use mrt to improve picking perfomace

[ouzhou]

Logically speaking, to implement picking functionality, it typically requires rendering twice. However, by utilizing WebGL2's MRT functionality, two results can be output in a single render, with one directed to the framebuffer for pixel color computation. This approach could potentially improve performance.

[jacomyal]

Hi @ouzhou,

I completely agree. I found out about this recently, while trying to implement force-directed layout algorithms using WebGL. Unfortunately, it was after having implemented the picking in sigma.

I hope to rework the picking that way for the v4 hopefully in 2025, in the same time we tackle the z-indexing issues, probably using OIT.

[ouzhou]

I am also implementing various GPGPU layout algorithms, and I find some open-source code that you can refer to.

https://github.com/cosmograph-org
https://github.com/antvis/g-webgl-compute

[ouzhou]

Moreover, I believe reducing boilerplate code is important; you can consider using foundational libraries like regl.

[ouzhou]

You can definitely use depth to solve z-indexing issues