Ellipsoid Fragment Count Estimation for GPU Scheduling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In automotive driver information display systems, estimating the runtime duration for command sequences on a shared graphics processing unit (GPU) is challenging due to the need to accurately determine the number of fragments to be displayed without executing all transformation steps for all vertices, which is processor-intensive and leads to over or under estimation.
Innovation Solution
A method using an ellipsoid representative of a set of vertices to estimate the fragment count by applying transformations and calculating the projection area, allowing for improved predictive GPU scheduling and load balancing without executing all transformation steps for all vertices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all transformation steps for all vertices are executed to accurately determine fragment count, then measurement precision is improved, but use of energy and processing time increase significantly
Solution Approach 1:
The patent creates a simplified copy of the 3D object in the form of an ellipsoid that approximates the original object's bounding volume. This ellipsoid copy contains only a subset of the original vertices, allowing fragment count estimation without processing all original vertices. The ellipsoid serves as a computational proxy that preserves the essential spatial characteristics needed for accurate fragment count prediction while dramatically reducing processing requirements.
Solution Approach 2:
The patent extracts only the necessary geometric information from the full set of vertices by fitting an ellipsoid to the 3D object. This extraction process identifies and utilizes only the critical bounding characteristics needed for fragment count estimation, discarding the redundant detailed vertex information. The ellipsoid representation extracts the essential spatial envelope without requiring processing of all original vertex coordinates and transformation steps.
2Measurement precision
If all transformation steps for all vertices are executed to accurately determine fragment count, then measurement precision is improved, but productivity decreases due to processing time
Solution Approach 1:
The ellipsoid copy enables rapid fragment count estimation by replacing the computationally intensive processing of all original vertices with a simplified geometric model. This copy contains minimal vertex data that can be transformed and processed much faster, allowing the system to generate accurate runtime duration estimates without bottlenecking the GPU scheduling pipeline.
Solution Approach 2:
The patent performs preliminary action by pre-computing the ellipsoid representation and its transformation properties before actual rendering or fragment processing occurs. By establishing the bounding ellipsoid and its characteristics in advance, the system prepares the necessary estimation data structure that can be quickly queried for fragment count predictions, eliminating the need for exhaustive vertex processing during time-critical scheduling decisions.
3Device complexity
If a shared GPU is used to reduce costs and complexity, then device complexity is reduced, but reliability of timely information display decreases due to scheduling conflicts
Solution Approach 1:
The patent implements preliminary action by calculating runtime duration estimates for command sequences before they are executed on the shared GPU. This advance estimation allows the instrument cluster module to reserve or prioritize GPU access based on predicted execution times of other modules. By knowing in advance how long other modules will occupy the GPU, the system can schedule instrument cluster rendering to guarantee timely display despite shared resource constraints.
Solution Approach 2:
The fragment count estimation mechanism provides feedback about the computational workload of upcoming command sequences. This feedback information about expected fragment counts and runtime durations allows the GPU scheduler to make informed decisions about resource allocation and timing, ensuring that critical instrument cluster updates receive appropriate priority while still utilizing the shared GPU efficiently.
Data Source
AI summary
A method of estimating a fragment count for the display of at least one three-dimensional (3D) object. The method comprises determining an ellipsoid representative of a set of vertices defined by coordinates of the at least one 3D object, applying a transformation to the ellipsoid, calculating a projection area of the transformed ellipsoid, and estimating the fragment count for the display of the 3D object based at least partly on the calculated projection area of the transformed ellipsoid.