Ibraheem Ahmed

That terrible feeling when you have a great blog post idea but know you'll never have time to write it

Writing fast code under optimizing compilers feels similar to *prompting: 1) Developing an intuitive understanding of how the model combines your patterns to outputs 2) Knowing what a "good" output is, in regards to correction. * any deterministic ML inference with text input

Today, we’re announcing the preview release of ty, an extremely fast type checker and language server for Python, written in Rust. In early testing, it's 10x, 50x, even 100x faster than existing type checkers. (We've seen >600x speed-ups over Mypy in some real-world projects.)

I was recently made aware that every single additional *cycle* in the hash function used by rustc increases its runtime by a whopping 0.25%

@charliermarsh Not sure if there's anything written up other than that PR and the rustc perf runs.

@ibraheemdev Bro I'm spreading disinfo?? Where can I read about it?

The latest version of rustc-hash includes the newly designed hash function from orlp!

@charliermarsh The new rustc-hash algorithm is actually different than foldhash. rustc-hash is optimized specifically for the compiler's usage, while foldhash is more general purpose.

Here are the benchmarks, where `foldhash-f` represents the new hash function, and fxhash represents the previous hash function.

@chillbert32 @Lucretiel @mitsuhiko This is better than multiple points of failure in this case.

@Lucretiel @mitsuhiko Except for the fact that it's a single point of failure.

So I looked through some more reasons for why zerocopy now shows up in my dependency tree and from what I can tell it's all coming from the zerocopy developers themselves making PRs against other crates introducing the dependency. Not sure what to think of that.

@mitsuhiko Would it feel different if zerocopy was an official rust-lang repository backed by a working group? I think that's essentially the idea here.

@mitsuhiko Not exposed, you say 🙃 RandomState::new().build_hasher().finish()

Fun fact: rust needs to generate random numbers in the standard library for hash maps and sets. But that's not exposed. This is how many dependencies adding the latest version of the popular rand crate pulls in that everybody uses:

Very high quality code here too

No I don't want AI generated PRs on my repository thank you very much

@charliermarsh The only time Charlie ever takes a break

This is good actually

Unfortunately all of my short blog post ideas evolve into books

Going to try and publish more than one blog post this year...

@GustavSterbrant @olson_dan If this level of safety is not important to you then Rust might not be for you, but once you appreciate it the experience becomes a lot more enjoyable. And to be honest there's only a handful of patterns you have to come to understand before feeling relatively comfortable.

@GustavSterbrant @olson_dan That's fair, but Rust's goal is to force you to think more about what exactly you're doing. Another language would let you trivially create globally mutable state, which you can also do in Rust, it just makes it less accessible because that's probably not what you actually want.

@gwenzek Rust is really good at this too, I just happen to be working with a lot of raw pointers and atomics.

@ibraheemdev Switching from C++ to Zig removes 90% of them

Eliminating LLVM optimization remarks is my passion.