What you can do with WebGPU? By Corentin Wallez, François Beaufort
Vložit
- čas přidán 5. 10. 2023
- WebGPU is the successor to WebGL that brings the advancements of modern graphics APIs to the web. It is designed to be fast, portable, and easy to use.
In this talk, we will introduce WebGPU and show how it can be used to create high-performance graphics applications and perform data-parallel computation on the web and native platforms. We will also discuss the future of WebGPU.
This talk is aimed at developers who are interested in learning about WebGPU and how to use it. No prior experience with WebGPU is required.
FRANÇOIS BEAUFORT
François Beaufort, Developer Program Engineer at Google
developer.chrome.com/authors/...
François loves to dive into open source code to understand how things work in web browsers.
CORENTIN WALLEZ
Corentin is the co-chair of the WebGPU standardization group since it was created and also leads the implementation of WebGPU in Chromium. Before that he grew up in the hills of Pechabou and worked on other GPU technologies for the Web. - Věda a technologie
such a good presentation.
Love the interaction between Corentin and Francois 😂
I gotta comment again. Love this presentation
It was fun and informative, thank you
Great talk!
Well, good to know webgpu will be available for Linux soon. I want to try it out
WebGPU is great idea in common. But I still cannot find constructive answer why WGSL was invented. One more language to support.
They did that to support WebGPU in not only web but also in native apps as well.
@@abhishekbu1087 Why it's not possible with GLSL? It's rhetoric question. GLSL supported by all platforms, operating systems, GPU manufacturers and game engines. Even both OpenGL and Vulkan. Can you give real CONSTRUCTIVE answer? What's advantage in WGSL? Outweight of new syntax, support of compiler to SPIR-V and millions of already written shaders by thausands people to copy-paste?
I see huge strategic mistake with the price of promotion of this technology. I mean WebGPU, not WGSL.
@@alexfrozen Maybe they wanted to make the API easier to work with. Also it was designed from ground up, maybe it was harder to make it fully compatible with other apis. Also this is for the web, so the scope is very different.
@@abhishekbu1087 It wasn't designed from ground up for sure. It's made around already created gem. A kind of wrapper. About compatibility, there is already solved solution SPIR-V family. What about API... It's good. Just good. But please note, I am talking about creation of programming language, not API. They're as same as table and crocodile. Ofc there is a reason, like everithing has cause. My personal deep technical research gave me an answer. Same reason as for Rust, which "better" than Go. A coin to create personal name. What about "make things better to work", nope. I said it above. It's one more language. You should learn it. And you should rewrite existing shaders from GLSL/HLSL. Or be very familiar with SPIR-V tools to transpile it. And syntax... It's more unintuitive.
1. Because if I remember correctly, Apple said no to SPIR-V.
2. Because GLSL has a billion different versions ... and not just version numbers, but preprocessor hints about which features to use, inside of which versions... which they would then need to juggle to make work across OpenGL, OpenGL ES2/3, DX11, DX12, Vulkan, and Metal as back-ends; because it needs to run on Windows / Linux / Mac, and on the mobile forms of those OSes.
3. Because they wanted to make it statically analyzable (like SPIR-V), so that they could make #2 happen.
4. Because Apple said no to SPIR-V as a front-end for Metal.
My memory might be off about #1 or #4. ...those are conversations from 5+ years ago.
No GLSL == dead on arrival. I'm not supporting ANOTHER shading language.
Understandable but it's very easy to port.