Arbitrary Computation on the GPU Using WebGL

on YouTube ↗23m

" summary

WebGL is used to pass arbitrary arrays as WebGL textures as input to fragment shaders, compute things like matrix multiplication on the GPU, and serialize results back by reading pixels off the canvas readPixels bottleneck. Benchmarks show small matrices are slower on GPU but large ones hit roughly 200x speed; a 1024×1024 multiply runs in minutes on naive code but is far faster on the GPU. Missing bitwise ops in GLSL prevents hashes and Bitcoin mining.

AI-written summary · about

" chapters (8)

1. 0:00Start
2. 0:17FOLLOW ALONG: SETHSAMUEL.GITHUB.IO
3. 1:29WHAT IS WEBGL?
4. 7:07CPU V GPU
5. 8:47COMPUTE SHADERS: COMING SOON(TW)
6. 9:39THE HIGH COST OF SERIALIZATION
7. 10:01MATRIX MULIPLICATION
8. 18:46POTENTIAL APPLICATIONS

" related talks

• 25mArbitrary Computation on the GPU Using WebGLSeth Samuel · CascadiaFest 2016
• 24mArbitrary Computation on the GPU Using WebGLSeth Samuel · JS Kongress 2016
• 22mWebGL For More Than Graphics: Is It Possible?Seth Samuel · JSConf Colombia 2016
• 38mWebGPU: Next-generation 3D Graphics on the WebCorentin Wallez · DevFest Toulouse 2019
• 25mHow to Program Your GPU (and Not Do It Alone)Max Bittker · DinosaurJS 2018
• 34mNext-Generation 3D Graphics on the WebRicardo Cabello, Corentin Wallez · Google I/O 2019
• 05mA Brief Intro to GPU Shader Programming in WebGLSuby Raman · dotJS 2017
• 28mTurbo charging 2D on the webMat Groves · JSNation 2019