Detail publikace

Accelerating Hybrid Local Domain Decomposition for the k-Wave Toolbox on Multi-GPU Systems

KUNÍK, O. JAROŠ, J.

Originální název

Accelerating Hybrid Local Domain Decomposition for the k-Wave Toolbox on Multi-GPU Systems

Typ

prezentace, poster

Jazyk

angličtina

Originální abstrakt

The k-Wave toolbox is designed for high-fidelity acoustic wave simulations using Fourier collocation for spatial derivatives, but its performance is constrained by communication overhead on multi-CPU and multi-GPU systems. We present a hybrid local domain decomposition approach that partitions the simulation domain into subdomains, each assigned to a GPU with configurable resolution. Using CUDA and cuFFT for Fourier transforms and NVLink for halo exchanges, our method minimizes inter-subdomain communication and accelerates multi-GPU performance. Testing on the Karolina supercomputer shows strong scalability and accuracy, especially with uniform-resolution subdomains, and proves effective even for large-scale simulations beyond single-GPU memory limits.

Klíčová slova

k-Wave, HPC, Hybrid decomposition, Local decomposition, CUDA, Multi-GPU

Autoři

KUNÍK, O.; JAROŠ, J.

Vydáno

4. 11. 2024

Místo

Ostrava

Strany počet

1

URL

https://www.fit.vut.cz/research/publication/13293/

BibTex

@misc{BUT193367,
  author="Oliver {Kuník} and Jiří {Jaroš}",
  title="Accelerating Hybrid Local Domain Decomposition for the k-Wave Toolbox on Multi-GPU Systems",
  year="2024",
  pages="1",
  address="Ostrava",
  url="https://www.fit.vut.cz/research/publication/13293/",
  note="presentation, poster"
}

Dokumenty

ikunik-A1-posterIT4I.pdf