Adaptive Mesh Approach for Predicting Algorithm Behavior with Application to Visibility Culling in Computer Graphics

TL;DR

This paper introduces an adaptive mesh method that uses random sampling to efficiently predict algorithm performance across input spaces, demonstrated on visibility culling in computer graphics.

Contribution

It presents a novel adaptive sampling technique for modeling algorithm behavior, with formal correctness proofs and practical application in graphics performance prediction.

Findings

Accurately predicts performance benefits of occlusion culling

Efficiently models algorithm behavior with fewer samples

Proves formal correctness of the adaptive approach

Abstract

We propose a concise approximate description, and a method for efficiently obtaining this description, via adaptive random sampling of the performance (running time, memory consumption, or any other profileable numerical quantity) of a given algorithm on some low-dimensional rectangular grid of inputs. The formal correctness is proven under reasonable assumptions on the algorithm under consideration; and the approach's practical benefit is demonstrated by predicting for which observer positions and viewing directions an occlusion culling algorithm yields a net performance benefit or loss compared to a simple brute force renderer.