How to pass an array of size changing structures into a glsl shader

Question

In order to add dynamic lighting to my application, I was thinking that I could represent each shadow casting object as an array of floats (every 3 floats is a vector and so every 9 floats is a triangle). And so, in order to cast ALL possible shadows, I need an array of these arrays (if this was C++ I would have done an array of vectors).

Is there a way to construct somehting similar in a glsl shader, throught either SSBOs or uniforms?

And if not, how could I pass this information. The problem here is that I don't know how big the number of triangles per shading object is, so I have no way of determining a constant sized structure to make my array of solid objects.

Shadowmaps is todays most widely used technique (at least in games, there is active research for incorporating soft-shadows, etc.). Shadow Volumes are another technique. Should be fairly easy to find a OpenGL tutorial if you google for them. — BDL
– BDL, Commented Jan 14, 2018 at 22:32
Textures won't work, they require both dymensions to be the same, which is exactly my problem — Makogan
– Makogan, Commented Jan 14, 2018 at 22:53
Since when do Texture require both dimensions to be the same? You can create texture of any size you want. — BDL
– BDL, Commented Jan 14, 2018 at 23:00
If you find a method for rendering shadows without any artifacts which is in addition faster than traditional techniques (which I doubt, but who knows), let me know, I'd be highly interested how you do it. My current guess is that you're going to use some kind of ray-tracing technique, but I'm fairly sure that this won't work out. I really don't wont to discourage you from trying, just want to make you aware that you might be wasting a lot of time. — BDL
– BDL, Commented Jan 14, 2018 at 23:17

Nicol Bolas · Accepted Answer · 2018-01-15 01:30:02Z

You're thinking simultaneously too high-level and too low-level. You say that you need a bunch of arrays, but you don't. What you need is a way to get the vertex data for a particular object in the scene. This doesn't have to be encoded as "an array of these arrays" at all.

Instead, encode it as a single massive array of vertex data. Each object has indices that specifies the location in this array for the vertex data.

layout(std430) buffer obj_data
{
  uvec2 objects[];
};

layout(std430) buffer vertex_data
{
  vec4 vertices[]; //NEVER use `vec3`s in storage blocks
};

objects is an array, where each uvec2 represents a specific object. The x component of that uvec2 is the offset into vertices where its vertex data starts. And y is the number of vertices to read, starting at x.

So vertices[objects[10].x] is the first vertex of the data for the object with index 10.

And BTW:

if this was C++ I would have done an array of vectors

If this were C++, I'd encode it more or less like I would for GLSL: store all of the vertex data in a single array, and have each object reference its contiguous slice of that array. It's more efficient to access than an array of vector, and requires far fewer allocations.

Collectives™ on Stack Overflow

How to pass an array of size changing structures into a glsl shader

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related