跳转至主要内容
User story

“ Taichi, a perfect programming framework for computer graphics courses.”

In Fall 2021, the Computer Graphics Laboratory (CGL) at ETH Zurich offered a course on physically based simulation, which requires students to create a small game or a demo scene using techniques they learned in class.

The vast majority of the student groups chose the Taichi programming language for real-time physical simulation.

GOAL

Select an efficient parallel programming language and use it to implement a physically based simulation project.

Challenges

  • The bar to parallel programming is set high: Few students have previous experience with CUDA.
  • The language must provide rich utilities, such as 3D visualization, to facilitate development.
  • Run the same parallel programming projects from different computing platforms.

为什么要 Taichi

  • Seamlessly embedded in Python: Easier to learn, more intuitive, and as fast as CUDA.
  • Taichi's GGUI system makes 3D rendering a breeze.
  • A single-source framework: Automatically falls back to CPUs if your system doesn't have CUDA or a powerful GPU.
  • An active community and a wide range of reference code make it easy to get started.

Results

  • 90% of the projects chose Taichi, and all were completed with flying colors.
  • Achieved comparable performance to C++/CUDA with much less code.

Requests for improvement: The future

  • Enriched GGUI tooling.
  • More descriptive error reports.
  • Move comprehensive benchmark reports.
  • Further reduced compile time.