Lithe: Enabling Efficient Composition of Parallel Libraries
H. Pan, B. Hindman, and K. Asanović. Proceedings of the First USENIX conference on Hot topics in parallelism, page 6. Berkeley, CA, USA, USENIX Association, (2009)
Abstract
For the software industry to take advantage of multicore processors, we must allow programmers to arbitrarily compose parallel libraries without sacrificing performance. We argue that high-level task or thread abstractions and a common global scheduler cannot provide effective library composition. Instead, the operating system should expose unvirtualized processing resources that can be shared cooperatively between parallel libraries within an application. In this paper, we describe a system that standardizes and facilitates the exchange of these unvirtualized processing resources between libraries.
%0 Conference Paper
%1 Pan:2009:LEE:1855591.1855602
%A Pan, Heidi
%A Hindman, Benjamin
%A Asanović, Krste
%B Proceedings of the First USENIX conference on Hot topics in parallelism
%C Berkeley, CA, USA
%D 2009
%I USENIX Association
%K Scheduling Threads
%P 6
%T Lithe: Enabling Efficient Composition of Parallel Libraries
%U http://portal.acm.org/citation.cfm?id=1855591.1855602
%X For the software industry to take advantage of multicore processors, we must allow programmers to arbitrarily compose parallel libraries without sacrificing performance. We argue that high-level task or thread abstractions and a common global scheduler cannot provide effective library composition. Instead, the operating system should expose unvirtualized processing resources that can be shared cooperatively between parallel libraries within an application. In this paper, we describe a system that standardizes and facilitates the exchange of these unvirtualized processing resources between libraries.
@inproceedings{Pan:2009:LEE:1855591.1855602,
abstract = {For the software industry to take advantage of multicore processors, we must allow programmers to arbitrarily compose parallel libraries without sacrificing performance. We argue that high-level task or thread abstractions and a common global scheduler cannot provide effective library composition. Instead, the operating system should expose unvirtualized processing resources that can be shared cooperatively between parallel libraries within an application. In this paper, we describe a system that standardizes and facilitates the exchange of these unvirtualized processing resources between libraries.},
acmid = {1855602},
added-at = {2011-08-10T21:46:42.000+0200},
address = {Berkeley, CA, USA},
author = {Pan, Heidi and Hindman, Benjamin and Asanovi\'{c}, Krste},
biburl = {https://www.bibsonomy.org/bibtex/2783d4567cb6b3bf266d0674a5195f626/gron},
booktitle = {Proceedings of the First USENIX conference on Hot topics in parallelism},
description = {Lithe},
interhash = {b7fed0316b1a8568cf2b0b40aa415d4c},
intrahash = {783d4567cb6b3bf266d0674a5195f626},
keywords = {Scheduling Threads},
location = {Berkeley, California},
numpages = {1},
pages = 6,
publisher = {USENIX Association},
series = {HotPar'09},
timestamp = {2011-08-10T21:46:42.000+0200},
title = {Lithe: Enabling Efficient Composition of Parallel Libraries},
url = {http://portal.acm.org/citation.cfm?id=1855591.1855602},
year = 2009
}