Taichi Programming Language / Quadrants

A high-performance parallel programming language embedded in Python; the engine underneath the Genesis simulator.

Taichi is a high-performance, parallel programming language embedded in Python, proposed by Yuanming Hu during his PhD at MIT (SIGGRAPH Asia 2019). Users write computational kernels using Python syntax, and Taichi just-in-time compiles them to run in parallel on CUDA, Vulkan, Metal, or CPU — a particularly good fit for physics simulations like the material point method, fluids, and soft bodies. The Genesis simulator's physics computation is built on Taichi; the Genesis team reportedly later maintained their own fork of Taichi, named Quadrants, so they could keep improving and adapting it for their needs. Students in embodied AI mostly encounter it when reading Genesis's source code or writing custom simulations.

ExampleDecorate a Python function with @ti.kernel to write a million-particle material-point-method simulation that runs directly on the GPU.

Also called

Taichi, Quadrants (Genesis fork)

Related

Genesis · Material Point Method · Differentiable Simulation · NVIDIA Warp · GPU-Accelerated Parallel Simulation

Sources

taichi-dev/taichi (GitHub)
Genesis-Embodied-AI/Genesis (GitHub)

As of

2026-09