Arbitrary Computation on the GPU Using WebGL
summary
Claim: WebGL can run arbitrary parallel computation by encoding data as pixels, shipping it through the WebGL pipeline, and reading back with canvas getImageData. Demos show 300,000 particles near 60 fps, matrix multiplication moves from 170,000 ms on CPU to GPU in many cases, but costs include expensive data transfer, OS/driver shader timeouts and no bitwise ops in the WebGL spec; a turbojet-style library packages the boilerplate and some examples do and don’t work.
AI-written summary · about
chapters (12)
- 0:00Intro
- 1:00WebGL
- 2:46Gradients
- 3:56Edge Finder
- 4:36Edge Threshold
- 5:22Particles
- 6:21Why WebGL
- 9:18Matrix Multiplication
- 11:54Demo
- 12:57CSS Animation
- 16:28Big O Notation
- 20:15No Bid Operations
related talks
25mArbitrary Computation on the GPU Using WebGL
23mArbitrary Computation on the GPU Using WebGL
22mWebGL For More Than Graphics: Is It Possible?
39mThe Pixel Factory
38mWebGPU: Next-generation 3D Graphics on the Web
32mGPU.JS - GPU Accelerated JavaScript
25mHow to Program Your GPU (and Not Do It Alone)
34mNext-Generation 3D Graphics on the Web