Parallel Computation of EM Backscattering from Large Three-Dimensional Sea Surface with CUDA.

Area of Science:

• Electromagnetics
• Computational Physics
• Oceanography

Background:

• Accurate electromagnetic (EM) backscattering analysis of large 3D sea surfaces is crucial for radar remote sensing.
• Current computational models, like the slope-deterministic composite scattering model (SDCSM), become inefficient for high-resolution sea surfaces with millions of facets.
• Existing methods struggle with the computational demands of simulating complex sea surface interactions.

Purpose of the Study:

• To develop and verify an efficient parallel computation method for EM backscattering from large 3D sea surfaces using graphics processing units (GPUs).
• To optimize the calculation of normalized radar cross section (NRCS) for sea surfaces.
• To demonstrate the feasibility of GPU acceleration for the SDCSM.

Main Methods:

• Implemented a slope-deterministic composite scattering model (SDCSM) on an NVIDIA Tesla K80 GPU.
• Utilized four Compute Unified Device Architecture (CUDA) optimization strategies: coalesced memory access, constant memory, fast math compiler options, and asynchronous data transfer.
• Analyzed the impact of block size and registers per thread on computation speed.

Main Results:

• Achieved a significant speedup of 748.26x for GPU-based SDCSM calculations compared to a CPU-based implementation.
• Demonstrated the efficiency of GPU acceleration for complex EM backscattering simulations.
• Validated the effectiveness of CUDA optimization techniques for large-scale sea surface modeling.

Conclusions:

• GPU acceleration provides a highly efficient solution for computing EM backscattering from large 3D sea surfaces.
• The developed GPU-based SDCSM significantly reduces computation time, enabling more detailed sea surface analysis.
• This approach enhances the feasibility of high-resolution radar studies of oceanic environments.