IT lexicon › Programming › Micropolygon

Micropolygon

Programming På svenska → Updated: 2026-10-26

A polygon diced down to roughly half a pixel, so that shading and displacement can happen per surface-covering fragment without visible edges.

The concept comes from Pixar's REYES architecture (Cook, Carpenter, Catmull, 1987). Instead of shading large triangles, surfaces are diced into a grid of micropolygons, each covering less than a pixel. Displacement, motion blur and shading can then be computed uniformly and finely, which gave film rendering its smooth surfaces.

The snag is efficiency: fixed-function rasterization hardware wastes nearly all its work when a triangle is smaller than a pixel, because it always processes at least a 2×2 quad. That is why Nanite in Unreal Engine 5 built its own software rasterizer specifically for micropolygons. The same insight drives research into efficient sub-pixel geometry on the GPU.

← Back to the lexicon