#3ds max 2011 clipping plane series#

Given the clipcodes, all this is just a few gates worth of hardware вЂ“ simple stuff.ДЄdditionally, the shaders can also generate a set of вЂњcull distancesвЂќ (a triangle will be discarded if any one cull distance for all vertices is less than zero), and a set of вЂњclip distancesвЂќ (which define additional clipping planes). Then, for each primitive, the bitwise AND of the clip-codes will tell you all the view-frustum planes that all vertices in the primitive are on the wrong side of (if thereвЂ™s any, that means the primitive is completely outside the view frustum and can be thrown away), and the bitwise OR of the clip-codes will tell you the planes that you need to clip the primitive against. You compute the clip out-code (or just clip-code) for each vertex (this can be done at vertex shading time and stored along with the positions, for example). This can be done very efficiently using Cohen-Sutherland-style out-codes. We first need to find out if the triangle is partially or even completely outside any of these clip planes. Clip space is chosen to make the equations that describe the view frustum as simple as possible in the case of D3D, they are, ,, and note that all the last equation really does is exclude the homogeneous point (0,0,0,0), which is something of a degenerate case. The short version is this: Your vertex shader returns vertex positions on homogeneous clip space. So IвЂ™m not gonna explain polygon clipping in general here, you can look that up in any computer graphics textbook, although most make a terrible mess of it if you want a good explanation, use Jim BlinnвЂ™s (chapter 13 of this book), although you probably want to pass on his alternative clip space these days, to avoid confusion if nothing else.ДЄnyway, clipping.

Yeah, I guess weвЂ™d better do that first, huh? This is one part of pipeline that really does exactly what youвЂ™d expect, pretty much the way you would expect it too (i.e. So can we rasterize them already? Not quite. In other words, we now have complete triangles, not just a bunch of vertices. Either way works fine.ДЄnd now we have expanded out all the vertices that make up a primitive. We can either do this by reading the original index buffer and keeping a copy of our vertex index->cache position map around (as I described), or we can store the indices for the fully expanded primitives along with the shaded vertex data, which might take a bit more space for the output buffer but means we donвЂ™t have to read the indices again here. In short, all that happens here is that we gather vertices. And so on for patches with larger numbers of control points.

If that primitive happens to be a point, this just reads exactly one vertex and passes it on. The next step is to assemble all the vertices belonging to a single primitive (hence вЂњprimitive assemblyвЂќ). This is important, because it means we can truly treat each block independently and never need to buffer more than one block of shader output вЂ“ we can, of course, but we donвЂ™t have to. When we left the vertex pipeline, we had just gotten a block of shaded vertices back from the shader units, with the implicit promise that this block contains an integral number of primitives вЂ“ i.e., we donвЂ™t allow triangles, lines or patches to be split across multiple blocks. So much so in fact that weвЂ™re not going to see any rasterization in this post вЂ“ thatвЂ™ll have to wait until next time. You see, thereвЂ™s a bunch still left to do before we actually start rasterizing primitives. WeвЂ™re done with vertex shading, so now we can start actually rendering stuff, right? Well, not quite.

#3ds max 2011 clipping plane series#

This post is part of the series вЂњA trip through the Graphics Pipeline 2011вЂќ.ДЄfter the last post about texture samplers, weвЂ™re now back in the 3D frontend.