Use a Vogel spiral in the sampling method of PCSS for directional light

[querielo]

Vogel spiral distribution reduces visible noise and achieves more uniform sample coverage.

Main	PR
https://playcanvas.vercel.app/#/graphics/shadow-soft	https://engine-h6nesoji0-playcanvas.vercel.app/#/graphics/shadow-soft

I confirm I have read the contributing guidelines and signed the Contributor License Agreement.

[mvaligursky]

The noise improvement is great. I'm just hoping something can be done with this banding on the long pole.

I wonder if similar factor applied to radius would help:
pow(data.currentRadius, 0.75)

[querielo]

@mvaligursky Now I know why we use pow here.

Falloff: 10

Main

vec2 offset = vec2(cos(theta), sin(theta)) * pow(r, 0.75);

vec2 offset = vec2(cos(theta), sin(theta)) * r;

vec2 offset = vec2(cos(theta), sin(theta)) * pow(r, 1.333); // 1.0 / 0.75

What do you think about it?

[mvaligursky]

It seems like this could work? Update the PR so I can run it locally to test.

[querielo]

@mvaligursky Done: vec2 offset = vec2(cos(theta), sin(theta)) * pow(r, 1.333);

[mvaligursky]

Perfect @querielo, this behaves equivalently now with visibly less noise, especially when the number of samples is low.
Please remove the poisson sampling in favour of vogel, and we're done!

[querielo]

@mvaligursky Done

[mvaligursky]

fantastic change, thanks!