Detail publikace

Use of Multiple GPUs on Shared Memory Multiprocessors for Ultrasound Propagation Simulations

JAROŠ, J.; TREEBY, B.; RENDELL, A. Use of Multiple GPUs on Shared Memory Multiprocessors for Ultrasound Propagation Simulations. In Australasian Symposium on Parallel and Distributed Computing (AusPDC 2012). Melbourne: 2012. p. 43-52. ISBN: 978-1-921770-08-1.
Typ
článek ve sborníku konference
Jazyk
anglicky
Autoři
Jaroš Jiří, doc. Ing., Ph.D. (UPSY)
Treeby Bradley
Rendell Alistair
URL
Klíčová slova

Ultrazvuková simulace, 7-GPU systém, Matlab, FFT, PCI-Express, propustnost, multi-core

Abstrakt

This paper outlines our effort to migrate a compute intensive application of ultrasound propagation being developed in Matlab to a cluster computer where each node has seven GPUs. Our goal is to perform realistic simulations in hours and minutes instead of weeks and days. In order to reach this goal we investigate architecture characteristics of the target system focusing on the PCI-Express subsystem and new features proposed in CUDA version 4.0, especially simultaneous host to device, device to host and peer-to-peer transfers that the application is going to highly benefit from. We also present the results from a CPU based implementation and discuss future directions to exploit multiple GPUs.

Rok
2012
Strany
43–52
Sborník
Australasian Symposium on Parallel and Distributed Computing (AusPDC 2012)
Konference
Australské sympoziom o paralelních a distribuovaných výpočtech (AusPDC 2012), Melbourne, AU
ISBN
978-1-921770-08-1
Místo
Melbourne
EID Scopus
2-s2.0-84869098403
BibTeX 
@inproceedings{BUT91284,
  author="Jiří {Jaroš} and Bradley {Treeby} and Alistair {Rendell}",
  title="Use of Multiple GPUs on Shared Memory Multiprocessors for Ultrasound Propagation Simulations",
  booktitle="Australasian Symposium on Parallel and Distributed Computing (AusPDC 2012)",
  year="2012",
  pages="43--52",
  address="Melbourne",
  isbn="978-1-921770-08-1",
  url="http://crpit.com/confpapers/CRPITV127Jaros.pdf"
}
Soubory
Nahoru