and conflating assembly/platform-specific ISAs with C for... some reason
C is dead simple
>>
>>108005650
if you only use languages with hashtables you can burn a whole career without implementing one.
if you use a language without hashtables you'll more readily learn something about them.
C is bad but not because there's "not a lot of stuff to learn". Programming is a broad field and you can do anything in anything.
>>
>Programming is a broad field and you can do anything in anything.
It's actually wild how deep the hole goes. You can go from monoids haskell all the way down to assembly and keep digging into you get to math and hardware. But once the math gets hard enough you loop back around to monoids and haskell. It's like how planes have to fly through the earths core to reach australia.
>>
>>108005301
pipe shit to ollama run. It's a one-line binding in your ~/.vimrc ... and a short script because ollama's CLI kinda sucks
// foo.slop
int count_r_in_word(char *word) { }
void another_function(void) {}
" .vimrc
au Bufread,BufNewFile *.slop map <C-p> :%!slop<Return>
# ~/bin/slop
#!/bin/bash
( echo "Respond with only code that expands on the code provided. Be terse. No markdown.:"; cat ) | ollama run qwen3-coder 2>&1 | perl -pe 's/\e\[[0-9;?]*\w//g; tr/[^\x01-\x7F]//'

visually select the first function, hit ctrl-P
foo.slop result:
int count_r_in_word(char *word) {
int count = 0;
for (int i = 0; word[i] != '\0'; i++) {
if (word[i] == 'r' || word[i] == 'R') {
count++;
}
}
return count;
}

void another_function(void) {}

... plus some of the spinner that didn't get removed by the perl. fix it yourself.
>>
>>108005301
llama.cpp comes with llama-server which you can easily connect to opencode and it just works. been using qwen3 coder 30b a3b 2507 gguf 4bit version. works ok
>>
>>108004908
linux, ffmpeg, the JVM
>>
>>108005334
One day I got lucky and it implemented 2 complex functions well, I started overhyping it.
I still think it helps me personally, but I now get so much slop from my coworkers to fix, it's just sad.
>>
>>108006114
And the worst thing is that they get 80% there in a short time, so they expect me to just write a single prompt and fix the remaining 20% in 5 minutes. If AI failed to fix if for hours, what makes you think I have the magic prompt that will make ai unfuck it? If the fix requires real programming it will take as long as real programming does.
>>
this is the daily programming thread not the vibe coding thread
>>
File: hit.webm (227 KB, 696x352)
227 KB
227 KB WEBM
>>108004908
>>
>>108004999
>id expect scientific shit on supercomputers gets written in c
Or C++ or Fortran.
Or Python because who really gives a shit about performance. (This works when the Python is just coordinating some C code behind the scenes.)
>>
>>108006225
I would beat the shit out of them
>>
xoomers be like
>yeah zoomie u need to learn this dead language that's exclusively used in backend and low level systems ull never be able to get a job in because they all require 20 yoe or are hobbyist dogshit that doesn't pay
i think ill stick to python and javascript, thanks
>>
>>108006283
i know this is bait, but i'll still bite. there are plenty of entry level C developer jobs paying 6 figures
i only have a few myself, and i'm making roughly 150k, and i'm not in a particularly well-paying industry
>>
Based AI programmers, you now have your own general.
You're welcome.
>>108006287
>>108006287
>>108006287
>>108006287
>>108006287
>>108006287
>>
>>108006310
those threads never get more than 5 replies
>>
>>108006342
yeah I know. can't hurt to try though
>>
>>108006225
>Maybe Khans kill people without looking them in the face, but I ain't a fink, dig?
>>
>>108006310
good idea, hope it takes off. this really isn't the right place for most of these AI posts I see here as they almost never have any code or programming concepts to discuss, just some guy coming in "wow claude code did this for me" yea thanks for the info
>>
>>108006310
needs an anime girl in the OP, not a jeeta
>>
>>108006595
you're vibe coders
>>
>>108006595
don't try to rise above your station, slopslinger
>>
reflection and reification in mainline c++ compilers when
i don't want to have to use the bloomberg fork
>>
>>108006820
reflection is a modern programming requirement
so quit moaning
>>
>>108006820
>reflection
What's the ideal syntax? something like reflect(struct mystruct).fields ... ?
>reifiction
what's that?
>>
>>108006885
>What's the ideal syntax? something like reflect(struct mystruct).fields ... ?
reflect(thing) was the proposal syntax but for the actual syntax, they're going for is ^^ as a prefix operator to get a metaobject because apparently ^ is already used by fucking apple's bootleg lambda function syntax for objective C and oh no we can't possibly hurt the unofficial language of objective C++
then library functions that wrap compiler builtins for queries, so something like std::meta::fields(^^struct mystruct) (although I think it's a different function iirc)
>>reifiction
>what's that?
opposite of reflection, static code injection using reflected metaobjects
it was originally included as part of the metaclasses proposal but was moved to reflection
it only really makes sense in languages with static reflection or AOT reflection compilation
the syntax they've gone with is ugly as sin [: :] in order to inject the unreflected thing and [# #] i think to inject a metaobject's name but it has syntax at the very least
it makes template metaprogramming completely obsolete
>>
File: use case ???????.png (113 KB, 1037x1009)
113 KB
113 KB PNG
>don't have a CS degree
>every day I learn about some new autistic algorithm named after some professor that optimally solves a very specific problem (kahns, kadane's, knuth morris, floyd warshall in this case)
>>
File: 1741792066240061.png (219 KB, 432x621)
219 KB
219 KB PNG
>todays problem
https://leetcode.com/problems/minimum-cost-to-convert-string-ii/description/
>>
>>108006595
As long as it's AI generated, it fits the theme of that thread (slop for the slop god).
>>
>>108006963
>use case ????
bioinformatics
>>
>>108006953
>reflect(thing) was the proposal syntax
nice
>std::meta::fields(^^struct mystruct)
ewww
meta is a sick keyword but metaprogramming is very broad and it doesn't feel right to claim "meta" just for reflection

