Faculty of Information Technology, BUT

Publication Details

Spectral Domain Decomposition Using Local Fourier Basis: Application to Ultrasound Simulation on a Cluster of GPUs

JAROŠ Jiří, VAVERKA Filip and TREEBY Bradley E. Spectral Domain Decomposition Using Local Fourier Basis: Application to Ultrasound Simulation on a Cluster of GPUs. International Journal of Supercomputing Frontiers and Innovations, vol. 3, no. 3, pp. 39-54. ISSN 2313-8734. Available from: http://superfri.org/superfri/article/download/108/199
Czech title
Dekompozice spektrálních metod pomocí lokálních Fourierových bází
Type
journal article
Language
english
Authors
Jaroš Jiří, doc. Ing., Ph.D. (DCSY FIT BUT)
Vaverka Filip, Ing. (DCSY FIT BUT)
Treeby Bradley E. (UCL)
URL
Keywords
Spectral methods, Local Fourier Basis domain decomposition, Ultrasound simulations, GPUs, HPC, Supercomputing, cancer treatment.
Abstract
The simulation of ultrasound wave propagation through biological tissue has a wide range of practical applications. However, large grid sizes are generally needed to capture the phenomena of interest. Here, a novel approach to reduce the computational complexity is presented. The model uses an accelerated k-space pseudospectral method which enables more than one hundred GPUs to be exploited to solve problems with more than 3*10^9 grid points. The classic communication bottleneck of Fourier spectral methods, all-to-all global data exchange, is overcome by the application of domain decomposition using local Fourier basis. Compared to global domain decomposition, for a grid size of 1536 x 1024 x 2048, this reduces the simulation time by a factor of 7.5 and the simulation cost by a factor of 3.8.
Annotation
The simulation of ultrasound wave propagation through biological tissue has a wide range of practical applications. However, large grid sizes are generally needed to capture the phenomena of interest. Here, a novel approach to reduce the computational complexity is presented. The model uses an accelerated k-space pseudospectral method which enables more than one hundred GPUs to be exploited to solve problems with more than 3 x 10^9 grid points. The classic communication bottleneck of Fourier spectral methods, all-to-all global data exchange, is overcome by the application of domain decomposition using local Fourier basis. Compared to global domain decomposition, for a grid size of 1536 x 1024 x 2048, this reduces the simulation time by a factor of 7.5 and the simulation cost by a factor of 3.8.
Published
2016
Pages
39-54
Journal
International Journal of Supercomputing Frontiers and Innovations, vol. 3, no. 3, ISSN 2313-8734
Publisher
Association for Computing Machinery
DOI
BibTeX
@ARTICLE{FITPUB11149,
   author = "Ji\v{r}\'{i} Jaro\v{s} and Filip Vaverka and E. Bradley Treeby",
   title = "Spectral Domain Decomposition Using Local Fourier Basis: Application to Ultrasound Simulation on a Cluster of GPUs",
   pages = "39--54",
   journal = "International Journal of Supercomputing Frontiers and Innovations",
   volume = 3,
   number = 3,
   year = 2016,
   ISSN = "2313-8734",
   doi = "10.14529/jsfi1603",
   language = "english",
   url = "https://www.fit.vut.cz/research/publication/11149"
}
Back to top