Adaptive Implementation Selection in the SkePU Skeleton Programming Library

Usman Dastgeer; Lu Li; Christoph Kessler

doi:10.1007/978-3-642-45293-2_13

Adaptive Implementation Selection in the SkePU Skeleton Programming Library

Usman Dastgeer, Lu Li, Christoph Kessler

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

In earlier work, we have developed the SkePU skeleton programming library for modern multicore systems equipped with one or more programmable GPUs. The library internally provides four types of implementations (implementation variants) for each skeleton: serial C++, OpenMP, CUDA and OpenCL targeting either CPU or GPU execution respectively. Deciding which implementation would run faster for a given skeleton call depends upon the computation, problem size(s), system architecture and data locality.

Original language	English
Title of host publication	Advanced Parallel Processing Technologies
Editors	Chenggang Wu, Albert Cohen
Place of Publication	Berlin, Heidelberg
Publisher	Springer
Pages	170-183
Number of pages	14
ISBN (Electronic)	978-3-642-45293-2
ISBN (Print)	978-3-642-45292-5
DOIs	https://doi.org/10.1007/978-3-642-45293-2_13
Publication status	Published - 2013
Event	International Conference on Advanced Parallel Processing Technologies - Stockholm, Sweden Duration: 27 Aug 2013 → 28 Aug 2013 http://appt2013.tcarch.org/about.asp-bid=1&id=20.html

Conference

Conference	International Conference on Advanced Parallel Processing Technologies
Abbreviated title	APPT2013
Country/Territory	Sweden
City	Stockholm
Period	27/08/13 → 28/08/13
Internet address	http://appt2013.tcarch.org/about.asp-bid=1&id=20.html

Access to Document

10.1007/978-3-642-45293-2_13

https://link.springer.com/chapter/10.1007/978-3-642-45293-2_13

Cite this

@inproceedings{9a7ff8b7f63845cdada617f3444c3926,

title = "Adaptive Implementation Selection in the SkePU Skeleton Programming Library",

abstract = "In earlier work, we have developed the SkePU skeleton programming library for modern multicore systems equipped with one or more programmable GPUs. The library internally provides four types of implementations (implementation variants) for each skeleton: serial C++, OpenMP, CUDA and OpenCL targeting either CPU or GPU execution respectively. Deciding which implementation would run faster for a given skeleton call depends upon the computation, problem size(s), system architecture and data locality.",

author = "Usman Dastgeer and Lu Li and Christoph Kessler",

year = "2013",

doi = "10.1007/978-3-642-45293-2_13",

language = "English",

isbn = "978-3-642-45292-5",

pages = "170--183",

editor = "Chenggang Wu and Albert Cohen",

booktitle = "Advanced Parallel Processing Technologies",

publisher = "Springer",

address = "United Kingdom",

note = "International Conference on Advanced Parallel Processing Technologies, APPT2013 ; Conference date: 27-08-2013 Through 28-08-2013",

url = "http://appt2013.tcarch.org/about.asp-bid=1&id=20.html",

}

Dastgeer, U, Li, L & Kessler, C 2013, Adaptive Implementation Selection in the SkePU Skeleton Programming Library. in C Wu & A Cohen (eds), Advanced Parallel Processing Technologies. Springer, Berlin, Heidelberg, pp. 170-183, International Conference on Advanced Parallel Processing Technologies, Stockholm, Sweden, 27/08/13. https://doi.org/10.1007/978-3-642-45293-2_13

TY - GEN

T1 - Adaptive Implementation Selection in the SkePU Skeleton Programming Library

AU - Dastgeer, Usman

AU - Li, Lu

AU - Kessler, Christoph

PY - 2013

Y1 - 2013

N2 - In earlier work, we have developed the SkePU skeleton programming library for modern multicore systems equipped with one or more programmable GPUs. The library internally provides four types of implementations (implementation variants) for each skeleton: serial C++, OpenMP, CUDA and OpenCL targeting either CPU or GPU execution respectively. Deciding which implementation would run faster for a given skeleton call depends upon the computation, problem size(s), system architecture and data locality.

AB - In earlier work, we have developed the SkePU skeleton programming library for modern multicore systems equipped with one or more programmable GPUs. The library internally provides four types of implementations (implementation variants) for each skeleton: serial C++, OpenMP, CUDA and OpenCL targeting either CPU or GPU execution respectively. Deciding which implementation would run faster for a given skeleton call depends upon the computation, problem size(s), system architecture and data locality.

U2 - 10.1007/978-3-642-45293-2_13

DO - 10.1007/978-3-642-45293-2_13

M3 - Conference contribution

SN - 978-3-642-45292-5

SP - 170

EP - 183

BT - Advanced Parallel Processing Technologies

A2 - Wu, Chenggang

A2 - Cohen, Albert

PB - Springer

CY - Berlin, Heidelberg

T2 - International Conference on Advanced Parallel Processing Technologies

Y2 - 27 August 2013 through 28 August 2013

ER -

Adaptive Implementation Selection in the SkePU Skeleton Programming Library

Abstract

Conference

Access to Document

Fingerprint

Cite this