>but for the actual syntax, they're going for is ^^ as a prefix operator to get a metaobject
jesus chirst, I feel like I'm reading about Raku
>.^method
https://docs.raku.org/language/typesystem
https://docs.raku.org/language/mop

>because apparently ^ is already used by fucking apple's bootleg lambda function syntax for objective C and oh no we can't possibly hurt the unofficial language of objective C++
fucking hell. collusion much?

I still don't get reification. I understand static code generation and how this could be better than templates but I don't get how code generation is the related to or the opposite of reflection, or related to metaclasses.
Metaclasses are one of the most retarded thing in PLs. When it comes to metaprogramming, they don't do anything, don't solve anything, don't have any semantics. They are empty shells made by midwits.
>>
>>108005679
>C is dead simple
Not once you get to higher order pointers. When you see:
int *****v;/
then you definitely start thinking WTF's going on.
Yes, I've seen that in production code. That was code I really didn't want to change.
>>
>>108007487
>fucked the syntax
I must need sleep more than I thought.
>>
>>108007487
you should realistically never need to go above a triple pointer (address of an array of pointers). if you find yourself needing to, you've probably fucked something up
>>
>>108007487
how did you even manage that? can you nest code blocks or something?
int x;
int nested;  // ?
;
int y;
>>
test;/test
>>
test zero
test one;/
test two
test three

test four

test five
>>
before
long line long line long line long line long line long line long line long line long line long line long line long line long line long line long line long line long line long line long line long line 

