MRMark:面向三维视频融合的机器性能评测方法

MRMark:A Machine Performance Evaluation Method for Three-dimensional Video Fusion

三维视频融合系统涉及视频和图形的计算,对机器性能有较高要求,但现有商用评测工具缺乏对应考虑,会出现机器评分与实际运行性能不相符的情况。提出一种面向三维视频融合的机器性能评测方法,并实现了评测工具MRMark。该方法根据视频融合系统的特点设计了多组标准三维测试场景,统计测试过程中的帧率、播放的视频路数等参数,来计算GPU、CPU分数和耦合分数,以及机器的整体性能分数。实验结果表明,相比现有评测工具,该工具的评测结果能更有效地检测机器进行三维视频融合计算的能力。

The three-dimensional video fusion system needs the calculation of video and graphics,and requires high machine performance.The existing commercial evaluation tools lack the corresponding considerations,which may cause the inconsistency of a machine's performance from the score.A machine performance evaluation method for three-dimensional video fusion is proposed and an evaluation tool MRMark is implemented.Several sets of standard 3D test scenes are designed according to the characteristics of the video fusion system,the statistics on the frame rate,the number of video channels played and the other parameters during the test are collected to calculate the GPU,CPU and coupling score,as well as the overall performance score of the machine.The results show that,compared to the existing evaluation tools,the tool can more effectively detect the ability of the machine to perform the three-dimensional video fusion calculations.