研究目的
To accelerate the computation of the fast Fourier transform beam propagation method (FFT-BPM) using GPU and multi-core processors for faster processing of optical beam propagation.
研究成果
The proposed parallel FFT-BPM significantly accelerates the computation of optical beam propagation, with speedup factors of 150× and 5× using GPU and eight-core multiprocessor, respectively, compared to traditional CPU. This method is particularly effective for high-resolution beams and large numbers of propagation steps, offering a practical solution for time-consuming calculations in optical device design and analysis.
研究不足
The study is limited by the computational capabilities of the hardware used (GPU and multi-core processors). The accuracy and speed of the FFT-BPM are dependent on the resolution of the input beam and the number of propagation steps.
1:Experimental Design and Method Selection:
The FFT-BPM is implemented on GPU and multi-core processors to speed up computations. The methodology includes the use of CUDA for GPU and OpenMP for multi-core processing.
2:Sample Selection and Data Sources:
The study uses a Gaussian light beam as the input for propagation through optical devices, with specific focus on the Goos-H?nchen shift calculation.
3:List of Experimental Equipment and Materials:
NVIDIA Tesla K20C and Quadro K2000 GPUs, Intel? Xeon(R) CPU E5-2609 v2 @ 2.50 GHz.
4:50 GHz.
Experimental Procedures and Operational Workflow:
4. Experimental Procedures and Operational Workflow: Initialization of beam parameters on CPU, followed by parallel computation of FFT, IFFT, and complex multiplication on GPU or multi-core processors.
5:Data Analysis Methods:
Comparison of execution times and speedup factors between serial CPU, multi-core, and GPU implementations.
独家科研数据包,助您复现前沿成果,加速创新突破
获取完整内容