after
>>
File: embedded code block.png (107 KB, 1663x801)
107 KB
107 KB PNG
>>108007804
>>
>>108007481
Reification is the process of taking a value and turning it into code structure. Reflection turns code structure into values.
When data is code, reification and reflection are the same. You can turn a value into a class (reify) and shove that class into the symbol as the new thing to be interpreted. You can grab a class value and access a list of its methods, its members, if you can access them or not (reflection).
When compiled languages talk about reification they usually mean "we do one compilation stage to generate code for the second compilation stage", as the code structure is gone when the program is finally compiled.
>>
In C is it ok to use gotos where I would use recursive tailcalls in other languages? I mean, obviously I can do whatever the hell I want in my code, but is it anathema?
>>
>>108008400
If you can't fall through the end of the loop, or use a continue statement, yes of course you can.
>>
>>108007481
>meta is a sick keyword but metaprogramming is very broad and it doesn't feel right to claim "meta" just for reflection
meta isn't a keyword, it's a namespace, the function fields would be under it, so it isn't just necessarily for reflection, although that's all it's being used for now

reflection gives you a something representing something in the AST (metaobject)
reification takes a metaobject and turns it back into something in the AST

in the C++ reflection proposal this would just take the form of injecting arbitrary stuff into places via reflected data, it's literally the thing that fills out a "template (not necessarily a c++ template)" as opposed to have to the very limited and extremely retarded shit you could do by like functional programming using templates
it's hard to explain but the full text of the proposal is here https://isocpp.org/files/papers/P2996R13.html and the reification operators are called splicers because they just had to come up with a new name for it

and the C++ metaclasses proposal defines metaclasses fairly well, a metaclass is just syntax sugar for a compile time function that takes a "fake" object as a parameter via its reflection and reification would be used to inject whatever was to constitute what would actually be the "real" object with the same name (while at least in the example compiler implementation the fake one then be inserted as a declaration under that scope and would be renamed something like metaclass_original)
void example_metaclass(std::meta::info /* the reflection metaobject */ info) {
//stuff
}
class object(example_metaclass) {
//body of the "fake" object to be used to programmatically generate code
};


just a simple one pass transform and the advantage it has over reflection/reification on their own is that it replaces and renames the original while reification/reflection proposals type introduction functions can only introduce new names when they introduce names
>>
>>108008400
i use gotos all the time. they're perfectly fine to use in moderation, but as much as possible, i would recommend only ever putting labels at (or near) the beginning and end of your function. that will solve about 99% of all goto related concerns about spaghettification
>>
>>108004738
>but from 1990
Are you on drugs?
>>
>>108007904
>>108008456
>meta isn't a keyword, it's a namespace
I know. I was contemplating how it would be to have meta(type).fields / meta(value).fields instead of reflect(type).fields) / reflect(value).fields.
>the function fields would be under it, so it isn't just necessarily for reflection, although that's all it's being used for now
I see, but then again I don't see for what else you would use it outside of reflection.

Also it doesn't make sense that fields is a function, it's not, it's a special hook in the compiler.

>Reification is the process of taking a value and turning it into code structure. Reflection turns code structure into values.
>reflection gives you a something representing something in the AST (metaobject)
>reification takes a metaobject and turns it back into something in the AST
>it's hard to explain but the full text of the proposal is here https://isocpp.org/files/papers/P2996R13.html and the reification operators are called splicers because they just had to come up with a new name for it
Now I understand but to me to me it's all reflection. God this is stupid, I knew the symmetry was bullshit.

The way I would implement reflection is that when you do reflect(type) or relect(name), you get a user level value with a specific user level type, a struct that contains everything of interest. It doesn't make sense to split this is 2 (what is there even to split).

