Update multisplat16.shader

[QbieShay]

My own attempt at more organic blending for the multisplat plugin and preventing hard cuts when painting with low opacity

[Zylann]

This needs explanation because I have no clue how this works^^"

Also you updated that shader but it's not the only one using that technique (there is a fair amount of copypasta to make once this is finalized)

[Zylann]

Should respect naming convention, uniforms start with u_ (I wish I could use _ only but Godot reserves that).
Also I have no idea what these do just from the name. I see they all affect how textures are blended (when depth blending is on) but their name alone doesnt reflect that either.

[Zylann]

Careful with pow, it's a transcendental math function, it's expensive. Here you raise to power 0.5, which is actually square root, so maybe use sqrt instead?
Not sure what this formula does

[Zylann]

pow again, prefer squaring by multiplying by itself. Are GLSL compilers smart enough to replace the call with that? Still no idea why you do this though, it's like some textures are more important than others or something

[Calinou]

Could you add comparison screenshots between the old and new shader?

[QbieShay]

The pow in the function is to alleviate issues when there are more than 4 textures blending at the same time. If you paint with low opacity brushes with more than 4 textures you'll see abrupt cuts starting to show.

I can't show the assets i tested with but I can show in a clean project later.

As for the names, i have no idea either exactly what they do :D that's why they are not very descriptive. I'm open to any name you will propose ^^

[TokisanGames]

When I have 5+ textures in one area, I often get lines like the dashed line across textures as you can see on the rock faces on the right and left (not the rock cracks). By pushing around the textures more I can eventually brush these out.

I inserted the changes to bump into my get_depth_blended_weights() function, however none of them, nor any other parameter in that function affected those line artifacts. They did change other blended edges, but I have no issue with the default blend.

Regarding the functions:

• 1-sqrt(1-x) is indeed equivalent to 1-pow(1-x,0.5) shown on graphtoy below.
• x*x is indeed preferable to pow() for reasons discussed here: https://community.khronos.org/t/pow-x-2-different-then-x-x/70839/3
• x*x*x and x*x*x*x are probably also better than pow for the above reasons.
• Graphtoy shows the shape of all curves used. 1-sqrt(1-x) is basically a slow linear line with an ease in ending.
  https://graphtoy.com/?f1(x,t)=1-pow(1-x,0.5)&v1=true&f2(x,t)=1-sqrt(1-x)&v2=true&f3(x,t)=x*x&v3=true&f4(x,t)=x*x*x&v4=true&f5(x,t)=x*x*x*x&v5=true&f6(x,t)=0&v6=false&grid=1&coords=0,0,2.3741360268016223