GPU Image Rendering and Encoding Without CPU-GPU Data Replication

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image rendering and coding processes in cloud gaming consume excessive CPU and GPU resources due to the need for data replication and format conversion between processors, leading to reduced efficiency.

Innovation Solution

The process is optimized by performing rendering, format conversion, and coding directly on a second processor without additional data transmission, using dedicated application programming interfaces to manage storage addresses and convert data formats directly on the second processor.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If data is replicated between CPU and GPU for format conversion and coding, then image processing can be completed, but processor resources are consumed and efficiency is reduced

Engineering Contradiction:
Improverendering and coding efficiencyVSAvoidprocessor resource consumption
Core Design Contradiction:
ProductivityVSUse of energy by moving object

Solution Approach 1:

The patent merges rendering, format conversion, and video coding operations into a single integrated process executed by the GPU. The GPU renders image data in RGBA format, converts it to YUV format, and performs H.264/H.265 coding all within its own memory space, eliminating the need to replicate data between CPU and GPU. This consolidation reduces processor resource consumption and improves rendering and coding efficiency.

Inventive Principle:
Principle #5Merging (Combining)

2Ease of operation

If data transmission between CPU and GPU is performed multiple times, then format conversion and coding can be achieved, but additional resources are consumed and latency increases

Engineering Contradiction:
Improvedata transmission simplicityVSAvoidprocessing latency
Core Design Contradiction:
Ease of operationVSLoss of time

Solution Approach 1:

The patent applies preliminary action by having the GPU perform format conversion and coding operations directly on the rendered data before it leaves the GPU memory. Instead of transmitting raw RGBA data to the CPU for conversion and then receiving converted data back for coding, the GPU prepares the data in the required format (YUV) and encodes it in advance, reducing the number of transmission cycles and minimizing processing latency.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3972253B1Image rendering and encoding method, and related apparatus
Publication Date: 2025.10.01 HUAWEI TECH CO LTD
  • EP3972253B1 patent drawingFigure 1
  • EP3972253B1 patent drawingFigure 2~3
  • EP3972253B1 patent drawingFigure 4

AI summary

Embodiments of the present invention provide an image rendering and coding method and a related apparatus. The method includes: first sending, by a first processor, to-be-rendered data to a second processor; then indicating the second processor to obtain first format data through rendering based on the to-be-rendered data, where the first format data is stored in first storage space of the second processor; indicating, by the first processor, the second processor to convert the first format data into second format data; and indicating the second processor to code the second format data into third format data, where a data capacity of the third format data is less than a data capacity of the second format data; and then sending the third format data to a client. According to the embodiments of the present invention, image rendering and coding efficiency can be improved.