GPU Tessellation Bandwidth Reduction via Texture Patch Sorting

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

The increase in bandwidth and power consumption associated with tessellation operations in graphics processing units (GPUs) due to the need to read and write large volumes of data for rendering finer images.

Innovation Solution

A GPU method that generates and rearranges display lists based on texture patch descriptors, removing invisible primitives and optimizing memory access by using a position-attribute map to sort and render primitives efficiently, thereby reducing memory-access bandwidth and power consumption.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If the GPU performs tessellation operation to generate finer image primitives, then the image rendering quality is improved, but the memory-access bandwidth and power consumption increase significantly

Engineering Contradiction:

Improveimage rendering qualityVSAvoidpower consumption

Core Design Contradiction:

Manufacturing precisionVSLoss of energy

Solution Approach 1:

The patent applies preliminary action by pre-sorting primitives into display lists organized by texture patch descriptors before the actual rendering process. This advance organization of primitive data according to texture patch requirements enables the rendering processor to access textures more efficiently, reducing redundant memory reads and lowering overall power consumption during tessellation-based rendering operations

Inventive Principle:

Principle #10Preliminary action

2Manufacturing precision

If the GPU reads and writes large volumes of data for tessellation operation, then the rendering detail is improved, but the memory-access bandwidth is wasted

Engineering Contradiction:

Improverendering detailVSAvoidmemory-access bandwidth

Core Design Contradiction:

Manufacturing precisionVSLoss of energy

Solution Approach 1:

The patent segments the rendering process into distinct phases: a geometry processor that generates primitives with texture patch descriptor indices, a texture patch processor that sorts primitives into display lists by texture patch descriptors, and a rendering processor that executes rendering operations. This segmentation allows each stage to optimize data organization and access patterns, reducing unnecessary bandwidth consumption while maintaining rendering detail

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The texture patch descriptor acts as an intermediary that links primitives to their required texture patches. By using this intermediary index structure, the system can efficiently match primitives with their texture data without requiring large volumes of raw data to be read and written, thereby reducing memory-access bandwidth waste while preserving rendering quality

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS9576396B2Graphics processing unit, graphics processing system including the same, and method of operating the same

Publication Date: 2017.02.21 SAMSUNG ELECTRONICS CO LTD

AI summary

A method of operating a graphics processing unit includes generating a first display list including a position information index, an attribute information index, and a texture patch descriptor index for each of a plurality of new primitives; generating a second display list by sorting the first display list according to texture patch descriptor indexes; and rendering the plurality of new primitives based on the second display list.