Dynamic Asynchronous Time Warping for GPU Rendering Efficiency

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image rendering mechanisms in AR/VR systems lead to waste of GPU resources due to waiting times and useless rendering, resulting in a low utilization rate of image rendering.

Innovation Solution

An image rendering method that involves receiving a synchronization signal, acquiring and rendering current data for a first viewpoint, and if rendering is not complete by a predetermined T2 moment, stopping the rendering and starting to render data for a second viewpoint, while performing asynchronous time warping on frame images and storing them in a display buffer.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If the rendering process waits for the T/2 moment to start ATW after completing left eye rendering, then the display synchronization is maintained, but GPU resources are wasted due to waiting time and useless rendering

Engineering Contradiction:
Improvedisplay synchronizationVSAvoidGPU utilization rate
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent makes the ATW start time dynamic rather than fixed. The ATW operation now starts at different times depending on when the rendering is actually completed, rather than waiting for a fixed T/2 moment. This dynamic adjustment eliminates idle waiting time while maintaining display synchronization, thereby improving GPU utilization without sacrificing reliability.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent performs ATW operation as soon as rendering is completed, rather than waiting for the scheduled T/2 moment. By taking preliminary action immediately when rendering finishes, the system avoids the idle waiting period and ensures GPU resources are productively used for the next rendering task rather than remaining idle.

Inventive Principle:
Principle #10Preliminary action

2Device complexity

If the rendering process uses a fixed T/2 moment for starting ATW, then the rendering timeline is simple to manage, but there is waste of GPU resources due to waiting periods

Engineering Contradiction:
Improverendering timeline managementVSAvoidGPU resource waste
Core Design Contradiction:
Device complexityVSLoss of energy

Solution Approach 1:

The patent introduces a feedback mechanism where the ATW start time is determined by the actual rendering completion time. The system continuously monitors when rendering finishes and adjusts the ATW start time accordingly, creating a closed-loop control that eliminates idle GPU periods while maintaining manageable complexity through event-driven timing.

Inventive Principle:
Principle #23Feedback

3Manufacturing precision

If the rendering process waits for previous frame completion before starting next frame, then rendering quality is maintained, but rendering throughput decreases due to sequential processing

Engineering Contradiction:
Improverendering qualityVSAvoidrendering throughput
Core Design Contradiction:
Manufacturing precisionVSProductivity

Solution Approach 1:

The patent enables continuous GPU utilization by overlapping rendering operations with ATW operations. Instead of waiting for one operation to complete before starting the next, the system maintains continuous useful action by initiating ATW as soon as rendering completes, allowing the GPU to immediately transition to the next rendering task without idle periods, thereby improving throughput while maintaining quality.

Inventive Principle:
Principle #20Continuity of useful action

Data Source

PatentUS12315029B2Image rendering method, device, equipment and computer-readable storage medium
Publication Date: 2025.05.27 GOERTEK INC
  • US12315029B2 patent drawing
  • US12315029B2 patent drawing
  • US12315029B2 patent drawing

AI summary

An image rendering method including acquiring a current first data to be rendered of a first viewpoint and starting rendering; if the rendering of the first data to be rendered has not been completed at a T2 moment, stopping rendering the first data, and acquiring a current second data to be rendered of the second viewpoint and starting rendering; if the rendering of the first data to be rendered has been completed before the T2 moment, acquiring the second data to be rendered and starting rendering; and at the T2 moment, performing asynchronous time warping on a frame image of the first viewpoint obtained from the last time of rendering and then storing it in a display buffer, and performing asynchronous time warping on a frame image of the second viewpoint obtained from the last time of rendering and then storing it in the display buffer.