GPU performance optimization analysis method based on GPU rendering program benchmark test set

By building a representative and extensive GPU rendering benchmark test set and combining machine learning algorithms to optimize the GPU performance, the problem of lack of representativeness and optimization indicators of GPU rendering benchmark test set in the existing technology is solved, and more accurate GPU performance evaluation and optimization.

CN119006672AInactive Publication Date: 2024-11-22SHENZHEN INST OF ADVANCED TECH CHINESE ACAD OF SCI
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202310568089.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2023-05-18
Publication Date
2024-11-22
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing GPU rendering benchmark test set lacks representation and breadth, fails to fully cover all mainstream rendering algorithms, and fails to provide effective GPU performance optimization metrics.

Method used

Build a benchmark test set based on GPU rendering programs, optimize the analysis of GPU performance through feature mining and machine learning algorithms, including selecting multiple rendering programs, extracting GPU features, and using methods such as random forest, factor analysis and principal component analysis. Conduct importance sorting to obtain GPU performance optimization indicators.

Benefits of technology

A representative and extensive set of GPU rendering benchmarks is implemented, enabling more accurate evaluation of GPU performance, providing in-depth optimization suggestions, and improving virtual reality experience and graphics rendering quality.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119006672A_ABST
    Figure CN119006672A_ABST
Patent Text Reader

Abstract

The invention discloses a GPU performance optimization analysis method based on a GPU rendering program benchmark test set. The method comprises the following steps: with universality and representativeness as targets, selecting multiple rendering programs from different types of GPU rendering programs to construct a GPU rendering benchmark test set; performing feature mining on the GPU rendering benchmark test set, extracting effective GPU features corresponding to each rendering program, and constructing a GPU system structure related feature set; and performing importance sorting on the related features of the GPU architecture by using multiple machine learning algorithms to obtain GPU performance optimization index suggestions. According to the method, the GPU rendering program is subjected to feature mining, and all the obtained GPU performance features are sorted in combination with a machine learning sorting algorithm, so that the rendering GPU can be more accurately and efficiently helped to find GPU performance related features and give rendering GPU performance optimization suggestions.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Cited By

  • A method for optimizing computing power configuration of intelligent micro-host graphics processor

    CN119762326A

  • Graph rendering mechanism identification method based on rendering pipeline alignment

    CN120492361A