Abstract: Terrain masking algorithms have found wide application in military, civil aviation, and meteorological analysis. However, as simulation scale expands and the demand for real-time simulation results increases, traditional computing models are no longer able to meet current requirements. To address this issue, a terrain occlusion detection algorithm is implemented on the compute unified device architecture (CUDA) parallel computing platform, which has successfully solved the issue of slow simulation computation speed. Firstly, the elevation data of discrete sampling points within the radar detection area is matrixed on the CPU end to improve the reading speed of elevation values in parallel computing. Then, the thread allocation method for radar simulation calculation parameters is optimized, and a loop comparison method is utilized to accelerate the calculation of terrain masking angle in parallel. Finally, device side thread synchronization and data alternating transmission technology are adopted to ensure the consistency of calculation results and maximize the utilization of GPU computing resources. Moreover, a multi-mode parallel computing method is employed, which includes both multi-threaded parallel computing and single-threaded serial computing, to support degraded computing when the GPU computing resources are insufficient. This ensures high availability of computing. The experimental results show that, compared to the terrain masking serial computation and multithreaded parallel computation of i7-12700H CPU with a simulation granularity of 3 600 detection beams, it achieves 48 times and 17 times acceleration at 3060 Laptop GPU, respectively. Therefore, this approach provides an effective engineering solution for real-time simulation.

Key words: parallel computing, compute unified device architecture (CUDA), elevation matrix, earth curvature, terrain masking algorithms

摘要： 地形遮蔽算法在军事、民航和气象分析等领域有广泛应用。随着仿真规模的扩大、仿真结果实时性要求越来越高，传统计算模型俨然不能满足当下的实时性要求。为解决这一不足，实现了在统一计算设备架构（CUDA）并行计算平台上的地形遮蔽探测算法，解决了仿真计算速度慢的问题。在CPU端将雷达探测区域内离散采样点的高程数据矩阵化，进而提升高程值在并行化计算中的读取速度。针对雷达仿真计算参数对线程分配方式进行优化，采用循环对比方式对地形遮蔽角的计算进行并行加速。采用设备端线程同步和数据交替传输技术，确保计算结果一致性和最大化利用GPU端计算资源。采用多模式并行化计算模式，使用多线程并行化计算和单线程串行化计算来支撑GPU端计算资源不足时的退化计算，从而保证计算的高可用。实验结果表明，相较于i7-12700H CPU在仿真粒度为3?600条探测波束下的地形遮蔽串行计算和多线程并行计算，在3060 Laptop GPU下分别获取了48倍和17倍加速，为仿真实时性提供了有效的工程解决方案。

关键词: 并行计算, 统一计算设备架构（CUDA）, 高程矩阵, 地球曲率, 地形遮蔽算法