Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Frame – Linux X server in Assembly

isene.org

Post reply on HN
Print view View original

101–110 of 115 posts

Re: Frame – Linux X server in Assembly

#101

Post by wojciii »

kibwen wrote: post #54
idiotsecant wrote: post #45

Earlier quoted context omitted.

There is evidence that LLMs are capable of making assembly that runs a great deal more efficiently than the compiler can manage on its own.

The purpose of an optimizing compiler is not merely to produce efficient assembly. The goal of an optimizing compiler is to produce efficient assembly while confidently preserving a program's observable logical semantics. Asking an LLM to spit out raw unstructured assembly based on inferred context from a specification given in English is a contender for one of the worst ideas I have ever heard; I award you no points…

Now imagine doing this for some system which actually matters and can cost lives if implemented incorrectly. That would be scary.

Re: Frame – Linux X server in Assembly

#102

Post by wojciii »

asveikau wrote: post #63

I can have a machine generated X server in assembly too. It's called cc -s. Or: $ objdump --disassemble /usr/lib/xorg/Xorg Maybe I'm getting too old for this, but I really don't see the point in having AI generate assembly code for this.

It's just you.

Everything that can be done with LLMs is better than without. Why use tooling at all when you just can use a LLM? Maybe just make the LLM generate machine code for your tager directly.

Just like making digital systems that are "better" than analog ones from the past because of digital.

Re: Frame – Linux X server in Assembly

#103

Post by cassianoleal »

yndoendo wrote: post #100
cassianoleal wrote: post #99

Earlier quoted context omitted.

I already see a lot of people say they've "read" a certain book when they mean they listened to the audiobook instead.

Words matter to me. This why I say "consumed" vs read. Audiobooks are great for walks and long car drives.

Yep, it's very weird. Consumed, listened to... There's ways to say the thing like it is. I guess "read" carries a certain superior connotation so people lean on it.

Note that I do think reading is superior. This is not to diminish anyone who chooses to listen to books. Some people do it because of accessibility, some because of time (can listen while commuting, or in the gym), or they just choose to listen to some books they don't care as much whilst still read the ones they do.

This is all great and people should be encouraged to do what works for them - but please don't pretend it's the same. Sometimes I even think we need a different word for reading e-books.

Re: Frame – Linux X server in Assembly

#104

Post by vidarh »

segmondy wrote: post #75

With the crazy lack of supply for hardware and ridiculous prices, seems we are going to have to start squeezing more juice out of our existing hardware. What I would love to see is how well this runs on an ancient system. Will this for instance run on an old PI and make it snappy?

An X11 server is rarely on the hot path for rendering. Most X11 apps does most of the rendering to shared memory buffers themselves.

Re: Frame – Linux X server in Assembly

#105

Post by vidarh »

vinceguidry wrote: post #10
yjftsjthsd-h wrote: post #6

Earlier quoted context omitted.

At a glance, that looks like an X client library, not an X server?

Ah! You're right, I had read somewhere that he'd implemented his own X windows but I suppose I was mistaken.

I had Claude prototype one, but it's not complete or public yet...

Re: Frame – Linux X server in Assembly

#106

Post by VortexLain »

wild_egg wrote: post #47
idiotsecant wrote: post #45

Earlier quoted context omitted.

There is evidence that LLMs are capable of making assembly that runs a great deal more efficiently than the compiler can manage on its own.

There are a ton of optimization opportunities that hinge on the intent of a piece of code which static compilers can never detect at scale. LLMs can actually navigate that and write surprisingly optimal assembly. I've had all my side projects being written in x64 for the last 6 months and it is shockingly effective.

They can navigate that, if specifically targeted towards that, with proper adversarial reviews (including human reviews).

The question is, if one takes their time and has enough competence to do that.

Re: Frame – Linux X server in Assembly

#107

Post by VortexLain »

orphereus wrote: post #78

At first I was so interested into this guy, writing his own everything, thinking this has got to be something to aspire to, to be as good a programmer as this guy. Then I realised LLMs wrote this, and I was so very disappointed. It was naive and stupid of me to think otherwise, but here I am.

This person's idea is that if LLMs can create personalized and efficient software on the fly, it's good enough to use such software instead of more bloated, existing software which has to accommodate for usecases outside of his.

Sortof like suckless.org, but vibecoded on the fly.

I've found his Rust android development framework quite interesting, and will probably try using that to vibecode some software I've always wanted to have for my own personal use, but I've never had the time to code properly, especially considering it won't be useful for anyone else. Of course, I'll probably release it under a FOSS license, in case somebody finds it useful anyways.

Re: Frame – Linux X server in Assembly

#109

Post by stelonix »

cogman10 wrote: post #69
Aurornis wrote: post #48

Earlier quoted context omitted.

It’s an interesting strategy, but I question how much it pays off, if at all. Very few parts of a program benefit from manually tuned assembly compared to the naive C implementation. Writing everything in assembly adds an extra layer of thought, which even for an LLM is additional effort that could have been used for targeted optimizations instead. It makes it harder to notice patterns that have been trained into the…

It will make the code slower. Writing maintainable assembly is at odds with writing fast assembly in most circumstances. A key optimization that's hard to pull off is inlining. An optimizing compiler can see that a method is small enough that it can be pulled into the caller, it can then further eliminate from that smaller method branches that can't be executed due to the nature of the caller (Imagine calling a funct…

Simply no. This is a catchphrase, a buzzword from compiler companies. We have seen here on HN time and time again many articles showing the truth to be the opposite of what you've said. It's time to put to rest the nonsense that compilers optimize better than humans. Here a few examples I've talked about:

https://news.ycombinator.com/item?id=8508923 https://news.ycombinator.com/item?id=36618344 https://news.ycombinator.com/item?id=41922295 https://news.ycombinator.com/item?id=44186843 https://news.ycombinator.com/item?id=44177446 https://news.ycombinator.com/item?id=36949314

Re: Frame – Linux X server in Assembly

#110

Post by asveikau »

wojciii wrote: post #102
asveikau wrote: post #63

I can have a machine generated X server in assembly too. It's called cc -s. Or: $ objdump --disassemble /usr/lib/xorg/Xorg Maybe I'm getting too old for this, but I really don't see the point in having AI generate assembly code for this.

It's just you. Everything that can be done with LLMs is better than without. Why use tooling at all when you just can use a LLM? Maybe just make the LLM generate machine code for your tager directly. Just like making digital systems that are "better" than analog ones from the past because of digital.

That hasn't been my experience at all. The quality of these things is absolutely terrible. Maybe if you have no taste and no actual skill, you can't tell the difference.

The fact that you have chosen to say this about generated amd64 assembly is telling. This is a terribly pointless exercise. Further, the llms are bad at more niche languages especially. Even experimenting with having them write C, which is quite a bit more common than writing straight assembly, I have seen them fall over quite severely.

Post reply on HN