Taichi.js

Play Taichi kernels online by Emscripten and Taichi's C backend.

TaichiHub

TaichiHub is a website where you can create and share your creativity by writting shaders using the Taichi programming langurage.

If you're familiar with Shadertoy, then Taichihub is to Taichi just as Shadertoy is to GLSL.

Preview it online at https://142857.red:3389/.

Warning

As the port abuse of :3389 suggested, the server is still at its early development stage. So there's no warranty, it might be buggy, and may temporary shutdown for maintainance.

Note

Your browser needs to support WebAssembly (WASM) to make this site functional.

How to browse and edit artworks?

Entering the website, you'll see several artworks created by users. Hovering on them to preview the animation. Click one of them and you'll be redirected to its 'EDIT' page.

The source code is on the left side, and the result is on the right side. You may edit the source code, then press the 'RUN' button to get the new result (may take a couple of time to compile).

Or, if you'd like to create a new artwork from scratch, press the 'CREATE NEW' button to get a blank one. It should gives you a pre-written template rendering a colorful animation.

Once your artwork is done, press the 'SAVE' button, give your artwork a 'title', and it will be displayed on the homepage.

Note

In fact till now there're only administrator's artworks in the database... so please feel free to create and share your own one if you'd like to :)

Difference between TaichiHub and ordinary Taichi

Directly copy-and-paste an ordinary Taichi program in https://github.com/taichi-dev/taichi/tree/master/examples won't work, cause they're using Taichi's builtin GUI event loop, which is deeply embed in Python. However TaichiHub is hosted on the web, its frontend is Javascript..

Background

As we all know, Taichi is a langurage embed in Python, while the browsers only support Javascript. But with the power of Taichi's C backend, we can export Taichi kernels to C source, and then compile them into Javascript and WebAssembly using Emscripten.

But we can only export the computations in Taichi-scope, not Python-scope this way. So computations must resides in Taichi kernels to write a functional TaichiHub artwork. The code in Python-scope are only executed on compile-time, only Taichi-scope code are called on run-time.

Kernels

So, in order to let the compiler know what kernels shall I export, we need to invoke the kernels to trigger JIT compilation. To do this, please invoke all the kernels required at runtime.

Or, equivalently, we provide a handy decorator to execute the kernel right after declaration:

import hub

@hub.kernel
def render():
    ...

# is equivalent to:

@ti.kernel
def render():
    ...

render()

You could replace all @ti.kernel to @hub.kernel while porting a program from ordinary Taichi to TaichiHub shaders.

Kernel names

TaichiHub will search kernels by name, for example:

• reset() will be called on initialization, or when 'RESET' button pressed.
• render(t: float) will be called on each frame, t is the current time in seconds.
• substep() will be called n-times per frame before render, hub.substep_nr(n) to specify n.
• onclick(x: float, y: float) will be called when LMB is clicked, x, y is the mouse position.

Rendering

So, we've pre-built the frontend GUI system and logic infrastructure in Javascript, which render the HTML5 canvas instead of Taichi GUI which is not available in browsers. Some changes are required to adapt to it:

1. To render images:

Ordinary:

img = ti.Vector.field(3, float, (512, 512))
...

while gui.running:
    ...
    gui.set_image(img)
    gui.show()

TaichiHub:

import hub

img = ti.Vector.field(3, float, (512, 512))
...

hub.bind_image(img)

2. To render particles:

Ordinary:

pos = ti.Vector.field(2, float, 8192)

while gui.running:
    ...
    gui.circles(pos.to_numpy())
    gui.show()

TaichiHub:

import hub

pos = ti.Vector.field(2, float, 8192)
...

hub.bind_particles(pos)

How to setup a TaichiHub server

1. Install Emscripten SDK according to their official documentation:

git clone https://github.com/emscripten-core/emsdk.git
cd emsdk
./emsdk install latest
./emsdk activate latest
source ./emsdk_env.sh
cd ..
emcc --version

2. Install requirements:

python3 -m pip install pymongo
python3 -m pip install flask
python3 -m pip install taichi
apt install mongodb

3. Clone this repo:

git clone https://github.com/taichi-dev/taichi.js.git
cd taichi.js

4. Build the container for executing user Python code:

docker build . -t taichihub

5. Start the server:

source /path/to/emsdk/emsdk_env.sh  # add `emcc` to PATH
cd taichihub
python3 -m flask run -h 0.0.0.0 -p 80

Acknowledgements

MDUI and jQuery are used in frontend page design.

Codemirror is used for the source editor.

BootCDN is used for hosting and accelerating their Javascript / CSS.

Flask is used as an infrastructure for a HTTP server.

MongoDB is used for storing database and artworks.

Emscripten is used for compiling C source emitted ny Taichi.

Docker is used for creating secure container executing user code.

Shadertoy is taken as a good reference for TaichiHub.