When rust ≠ performance. a lesson in developer experience | Hacker News Reader

When rust ≠ performance. a lesson in developer experience

suriya.cc

37 pointssuriya-ganesh24 comments

> but because we have this data (file name, etc.) passed back and forth between different layers in the code. we were doing lots of .clone() and db.open() at different layers to fetch the same data. The interesting part for me was that, this change reduced our

Reduced your what? The article seems to be cut off.

I'm assuming it was going to say lines of code. It is highlighting that more was deleted than added.

I agree. Though the line count is meaningless as the vibe coded diff has pointless formatting and whitespace changes.

I wrote that PR and can guarantee you that it was not vibe coded.

the reason why I thought this PR was interesting is that, it only needed ~ 4 lines of real change, everything else was cascades due to that change. and the performance did improve with that "pointless formatting and whitespace changes"

ah! sorry. I meant to say, "the interesting part for me was that we removed code to improve performance."

Of course rust is not performance.

Rust is a programming language. Performance is a mix of programmer's ability, clever design and compiler optimizations.

For a long time, Rust == performance because a _lot_ of people moved into Rust that had never really done programming with a concurrency-sane compiled programming language. For them, Rust was synonymous with performance!

Many supporting crates reinforced this. Rayon, Tokio (though not "performant", it was an improvement over naive impl), etc made jumping from hello world to parallel/concurrent execution pretty simple.

But I'm glad to see the downturn on that hype cycle.

What is not performant about tokio? Do you know a better async runtime? I also heard tokio's "multi-thread" scheduler had some performance issues.

Nothing is wrong with tokio, actually. I chose the wrong word

async != performance

async == concurrency

Unless you face a specific kind of slowdown, concurrency, and therefore async (and therefore tokio) is not a panacea. Many do face that slowdown, so for many it is.

This is extremely sparse on details.

It feels like a rage bait. It's clearly their own fault and has nothing to do with Rust.

maybe Rustaceans shouldn't rage so easily?

Rust is a general purpose systems programming language. There are very few situations where you cannot attain C like performance with Rust. Skill issue, most likely.

algorithm design and managing memory is something you have to think about regardless of what language you use, that should be obvious. Using rust doesn't guarantee correctness or performance, that should also be obvious.

Rust has features that make it easier to make correct and performant software, I think most programmers would agree?

Please stop this rust clickbait nonsense.

> Please stop this rust clickbait nonsense.

But then new content on HN would decline to nearly nothing.

> Using rust doesn't guarantee correctness or performance, that should also be obvious.

Judging from the most vocal segment of their community it's very much not obvious to them.

> make it easier

Define "easier." It's harder to make mistakes. I'm not sure this equates to developer "ease" in any way. In fact, if it's meant to be effective, it should be quite the opposite.

> rust clickbait nonsense.

It cuts both ways. There is a huge volume of "I switched to Rust and got 1% additional performance over C" posts here.

I wonder to what extent the complexity of the rust language hurts performance. We all only have so much mental capacity, if much of it is spent on the various different intersections of rust's features, that reduces how much we can spend on making things fast.

(I like Rust btw)

The alternatives with the potential to be as fast (C, C++, D, zig) are more complex in this regard because they make memory safety and lifetime tracking something that you have to keep track of in your head. Rust's biggest win is removing that mental overhead while allowing you to achieve the same performance as those other languages.

Rust also hides allocation, and both the standard library and community best practices encourage many smaller allocations, which makes it much harder to reason about performance characteristics of your code.

So what you say is definitely true if you do an allocation heavy, heap fragmenting, RAII style of programming. Which is the context Rust was born in, right? A kind of C++ app dev context where that was (is?) the prevailing meta.

You're also completely glossing over the incredible complexity you get in all thee weird intersection of rust features. And there are a LOT of features. Reasoning about those are not free from a mental overhead stand point.

> Rust also hides allocation.

Does it? In Rust you allocate by calling a function. That's exactly the same as C. Are there any langauges that don't allow you to hide allocation behind a function call?

> So what you say is definitely true if you do an allocation heavy, heap fragmenting, RAII style of programming.

I don't think Rust encourages lots of small allocations. Most of the Rust code I work with does a lot of arena allocation (using crates like https://github.com/orlp/slotmap) and reusing of allocations. And for that matter a lot of stack allocation avoiding the heap entirely. And it's borrowing system is fantastic for working with shared pointers to data without having to worry that something might accidentally overwrite it.

> You're also completely glossing over the incredible complexity you get in all thee weird intersection of rust features.

I don't really buy that Rust is complex. More complex than C I suppose (but C just pushes all the complexity into making you write 5x more application code), but closer to something like Java than something like C++.

It's significantly more complex than Java due to the borrow checker and lifetimes in generics.

I don't think Rust is all that complex. On the other hand Java has many obscure features like serialVersionUID. For that matter, even fucking JavaScript is more complex than Rust with its var vs let, == vs ===, how, e.g., an empty array is false, parsing Date, how to make a "deep copy," etc, and TypeScript's type system can run Doom.

The borrow checker definitely adds complexity over Java. But Rust also removes complexity that is in Java. For example, inheritance and exceptions. Overall I do think Rust is a bit more complex than Java, but I don't think the difference is all that great.

Other languages remove lifetime tracking by making you track it in very limited spots. Instead of "for each individual object", you track it for "this particular kind of object in this part of my application".

I.e. don't need to keep track of the memory for each allocation in my HTTP request and make sure I clean it up before closing the connection, I can just allocate some memory _per request_, put stuff in it, and at the end it gets cleaned up, whether I used the memory or not.

Some languages have the idea of "memory allocators" as a native construct, so that you can actually start thinking about managing memory in more sensible terms than "everything individually", e.g. Odin lang.

Reply on news.ycombinator.com