Meshlet Generation Using Seed Points to Cut Rendering Overhead
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current meshlet generating strategies produce computationally inefficient meshlets with a higher number of unique vertices than polygons, leading to increased computational expense during rendering.
Innovation Solution
Implement a data processing apparatus that determines seed points independently for each meshlet generation, allowing meshlets to be generated without dependence on other meshlets, thereby reducing elongation and low-polygon issues.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If advancing front iteration strategy is used to generate meshlets, then mesh subdivision can be performed systematically, but the generated meshlets have higher number of unique vertices than polygons leading to increased computational expense
Solution Approach 1:
The patent changes the parameter of meshlet generation from sequential advancing front iteration to simultaneous generation using multiple seed points. This parameter change in the generation strategy allows meshlets to be created independently and in parallel, reducing the vertex-to-polygon ratio and thereby decreasing computational expense during rendering.
Solution Approach 2:
The patent segments the mesh generation process into multiple independent operations by using multiple seed points to generate different meshlets simultaneously. This segmentation allows each meshlet to be optimized independently, avoiding the cumulative vertex accumulation problem of sequential generation while maintaining systematic subdivision.
2Productivity
If meshlets are generated with maximum number of polygons to balance culling efficiency, then culling effectiveness is improved, but the number of unique vertices increases leading to reduced rendering efficiency
Solution Approach 1:
The patent applies preliminary action by selecting multiple seed points before meshlet generation begins. These pre-selected seed points are positioned to ensure that subsequent meshlet growth will maximize polygon inclusion while minimizing unique vertex creation. This preliminary positioning resolves the contradiction by optimizing both culling efficiency and rendering efficiency before the actual generation process.
Solution Approach 2:
The patent changes the generation parameter from single-source sequential growth to multi-source simultaneous growth. This parameter change allows meshlets to reach maximum polygon counts more efficiently while sharing vertices across multiple meshlets, thereby improving both culling effectiveness and rendering efficiency simultaneously.
3Stability of the object's composition
If sequential meshlet generation is performed, then meshlets can be generated in an organized manner, but each subsequent meshlet depends on previous meshlets increasing complexity
Solution Approach 1:
The patent segments the meshlet generation into multiple independent parallel processes, each starting from its own seed point. This segmentation eliminates the sequential dependency chain while maintaining organized generation through structured seed point selection and independent meshlet growth, thereby reducing process complexity.
Solution Approach 2:
The patent performs preliminary action by pre-selecting and positioning multiple seed points before generation begins. This preliminary organization ensures that subsequent independent meshlet generation processes are systematically distributed across the mesh, maintaining organizational stability without requiring sequential dependencies during the actual generation phase.
Data Source
AI summary
A data processing apparatus includes: an input unit configured to receive data corresponding to at least part of a mesh, where the at least part of the mesh includes a plurality of vertices, where each vertex corresponds to a location within a virtual space, and a plurality of polygons, where each polygon includes a perimeter comprising three or more lines, where each of the three or more lines intersects two of the plurality of vertices; a generating unit configured to, in a first phase, generate, based on the received data, two or more seed points, where each seed point corresponds to a location within the virtual space, and in a second phase, where the second phase is different from the first phase, generate two or more meshlets, where each meshlet includes a subset of the at least part of the mesh, where each meshlet is generated in dependence upon the location of a respective one of the generated seed points; and an output unit configured to output data corresponding to one or more of the generated meshlets.