跨平台的海量波形数据并行绘制算法

A Cross-platform Parallel Painting Algorithm for Massive Waveform Data

针对当前第三方电力暂态数据分析软件在绘制海量波形数据时出现的效率低下、反应缓慢情况,结合多核并行计算技术,提出了一种可跨平台的海量COMTRADE波形数据并行绘制算法及其技术。该算法在分析传统串行绘制系统内部关系的基础上,提出了基于并行绘制的新型关系:将原有单一图层分为波形图层和用户控制图层,其中以并行方式绘制波形图层,而后通过融合图层方式完成最终绘制。通过试验分析Windows和UNIX下的各种图形绘制技术,找到了最适合海量波形数据并行绘制的跨平台技术组合:"QImage+QPainter"技术模式。结合跨平台的线程库Pthreads,详细论述了并行绘制算法的每个步骤,给出了让通道绘制线程能负载平衡运行的绘制工作量均分公式,同时给出了通道绘制线程和图层融合详细算法的伪码。试验证明所提出的并行绘制算法可获得较大加速比,并可随着绘制工作量的加大和CPU核的增多,获得线性加速比。

Since the current third party power system analysis software is low efficient and slow in printing massive waveform data, a cross-platform massive wave COMTRADE data parallel painting algorithm are proposed, based on the parallel multicore computing technology, on the basis of the internal relationship of tradition serial painting system, a new relationship based on the parallel printing is presented in the algorithm, the original single layer is divided into waveform layer and user control layer, the waveform layer is printed in parallel mode and the final printing is done by the way of image composition, By analyzing the printing technologies of Windows and Unix, Qlmage＋Qpainter is found to be the most suitable cross-platform technology for massive waveform parallel painting. Each step of the parallel printing algorithm is elaborated with a cross- platform thread library named Pthreads. the workload sharing formulas are given which can make channel printing thread run load balancing. The pseudo-code of the channel painting thread and image composition algorithms are presented in detail. Test results show that the proposed parallel printing algorithm has a high acceleration ratio and even linear acceleration ratio with the increasing of printing workload and central process unit （CPU） core number.