基于GPU的液晶大气湍流模拟器波面生成的并行实现

为了使液晶大气湍流模拟器具有实时大气模拟能力,在GPU通用计算架构下提出了基于GPU的液晶大气湍流模拟器实时波面生成计算方法。针对液晶湍流模拟器高分辨率、高精度的特点介绍了湍流波面生成计算方法,论述了CUDA通用计算架构。建立基于GPU的波面生成模型,并对该模型进行了并行化优化和共享存储器优化。给出了采用CPU与GPU进行波面生成的实验对比结果。结果表明:采用GPU生成分辨率为256×256,192项Zernike多项式进行波面生成的平均时间为2.5ms,生成速度比CPU少两个量级,满足实时波面生成的要求。

Parallel implementation of liquid crystal atmosphere turbulence simulator wavefront generation based on GPU

In order to enable the liquid crystal atmosphere turbulence simulator with the ability to simulate the atmospheric turbulence in real time, this paper puts forward a new method of liquid crystal atmosphere turbulence simulator wavefront generation based on Compute Unified Device Architecture. The calculation method of wavefront generation is introduced according to the characteristics of liquid crystal atmosphere turbulence simulator, which has high resolution and high precision. The Compute Unified Device Architecture is discussed. Then, a wavefront generation model is built with parallel optimization and share memory optimization. Finally, contrast experimental results by using CPU and GPU to generate the distortion wavefront are given. The results show that it takes only 2 ms for GPU to generate a wavefront with a resolution of 256256 and 192 Zernike polynomials, which is two orders of magnitude less than that using CPU, and satisfies the real-time wavefront generate requirements.