2023 LLVM Dev Mtg - Mojo 🔥: A system programming language for heterogenous computing

Channel:LLVM · Length:49:48 · Published:2023-11-22 · Views:61057

Watch on YouTube

Summary

A 2023 LLVM Developers' Meeting keynote by Abdul Dakkak, Chris Lattner and Jeff Niu introducing Mojo, a Python-family systems language built on MLIR and LLVM for heterogeneous computing.

Why HN recommends it

HN points to it as the best deep dive into Mojo's design, including MLIR use and SIMD-first scalars.

this video totally changed my view — HN, 2024

Programming language design · C, C++, Rust, Go & systems languages · Compilers, interpreters & program analysis

Subtitles (excerpt)

thank you Alex wonderful we're very happy to be here and talk a little bit about what we've been up to so we'll start with what is mojo um at a glance the top level points of Mojo is that it's a pythonic systems programming language so what does that mean that means we're here to do really cool things with systems and compilers and it happens to look like python but forget everything you know about python please please so this thing is about one year old so it's still pretty early still in development it's still quite interesting in doing some cool stuff though um and we also have a Vibrant Community we have over 150,000 users we have a b big community in Discord and there's a bunch of excitement around this so we'll dive today into why did we do this in the first place that's often something we're asked we'll talk about how we approach designing a new language from scratch we'll talk about internal implementation details including some of the horrible things we did to lvm talk about what this means for accelerators and compute and then wrap things up so first why why why why why why right so many of you are working on AI and if you work on AI the question I will ask of you all is if AI is so important to

On Hacker News

When mojo was initially released I didn’t think much of it because: 1. It was announced prior to being available, even in closed source form. 2. Emoji filenames meant to me that it wasn’t taking itself too seriously. 3. Many people have tried to make Python fast and failed, and that was my naive take on it, and despite the fact that LLVM allstars were working on it, it seemed like a lofty goal. 4… — nickysielicki, 2024-03
I think they had a good dive here [VIDEO] The gist is that MLIR (unlike LLVM) was designed to be multithreaded so the whole stack above LLVM IR is parallel, and then LLVM is used as a per-function codegen tool in parallel, so although LLVM itself is single threaded there are multiple seperate contexts of it. also I think they wrote their own linker. — MohamedMabrouk, 2026-08
They've given a talk abut this on the LLVM dev meeting: [VIDEO] They have some fancy tech around compile-time interpreter, mojopkg, and their code generator. They also have a graph compiler (not mentioned in the talk) that can fuse kernels written in Mojo. — melodyogonna, 2025-09
I was watching a recent talk about the new Mojo language [1]. There is a section on SIMD and how they treat scalar values as a special case of vector operations (around the 33 min time). It does seem that tensors are one of the core abstractions for modern ML systems. I've heard people joke that AI is just matrix multiplication. Python is such a flexible language that creating abstractions around… — zoogeny, 2023-11

Related videos