MPI-INF D4 Publications: Proceedings Article: Cache oblivious parallelograms in iterative stencil computations

Proceedings Article, Paper
@InProceedings
Beitrag in Tagungsband, Workshop

Show entries of:

this year (2013) | last year (2012) | two years ago (2011) | Notes URL

Action:

Options:

Library locked

Goto entry point

Author, Editor

Author(s):

Strzodka, Robert
Shaheen, Mohammed
Pajak, Dawid
Seidel, Hans-Peter

dblp
dblp
dblp
dblp

Not MPG Author(s):

Pajak, Dawid

Editor(s):

BibTeX cite key*:

StShPa_10CORALS

Title, Booktitle

Title*:	Cache oblivious parallelograms in iterative stencil computations
Attachment(s):	CORALS.pdf (215.87 KB)
Booktitle*:	ICS '10 : Proceedings of the 24th ACM International Conference on Supercomputing

Event, URLs

URL of the conference:	http://pcsostres.ac.upc.edu/ics-conference/archive/ics10/
URL for downloading the paper:	http://doi.acm.org/10.1145/1810085.1810096
Event Address*:	Tsukuba, Ibaraki, Japan	Language:	English
Event Date* (no longer used):		Organization:	Association for Computing Machinery (ACM)
Event Start Date:	1 June 2010	Event End Date:	5 June 2010

Publisher

Name*:	ACM	URL:
Address*:	New York, NY	Type:

Vol, No, Year, pp.

Series:

Volume:		Number:
Month:		Pages:	49-59
Year*:	2010	VG Wort Pages:
ISBN/ISSN:	978-1-4503-0018-6	Sequence Number:
DOI:	10.1145/1810085.1810096


(LaTeX) Abstract:	We present a new cache oblivious scheme for iterative stencil computations that performs beyond system bandwidth limitations as though gigabytes of data could reside in an enormous on-chip cache. We compare execution times for 2D and 3D spatial domains with up to 128 million double precision elements for constant and variable stencils against hand-optimized naive code and the automatic polyhedral parallelizer and locality optimizer PluTo and demonstrate the clear superiority of our results. The performance benefits stem from a tiling structure that caters for data locality, parallelism and vectorization simultaneously. Rather than tiling the iteration space from inside, we take an exterior approach with a predefined hierarchy, simple regular parallelogram tiles and a locality preserving parallelization. These advantages come at the cost of an irregular work-load distribution but a tightly integrated load-balancer ensures a high utilization of all resources.
URL for the Abstract:	http://www.mpi-inf.mpg.de/~strzodka/papers/info/StShPa_10CORALS.htm


Download Access Level:	Public

Correlation

MPG Unit:	Max-Planck-Institut für Informatik

MPG Subunit:	Computer Graphics Group
Appearance:	MPII WWW Server, MPII FTP Server, MPG publications list, university publications list, working group publication list, Fachbeirat, VG Wort

BibTeX Entry:

@INPROCEEDINGS{StShPa_10CORALS,
AUTHOR = {Strzodka, Robert and Shaheen, Mohammed and Pajak, Dawid and Seidel, Hans-Peter},
TITLE = {Cache oblivious parallelograms in iterative stencil computations},
BOOKTITLE = {ICS '10 : Proceedings of the 24th ACM International Conference on Supercomputing},
PUBLISHER = {ACM},
YEAR = {2010},
ORGANIZATION = {Association for Computing Machinery (ACM)},
PAGES = {49--59},
ADDRESS = {Tsukuba, Ibaraki, Japan},
ISBN = {978-1-4503-0018-6},
DOI = {10.1145/1810085.1810096},
}

Entry last modified by Anja Becker, 03/11/2011

Edit History (please click the blue arrow to see the details)

Attachment Section

View attachments here:

CORALS.pdf

CORALS.pdf