I am learning about screen space shadow techniques, and I do not understand one part: why we need to ray march in the loop for that short distance, why we just cannot project the depth buffer to the shadow map from the point of view of the light. In my mind, it is N-raymarching steps faster without losing any quality. I understand the limitation of screen space shadow techniques: only visible geometry will cast a shadow, and that argument applies to the N-raymarching steps. So don’t bring me that argument. Maybe constructing the highmap mesh is expensive?