>//body of the "fake" object to be used to programmatically generate code
Yeah I don't believe that. I've heard several times about metaclasses, often code generation is mentioned and it always either a lie or a lack of imagination.
Unless there are metaclasses for AST nodes, all you're going to be able to generate it types, at best. And even then I doubt you'll be able to generate arbitrary types because you also need an AST to express type definition.
>>
I finally finished implementing custom allocators and heap compaction to reduce my text generator's memory usage and reduce heap fragmentation. I'm using a bitmap-based allocator, with 2 byte slots, for allocations < 32 bytes and a free-list based allocator for allocations >= 32 bytes. I'm also using best-fit allocation to try to improve heap utilization before compaction. The current issue is it is way too slow. Pool size for the small allocation pool is 256 kiB and the pool size for the large allocation pool is 4 MiB. I think the primary bottleneck is best-fit allocation since it requires either searching the whole space or maintaining a sorted list of free holes in ascending order by size, similarly I keep the list of pools in order by increasing unused space. I'm also doing coalescence of the free-list nodes on every free, which might be too aggressive as it could be done only on failure to allocate (if it has not been performed since any frees occurred, etc.), and because the free-list is sorted by increasing size and not by offset, I have to search the entire free-list during coalescence. RB-tree's can make coalescence much faster, but the extra size overhead per free-list node is not ideal.
>>
>>108007804
>>108007814
incredible
>>
File: 1359438177782.png (1.02 MB, 1019x881)
1.02 MB
1.02 MB PNG
I know this has been asked a billion times before but is AI coding a meme? I don't like it but at the same time I don't want to be left behind.
>>
>>108009652
it is both shit and good at the same time and idk how the current business model for service providers is anywhere near sustainable
>>
>>108009652
fuck off >>108006287
>>
>>108009103
>Also it doesn't make sense that fields is a function, it's not, it's a special hook in the compiler.
a function that more than likely is going to wrap a compiler builtin (or at least it does in the reference implementation) in most implementations which is a special function that is a special hook in the compiler
in the case of fields (or rather members_of which is its real name) it takes a reflected declaration context (i.e. namespaces, classes, structs, unions, anything that can contain a sub-declaration) and returns a compile time array of reflections of its sub-declarations

>It doesn't make sense to split this is 2 (what is there even to split).
while they did kinda "unsplit" it in that reification is included in the reflection proposal and they dropped the name reification, splitting it conceptually still makes sense when you're dealing with static reflection, at least the way C++ intends on implementing it
reflected data is ephemeral and opaque outside of query functions and cannot exist at runtime at all
it has to be produced in very specific execution contexts, can only be consumed in even more limited execution contexts, and of those reification would be the most strict
>Unless there are metaclasses for AST nodes, all you're going to be able to generate it types, at best
in the context of the C++ metaclass proposal metaclasses are specifically for programmatically generating types from a provided skeleton type in a way that lets you avoid polluting declaration contexts with a bunch of names
there's functionality included in the reflection/reification proposal for building named AST nodes from scratch (including statically allocated strings which you could argue are "named" in a way) and unnamed AST nodes like expressions while unable to be built can be copied then unreflected in a different spot
>>
>>108009406
>>108009406
>I'm using a bitmap-based allocator
>I think the primary bottleneck is best-fit
If you use a bitmap, then you can search for consecutive blocks of zeroes. Of course it depends on the resolution of your blocks, but as long as the bitmap is consecutive it shouldn't be too slow - alternatively you can attempt to keep the position of the last allocation in your books so that you don't have to restart at the beginning if you're searching for a free bucket.
>>
>>108009406
Me and a couple other anons worked on a bitmap run finder a while ago, this is my implementation: https://files.catbox.moe/j6axom.txt

