Shadow mapping in two passes. The scene is drawn from the light into a depth-tested render target, each fragment recording its distance to the light; the lit pass projects every point back into that map and asks whether anything closer was already there. Nine taps per fragment soften the edge. The inset is the map itself.