ScholarLens Back to search
2012•Journal of Hefei University of TechnologyRequires access

Study of FDTD algorithm based on CUDA platform

Xianliang Wu

Open publisher page 0 citations

Abstract

In this paper,in view of the deficiencies in traditional finite-difference time-domain(FDTD) algorithm,the FDTD parallel algorithm based on Compute Unified Device Architecture(CUDA) platform is studied and realized by taking the graphics processing unit(GPU) as the core and using theoretical analysis and numerical simulation.The CUDA is a fairly new technology to program inexpensive multi-threaded GPUs.The FDTD algorithm is well suited for parallel processing across spatial domain,thus it is a good candidate for the execution on a GPU.The design and optimization of FDTD algorithm for CUDA programming model are discussed.The results of numerical experiments demonstrate that the computation time of GPU based parallel FDTD algorithm decreases significantly.The GPU based acceleration has been a hot topic in the studies of electromagnetic field numerical computation.

About this research paper

What this paper is about

In this paper,in view of the deficiencies in traditional finite-difference time-domain(FDTD) algorithm,the FDTD parallel algorithm based on Compute Unified Device Architecture(CUDA) platform is studied and realized by taking the graphics processing unit(GPU) as the core and using theoretical analysis and numerical simulation.The CUDA is a fairly new technology to program inexpensive multi-threaded GPUs.The FDTD algorithm is well suited for parallel processing across spatial domain,thus it is a good candidate for the execution on a GPU.The design and optimization of FDTD algorithm for CUDA programming model are discussed.The results of numerical experiments demonstrate that the computation time of GPU based parallel FDTD algorithm decreases significantly.The GPU based acceleration has been a hot topic in the studies of electromagnetic field numerical computation.

Why it matters

A significance statement is not available in the OpenAlex record.

Key contribution

A contribution statement is not available in the OpenAlex record.

Method / approach

Method details are not available in the OpenAlex metadata.

Main findings

Findings are not separately available in the OpenAlex metadata.

Limitations

Limitations are not available in the OpenAlex metadata.

Applications

Application details are not available in the OpenAlex metadata.

Available abstract

In this paper,in view of the deficiencies in traditional finite-difference time-domain(FDTD) algorithm,the FDTD parallel algorithm based on Compute Unified Device Architecture(CUDA) platform is studied and realized by taking the graphics processing unit(GPU) as the core and using theoretical analysis and numerical simulation.The CUDA is a fairly new technology to program inexpensive multi-threaded GPUs.The FDTD algorithm is well suited for parallel processing across spatial domain,thus it is a good candidate for the execution on a GPU.The design and optimization of FDTD algorithm for CUDA programming model are discussed.The results of numerical experiments demonstrate that the computation time of GPU based parallel FDTD algorithm decreases significantly.The GPU based acceleration has been a hot topic in the studies of electromagnetic field numerical computation.

Key concepts: CUDA, Finite-difference time-domain method, Graphics processing unit, Computer science, Computational science, Parallel computing, Computation, General-purpose computing on graphics processing units

Related papers

Related paper 1Related paper 2Related paper 3
Back to paper searchBrowse research topicsOriginal source
Study of FDTD algorithm based on CUDA platform — Research Paper | ScholarLens