laser VS rust

It depends.

You need 2~3 accumulators to saturate instruction-level parallelism with a parallel sum reduction. But the compiler won't do it because it only creates those when the operation is associative, i.e. (a+b)+c = a+(b+c), which is true for integers but not for floats.

There is an escape hatch in -ffast-math.

I have extensive benches on this here: https://github.com/mratsim/laser/blob/master/benchmarks%2Ffp...

Ah,

It was from an older implementation that wasn't compatible with Nim v2. I've commented it out.

If you pull again it should work.

> Anyway the reason for your competitive performance is likely that you are benchmarking with very small matrices. OpenBLAS spends some time preprocessing the tiles which doesn't really pay off until they become really huge.

I don't get why you think it's impossible to reach BLAS speed. The matrix sizes are configured here: https://github.com/mratsim/laser/blob/master/benchmarks/gemm...

It defaults to 1920x1920 * 1920x1920. Note, if you activate the benchmarks versus PyTorch Glow, in the past it didn't support non-multiple of 16 or something, not sure today.

Packing is done here: https://github.com/mratsim/laser/blob/master/laser/primitive...

And it also support pre-packing which is useful to reimplement batch_matmul like what CuBLAS provides and is quite useful for convolution via matmul.

For convolutions: - https://github.com/numforge/laser/blob/e23b5d63/research/convolution_optimisation_resources.md

You can train yourself on matrix transposition first. It's straightforward to get 3x speedup between naive transposition and double loop tiling, see: https://github.com/numforge/laser/blob/d1e6ae6/benchmarks/transpose/transpose_bench.nim#L238

Rust

It seems to be more fussy about compiler optimizations, though: https://github.com/rust-lang/rust/issues/125543

apimock-rs is one of my projects on API mock Server generating HTTP/JSON responses to help to develop microservices and APIs, written in Rust.

Rust, the language itself depends on 220 packages: https://github.com/rust-lang/rust/blob/e8753914580fb42554a79...

If you trust nobody, it is hard to use anything.

But about your second note, (environment, mismatched dependencies), I would argue that Rust provides the best tooling to solve or identify issues on that area.

rustc source code

First look is into The Unstable Book. Well, it does not look informative but gives us some background from the rust-lang Github project-const-generics. It says:

Rust has been around for several years and works well as a system and general programming language. There are many fine introductions to the language, a good place to start is here: https://www.rust-lang.org/

‍For the rest of this post I’ll list off some more tactical examples of things that you can do towards this goal. Savvy readers will note that these are not novel ideas of my own, and in fact a lot of the things on this list are popular core features in modern languages such as Kotlin, Rust, and Clojure. Kotlin, in particular, has done an amazing job of emphasizing these best practices while still being an extremely practical and approachable language.

> There are online Rust compilers and interpreters already if you just want to rapid prototype and develop ideas in Rust

You are responding to one of the key developers of Rust early on[1], who's been working with the language for 14 years at that point.

[1] https://github.com/rust-lang/rust/graphs/contributors?from=2... and he's still #16 in commits overall today, despite almost no activity on the rust compiler since 2014.

If you haven't dipped your touch-typing fingers into Rust yet, you really owe it to yourself. Rust is a modern programming language with features that make it suitable not only for systems programming -- its original purpose, but just about any other environment, too; there are frameworks that let your build web services, web applications including user interfaces, software for embedded devices, machine learning solutions, and of course, command-line tools. Since a custom GitHub Action is essentially a command-line tool that interacts with the system through files and environment variables, Rust is perfectly suited for that as well.