Tessellation Data Compression for GPU Power Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The high power consumption and large bandwidth requirements of graphics processing units (GPUs) during tessellation operations, due to the extensive data reading and writing needed for rendering realistic three-dimensional objects, are not efficiently addressed by existing methods.
Innovation Solution
A method of generating and compressing tessellation data by analyzing patch data to separate shared and non-shared data, and attribute data, and maintaining hierarchical data structures that include pointers for efficient data transmission, reducing redundant data storage and bandwidth usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If the GPU reads and writes extensive data corresponding to primitives for tessellation operations, then the rendering quality of three-dimensional objects is improved, but the power consumption increases
Solution Approach 1:
The patent segments tessellation data into three distinct categories: shared data (common to multiple patches), non-shared data (unique to each patch), and attribute data (control point attributes). This segmentation allows the GPU to process only necessary data for each patch, reducing redundant data reading and writing operations while maintaining rendering quality.
Solution Approach 2:
The patent extracts and separates shared data from patch-specific data, storing shared data once and referencing it across multiple patches. This extraction eliminates redundant storage and transmission of identical data, directly reducing power consumption during data access operations while preserving complete rendering information.
2Manufacturing precision
If the GPU reads and writes extensive data corresponding to primitives for tessellation operations, then the rendering quality of three-dimensional objects is improved, but the bandwidth requirement increases
Solution Approach 1:
By segmenting data into shared and non-shared portions, the patent enables efficient data transmission where shared data is sent once and referenced multiple times,而非 transmitting identical data repeatedly for each patch. This significantly reduces the total quantity of data transmitted through the memory interface.
Solution Approach 2:
The patent extracts shared data elements and creates references or pointers to them, allowing multiple patches to reference the same shared data without duplicating it in memory. This extraction approach reduces the actual data volume stored and transmitted, directly lowering bandwidth requirements.
3Speed
If all patch data is stored individually without compression, then the data access speed is improved, but the memory space requirement increases
Solution Approach 1:
The patent segments data into shared and patch-specific portions, storing shared data once and creating compact references for each patch. This segmentation reduces the total memory footprint while maintaining fast access to both shared and unique data through efficient memory addressing.
Solution Approach 2:
The patent merges shared data into a single storage location that is accessible by multiple patches, eliminating redundant storage. This merging reduces overall memory space requirements while maintaining data access speed through shared memory structures and efficient addressing mechanisms.
Data Source
AI summary
A method of generating tessellation data include analyzing patch data of each of a plurality of patches; generating shared data that is shared by the patches, non-shared data that are not shared by the patches, and attribute data on an attribute of control points of each of the patches from the patch data according to a result of the analyzing; and compressing the non-shared data and the attribute data.