Parallel Level-Limited BVH Reinsertions for Faster Ray Tracing

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for optimizing bounding volume hierarchies (BVHs) in ray tracing, such as Meister's method, incur significant processing resources due to multiple iterations of searching, scoring, and executing candidate reinsertions, without adequately balancing the complexity of the optimization search procedure with the potential computational savings.

Innovation Solution

Restrict the search for candidate reinsertions to nodes within a certain distance from the root node and allow both binary and non-binary reinsertions, while employing conflict retries and deterministic node allocation to reduce processing burden and enhance computational efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If multiple iterations of searching, scoring, and executing candidate reinsertions are performed to optimize BVH, then the quality of the bounding volume hierarchy is improved, but the processing resources and computational cost increase significantly

Engineering Contradiction:
ImproveBVH optimization qualityVSAvoidoptimization search procedure complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent segments the BVH optimization process by restricting candidate reinsertion searches to only those nodes within a certain distance from the root node. This segmentation divides the search space into manageable portions, allowing the algorithm to focus computational effort on the most impactful regions of the tree while ignoring less critical areas, thereby reducing overall processing complexity.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent extracts and removes less important optimization candidates from consideration by imposing distance constraints. By taking out nodes that are beyond the specified distance threshold from the root, the algorithm eliminates a large portion of the search space that would contribute minimally to overall BVH quality, thus reducing computational burden without significantly compromising optimization results.

Inventive Principle:
Principle #2Taking out (Extraction)

2Manufacturing precision

If the search scope for candidate reinsertions is expanded to include all nodes in the BVH, then more optimization opportunities are found, but the processing time and computational resources increase

Engineering Contradiction:
ImproveBVH optimization qualityVSAvoidprocessing time
Core Design Contradiction:
Manufacturing precisionVSLoss of time

Solution Approach 1:

The patent applies partial action by performing optimization searches only on a subset of nodes (those within distance d from the root) rather than exhaustively searching all nodes in the BVH. This partial search approach captures the most significant optimization opportunities in the upper levels of the tree while accepting that some optimizations in deeper levels may be missed, achieving a practical balance between quality improvement and processing time.

Inventive Principle:
Principle #16Partial or excessive action

3Device complexity

If only binary reinsertions are allowed in the optimization process, then the search procedure is simpler, but the potential computational savings are reduced

Engineering Contradiction:
Improvesearch procedure simplicityVSAvoidcomputational savings
Core Design Contradiction:
Device complexityVSProductivity

Solution Approach 1:

The patent changes the structural parameter of reinsertion operations by allowing both binary and non-binary reinsertions. Instead of restricting modifications to simple binary splits, the algorithm permits nodes to be reinserted in ways that create non-binary tree structures, thereby expanding the solution space and enabling greater computational savings through more flexible hierarchy reorganization.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4254345B1Parallel optimisation of bounding volume hierarchy according to tree levels
Publication Date: 2025.07.09 IMAGINATION TECH LTD
  • EP4254345B1 patent drawingFigure 1~2
  • EP4254345B1 patent drawingFigure 3a~4b
  • EP4254345B1 patent drawingFigure 5

AI summary

A method comprising, in each of one or more iterations: i) for each of a plurality of input nodes in a current bounding volume hierarchy (BVH), searching for a candidate reinsertion which would move the respective input node from an old parent to a new parent, and which would reduce the expected computational cost of searching the BVH for a ray intersection; and ii) updating to the current BVH with one or more selected reinsertions selected from among the candidates. In the search for candidate reinsertions, either or both of: a) the new parent is limited to being related to the old parent by an ancestor at no more than a predetermined number of hierarchal levels above the old parent, and/or b) the input nodes are limited to being at or above a hierarchical level a predetermined number of hierarchical levels below the root node.