Real-time dedispersion for fast radio transient surveys, using auto tuning on many-core accelerators

Alessio Sclocco, Joeri van Leeuwen, Henri Bal, Rob V. van Nieuwpoort

Research output: Contribution to JournalArticleAcademicpeer-review

Abstract

Dedispersion, the removal of deleterious smearing of impulsive signals by the interstellar matter, is one of the most intensive processing steps in any radio survey for pulsars and fast transients. We here present a study of the parallelization of this algorithm on many-core accelerators, including GPUs from AMD and NVIDIA, and the Intel Xeon Phi. We find that dedispersion is inherently memory-bound. Even in a perfect scenario, hardware limitations keep the arithmetic intensity low, thus limiting performance. We next exploit auto-tuning to adapt dedispersion to different accelerators, observations, and even telescopes. We demonstrate that the optimal settings differ between observational setups, and that auto-tuning significantly improves performance. This impacts time-domain surveys from Apertif to SKA.

Original languageEnglish
Pages (from-to)1-7
Number of pages7
JournalAstronomy and Computing
Volume14
DOIs
Publication statusPublished - 1 Jun 2016

Fingerprint

Particle accelerators
accelerators
Tuning
tuning
radio
interstellar matter
Telescopes
pulsars
Computer hardware
hardware
telescopes
Data storage equipment
Processing
Pulsars
Graphics processing unit
removal

Keywords

  • Astronomical instrumentation, methods and techniques
  • Pulsars: general
  • Techniques: miscellaneous

Cite this

@article{8ede710fc8ef441b8e02fa0f140d5f2f,
title = "Real-time dedispersion for fast radio transient surveys, using auto tuning on many-core accelerators",
abstract = "Dedispersion, the removal of deleterious smearing of impulsive signals by the interstellar matter, is one of the most intensive processing steps in any radio survey for pulsars and fast transients. We here present a study of the parallelization of this algorithm on many-core accelerators, including GPUs from AMD and NVIDIA, and the Intel Xeon Phi. We find that dedispersion is inherently memory-bound. Even in a perfect scenario, hardware limitations keep the arithmetic intensity low, thus limiting performance. We next exploit auto-tuning to adapt dedispersion to different accelerators, observations, and even telescopes. We demonstrate that the optimal settings differ between observational setups, and that auto-tuning significantly improves performance. This impacts time-domain surveys from Apertif to SKA.",
keywords = "Astronomical instrumentation, methods and techniques, Pulsars: general, Techniques: miscellaneous",
author = "Alessio Sclocco and Leeuwen, {Joeri van} and Henri Bal and {van Nieuwpoort}, {Rob V.}",
year = "2016",
month = "6",
day = "1",
doi = "10.1016/j.ascom.2016.01.001",
language = "English",
volume = "14",
pages = "1--7",
journal = "Astronomy and Computing",
issn = "2213-1337",
publisher = "Elsevier BV",

}

Real-time dedispersion for fast radio transient surveys, using auto tuning on many-core accelerators. / Sclocco, Alessio; Leeuwen, Joeri van; Bal, Henri; van Nieuwpoort, Rob V.

In: Astronomy and Computing, Vol. 14, 01.06.2016, p. 1-7.

Research output: Contribution to JournalArticleAcademicpeer-review

TY - JOUR

T1 - Real-time dedispersion for fast radio transient surveys, using auto tuning on many-core accelerators

AU - Sclocco, Alessio

AU - Leeuwen, Joeri van

AU - Bal, Henri

AU - van Nieuwpoort, Rob V.

PY - 2016/6/1

Y1 - 2016/6/1

N2 - Dedispersion, the removal of deleterious smearing of impulsive signals by the interstellar matter, is one of the most intensive processing steps in any radio survey for pulsars and fast transients. We here present a study of the parallelization of this algorithm on many-core accelerators, including GPUs from AMD and NVIDIA, and the Intel Xeon Phi. We find that dedispersion is inherently memory-bound. Even in a perfect scenario, hardware limitations keep the arithmetic intensity low, thus limiting performance. We next exploit auto-tuning to adapt dedispersion to different accelerators, observations, and even telescopes. We demonstrate that the optimal settings differ between observational setups, and that auto-tuning significantly improves performance. This impacts time-domain surveys from Apertif to SKA.

AB - Dedispersion, the removal of deleterious smearing of impulsive signals by the interstellar matter, is one of the most intensive processing steps in any radio survey for pulsars and fast transients. We here present a study of the parallelization of this algorithm on many-core accelerators, including GPUs from AMD and NVIDIA, and the Intel Xeon Phi. We find that dedispersion is inherently memory-bound. Even in a perfect scenario, hardware limitations keep the arithmetic intensity low, thus limiting performance. We next exploit auto-tuning to adapt dedispersion to different accelerators, observations, and even telescopes. We demonstrate that the optimal settings differ between observational setups, and that auto-tuning significantly improves performance. This impacts time-domain surveys from Apertif to SKA.

KW - Astronomical instrumentation, methods and techniques

KW - Pulsars: general

KW - Techniques: miscellaneous

UR - http://www.scopus.com/inward/record.url?scp=84956536870&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84956536870&partnerID=8YFLogxK

U2 - 10.1016/j.ascom.2016.01.001

DO - 10.1016/j.ascom.2016.01.001

M3 - Article

VL - 14

SP - 1

EP - 7

JO - Astronomy and Computing

JF - Astronomy and Computing

SN - 2213-1337

ER -