It can definitely be improved, there's still a big time spike in the smaller branches.
>>
>>108009652
realistically anything with actual value will still need real developers but yes it can make prototypes surprisingly well

but those prototypes cant be used in a real product later, it'll have to be written from scratch at some point by real devs
>>
i still continue to do nothing and achieve nothing its over
im lost
im done
there is no point anymore
>>
File: what_the_actual_fuck.png (42 KB, 1076x952)
42 KB
42 KB PNG
>>108011358
Wrong.
There is always informing people about and shaming developers' incompetence.
>>
>>108011397
Aw fuck you're back
>>
File: antiquated.png (23 KB, 1030x440)
23 KB
23 KB PNG
>>108011418
Was I ever gone?
>>
>this threads allows you to pick your schizo, just like Netflix
what a year to be on the internet
>>
Is it schizo if I have proof?
>>
i wish i would have gone to college
>>
>>108008400
>use gotos where I would use recursive tailcalls in other languages
Doesn't sound ok to me. Write your loop properly.
Only legitimate use for gotos that I've seen in C is for jumping to clean-up after error handling. And even that feels like there should be a better way.

But you know, this is just like my opinion, man. Do what you need to do to get your code working.
>>
File: 1752106579324559.png (57 KB, 240x240)
57 KB
57 KB PNG
claude is not a replacement for skill
I am a tadpole in a pond of sharks
>>
>>108011663
Welcome to the world. Start building skills.
>>
>>108011663
>sharks
Pffffff. Maybe sharks with wine gum teeth, if that-
>>
>>108011715
Maybe that anon is talking about (you)?

What are you even spamming about? I don't understand anything about what you're trying to show. I'm definitely not a shark.
>>
File: sovereign_serpent_coil.png (192 KB, 1000x1000)
192 KB
192 KB PNG
Quantum Sophia — Loom Edition Topological AI framework for deterministic cognition, negentropic memory, and 0.618‑aligned autonomous systems. Implements dimensional collapse, abundance heuristics, and physics‑agnostic simulation kernels.
>>
>>108011973
>What are you even spamming about?

>>108011397
This one shows a GetModuleFileNameW call reading information from the same process (0xFFFFFFFFFFFFFFFF) using NtReadVirtualMemory. Not always, but in almost every single case Nt and Zw functions issue a mode switch into kernel space, which are expensive - meaning that WinAPI code - which is supplied by Microsoft - switches between user and kernel mode several times *to read data from the same fucking process* ... when they could've just ... read the data directly.

>>108011450
This is AI essentially agreeing with me that there is no reason to be using malloc except maybe compatibility with systems that don't support paging.

>>108011475
This is the nVidia userspace driver performing the same UTF-8 => UTF-16 conversion of a static string containing all ASCII characters for no fucking reason whatsoever.
>they could've just turned their 0x01-0x7F string into a wide one and ship it statically

>>108011715
This is more WinAPI insanity, but you can see it in antiquated Linux interfaces (mmap) as well: instead of explicitly acquiring the lock for a resource once, then performing a bunch of operations on it, and then explicitly releasing the lock once, these nigger jews acquire and release the lock for every single fucking operation.
>but my other threads
Can suck a dick. I'm not gonna pay hundreds of cycles just because you fucking retards can repeatedly write 0x11 and 0x0 into a variable.
>>
>>108011475
how does the algorithm work?
>>
>>108012116
>utf8 to unicode
nevermind I thought it was the hex dumper
>>
File: RtlUTF8ToUnicodeN.png (19 KB, 840x780)
19 KB
19 KB PNG
>>108012128
The hex dump is just there to show that the string contains literally every single ASCII character. 0x01 - 0x7F.

>>108012128
>utf8 to unicode
I mean, sure, we can talk about how an RtlASCIIToUnicodeN function would be much more optimized because you'd only have to insert NUL bytes after every character, and how the assumption of UTF-8 complicates everything.
>>
>>108011663
lowkey stop asking claude to code for you. if you want to use AI you can have it pair study with you. read the fucking manual and your bot can also rtfm and then you can learn skills that you will remember
>>
>>108012235
the difference is that claude is not an agent
>>
>>108010628
Thanks, I will take a look. I need to run a profiler ultimately.
>>
>>108012249
they all look the same to me
>>
>>108012684
??? an agent will fix its own bugs eventually.
>>
New: >>108012888
>>
>>108006460
Kek



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.