Software Rasterizer

A minimal software rasterizer written in C. Goal wasn't to build a full rasterizer, it was to understand what actually happens inside one.

Kept it to a single triangle. Simple enough to read top to bottom and understand every step.

Output

A PNG image of an interpolated triangle rendered entirely on the CPU, no GPU involved.

How it works

1. Framebuffer

Just a flat array of bytes in memory width * height * 3 bytes for RGB. Everything gets written here, then saved as a PNG at the end using stb_image_write.

2. Edge function

float edgeFunction(Vec2 a, Vec2 b, Vec2 c) {
    return (c.x - a.x) * (b.y - a.y) - (c.y - a.y) * (b.x - a.x);
}

This is just a 2D cross product. Given two edge points a, b and a point c, it tells you which side of the edge c is on. positive, negative, or zero (on the edge).

It also doubles as area: edgeFunction(v0, v1, v2) gives you twice the area of the triangle. This is the same operation used for barycentric coordinates.

3. Bounding box

Instead of looping over every pixel on screen, compute the min/max of the triangle's vertices and only iterate within that box.

4. Per-pixel test

For each pixel in the bounding box, sample the center (x + 0.5, y + 0.5) and run the edge function against all three edges:

float w0 = edgeFunction(v1.position, v2.position, p);
float w1 = edgeFunction(v2.position, v0.position, p);
float w2 = edgeFunction(v0.position, v1.position, p);

If all three have the same sign, the point is inside the triangle.

5. Barycentric coordinates

Divide each weight by the total triangle area:

w0 /= area;  w1 /= area;  w2 /= area;

Now w0 + w1 + w2 = 1. These are the barycentric coordinates - how close the pixel is to each vertex, and you use them to interpolate any per-vertex attribute (color here, but same idea for UVs, normals, depth).

float r = w0*v0.color.r + w1*v1.color.r + w2*v2.color.r;

Build

gcc rasterizer.c -o rasterizer -lm
./rasterizer

Outputs output.png.

Dependencies

• stb_image_write