GPU MODE - (Mini) Competition #3. AlphaFold’s Triangle Multiplicative Update

[asdfefbhjy]

competion time(2025/6/24 - 9/30), trimul leaderboard

New competition problem: Triangle Multiplicative Update (June 24, 2025)

Announcing a new GPU_MODE kernel writing competition: our first featuring both NVIDIA and AMD hardware! The first problem will be the Triangle Multiplication operator essential to the AlphaFold 🧬 models! It's a particularly tricky problem with no good public implementation!

Edit. Funnily enough, while writing up this problem, we noticed that NVIDIA actually released that they had been working on this kernel in their cuEquivariance library — what a coincidence! We ask that participants do not use this library in their solutions, as it is closed source (we will be automatically and manually removing these solutions on the leaderboard). Honestly though, I’m confident that you all can easily beat their solution!

AlphaFold core update module open optimization: Triangle Multiplicative Update competition launched

GPU MODE - (Mini) Competition #3. AlphaFold’s Triangle Multiplicative Update

https://tinyurl.com/gpumode-trimul

https://github.com/gpu-mode/lectures

http://www.youtube.com/@GPUMODE

https://github.com/NVIDIA/cuEquivariance

[asdfefbhjy]

2025/8/1: Improve "Triangle Multiplicative Update" in Protenix:

bytedance#164

bytedance#165

[asdfefbhjy]

2025/8/19: Protenix seems to have used Nvidia's cuEquivariance to handle "Triangle Multiplicative Update"