Rust and Khal in OpenHarmony for single-source cross-platform GPU/CPU compute

With the emergence of World Models for robotics and training datasets generation, high-performance multiphysics simulators are becoming central to robotics training with libraries like qNewton (Nvidia Warp, Python) and Genesis (Taichi, Python). In this presentation, we introduce Project Nexus: a work-in-progress alternative written in Rust (compiled with rust-gpu), running on all GPU platforms including the web.

• Motivation for World Models, dataset generation, and the role physics simulation plays in it
• Introduction to Nexus and its main dependencies (Khal, Vortx, glamx)
• Benefits of rust-gpu for a 100% Rust codebase
• Live demo