SkelCL - A Portable Skeleton Library for High-Level GPU Programming

M. Steuwer; P. Kegel; S. Gorlatch

doi:10.1109/IPDPS.2011.269

SkelCL - A Portable Skeleton Library for High-Level GPU Programming

M. Steuwer, P. Kegel, S. Gorlatch

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

Abstract

While CUDA and OpenCL made general-purpose programming for Graphics Processing Units (GPU) popular, using these programming approaches remains complex and error-prone because they lack high-level abstractions. The especially challenging systems with multiple GPU are not addressed at all by these low-level programming models. We propose SkelCL - a library providing so-called algorithmic skeletons that capture recurring patterns of parallel computation and communication, together with an abstract vector data type and constructs for specifying data distribution. We demonstrate that SkelCL greatly simplifies programming GPU systems. We report the competitive performance results of SkelCL using both a simple Mandelbrot set computation and an industrial-strength medical imaging application. Because the library is implemented using OpenCL, it is portable across GPU hardware of different vendors.

Original language	English
Title of host publication	Parallel and Distributed Processing Workshops and Phd Forum (IPDPSW), 2011 IEEE International Symposium on
Publisher	Institute of Electrical and Electronics Engineers
Pages	1176-1182
Number of pages	7
ISBN (Print)	978-1-61284-425-1
DOIs	https://doi.org/10.1109/IPDPS.2011.269
Publication status	Published - 1 May 2011

Keywords / Materials (for Non-textual outputs)

abstract data types
coprocessors
formal specification
parallel programming
software libraries
CUDA
Mandelbrot set computation
OpenCL
SkelCL
abstract vector data type
algorithmic skeletons
data distribution specification
general-purpose programming
graphics processing units
high-level GPU programming
industrial-strength medical imaging application
parallel communication
parallel computation
portable skeleton library
Graphics processing unit
Image reconstruction
Kernel
Performance evaluation
Programming
Runtime
Skeleton

Access to Document

10.1109/IPDPS.2011.269

http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=6008967

Cite this

@inproceedings{536d8cc0702c4a39a2458956ed422243,

title = "SkelCL - A Portable Skeleton Library for High-Level GPU Programming",

abstract = "While CUDA and OpenCL made general-purpose programming for Graphics Processing Units (GPU) popular, using these programming approaches remains complex and error-prone because they lack high-level abstractions. The especially challenging systems with multiple GPU are not addressed at all by these low-level programming models. We propose SkelCL - a library providing so-called algorithmic skeletons that capture recurring patterns of parallel computation and communication, together with an abstract vector data type and constructs for specifying data distribution. We demonstrate that SkelCL greatly simplifies programming GPU systems. We report the competitive performance results of SkelCL using both a simple Mandelbrot set computation and an industrial-strength medical imaging application. Because the library is implemented using OpenCL, it is portable across GPU hardware of different vendors.",

keywords = "abstract data types, coprocessors, formal specification, parallel programming, software libraries, CUDA, Mandelbrot set computation, OpenCL, SkelCL, abstract vector data type, algorithmic skeletons, data distribution specification, general-purpose programming, graphics processing units, high-level GPU programming, industrial-strength medical imaging application, parallel communication, parallel computation, portable skeleton library, Graphics processing unit, Image reconstruction, Kernel, Performance evaluation, Programming, Runtime, Skeleton",

author = "M. Steuwer and P. Kegel and S. Gorlatch",

year = "2011",

month = may,

day = "1",

doi = "10.1109/IPDPS.2011.269",

language = "English",

isbn = "978-1-61284-425-1",

pages = "1176--1182",

booktitle = "Parallel and Distributed Processing Workshops and Phd Forum (IPDPSW), 2011 IEEE International Symposium on",

publisher = "Institute of Electrical and Electronics Engineers",

address = "United States",

}

SkelCL - A Portable Skeleton Library for High-Level GPU Programming. / Steuwer, M.; Kegel, P.; Gorlatch, S.
Parallel and Distributed Processing Workshops and Phd Forum (IPDPSW), 2011 IEEE International Symposium on. Institute of Electrical and Electronics Engineers, 2011. p. 1176-1182.

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

TY - GEN

T1 - SkelCL - A Portable Skeleton Library for High-Level GPU Programming

AU - Steuwer, M.

AU - Kegel, P.

AU - Gorlatch, S.

PY - 2011/5/1

Y1 - 2011/5/1

N2 - While CUDA and OpenCL made general-purpose programming for Graphics Processing Units (GPU) popular, using these programming approaches remains complex and error-prone because they lack high-level abstractions. The especially challenging systems with multiple GPU are not addressed at all by these low-level programming models. We propose SkelCL - a library providing so-called algorithmic skeletons that capture recurring patterns of parallel computation and communication, together with an abstract vector data type and constructs for specifying data distribution. We demonstrate that SkelCL greatly simplifies programming GPU systems. We report the competitive performance results of SkelCL using both a simple Mandelbrot set computation and an industrial-strength medical imaging application. Because the library is implemented using OpenCL, it is portable across GPU hardware of different vendors.

AB - While CUDA and OpenCL made general-purpose programming for Graphics Processing Units (GPU) popular, using these programming approaches remains complex and error-prone because they lack high-level abstractions. The especially challenging systems with multiple GPU are not addressed at all by these low-level programming models. We propose SkelCL - a library providing so-called algorithmic skeletons that capture recurring patterns of parallel computation and communication, together with an abstract vector data type and constructs for specifying data distribution. We demonstrate that SkelCL greatly simplifies programming GPU systems. We report the competitive performance results of SkelCL using both a simple Mandelbrot set computation and an industrial-strength medical imaging application. Because the library is implemented using OpenCL, it is portable across GPU hardware of different vendors.

KW - abstract data types

KW - coprocessors

KW - formal specification

KW - parallel programming

KW - software libraries

KW - CUDA

KW - Mandelbrot set computation

KW - OpenCL

KW - SkelCL

KW - abstract vector data type

KW - algorithmic skeletons

KW - data distribution specification

KW - general-purpose programming

KW - graphics processing units

KW - high-level GPU programming

KW - industrial-strength medical imaging application

KW - parallel communication

KW - parallel computation

KW - portable skeleton library

KW - Graphics processing unit

KW - Image reconstruction

KW - Kernel

KW - Performance evaluation

KW - Programming

KW - Runtime

KW - Skeleton

U2 - 10.1109/IPDPS.2011.269

DO - 10.1109/IPDPS.2011.269

M3 - Conference contribution

SN - 978-1-61284-425-1

SP - 1176

EP - 1182

BT - Parallel and Distributed Processing Workshops and Phd Forum (IPDPSW), 2011 IEEE International Symposium on

PB - Institute of Electrical and Electronics Engineers

ER -

SkelCL - A Portable Skeleton Library for High-Level GPU Programming

Abstract

Keywords / Materials (for Non-textual outputs)

Access to Document

Fingerprint

Cite this