HQL: A scalable synchronization mechanism for GPUs

Ayse Yilmazer; David Kaeli

doi:10.1109/IPDPS.2013.82

HQL: A scalable synchronization mechanism for GPUs

Ayse Yilmazer, David Kaeli

Northeastern University

Research output: Contribution to conference › Paper › peer-review

20 Citations (Scopus)

Abstract

Modern GPUs rely on atomic operations to perform global communication. These atomic operations can be used to construct finer-grained locks to provide support for mutual exclusion. However, equipped with only these basic synchronization primitives to support mutual exclusion results in inefficient use of resources. In this paper, we propose a new hardware-based blocking synchronization mechanism which uses hierarchical queuing for scalability and efficiency. We evaluate our design using a set of GPU applications for stressing synchronization mechanisms. We perform detailed simulation utilizing the Multi2Sim heterogeneous simulation infrastructure. Our results indicate that we can reduce the number of instructions executed by a GPU application by as much as 84%, while improving execution performance by as much as 73%.

Original language	English
Pages	475-486
Number of pages	12
DOIs	https://doi.org/10.1109/IPDPS.2013.82
Publication status	Published - 2013
Externally published	Yes
Event	27th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2013 - Boston, MA, United States Duration: 20 May 2013 → 24 May 2013

Conference

Conference	27th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2013
Country/Territory	United States
City	Boston, MA
Period	20/05/13 → 24/05/13

Keywords

GPUs
Mutual-exclusion
synchronization

Access to Document

10.1109/IPDPS.2013.82

Cite this

@conference{c262c05874a04b53865f0a54d1389bb3,

title = "HQL: A scalable synchronization mechanism for GPUs",

abstract = "Modern GPUs rely on atomic operations to perform global communication. These atomic operations can be used to construct finer-grained locks to provide support for mutual exclusion. However, equipped with only these basic synchronization primitives to support mutual exclusion results in inefficient use of resources. In this paper, we propose a new hardware-based blocking synchronization mechanism which uses hierarchical queuing for scalability and efficiency. We evaluate our design using a set of GPU applications for stressing synchronization mechanisms. We perform detailed simulation utilizing the Multi2Sim heterogeneous simulation infrastructure. Our results indicate that we can reduce the number of instructions executed by a GPU application by as much as 84%, while improving execution performance by as much as 73%.",

keywords = "GPUs, Mutual-exclusion, synchronization",

author = "Ayse Yilmazer and David Kaeli",

year = "2013",

doi = "10.1109/IPDPS.2013.82",

language = "English",

pages = "475--486",

note = "27th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2013 ; Conference date: 20-05-2013 Through 24-05-2013",

}

TY - CONF

T1 - HQL

T2 - 27th IEEE International Parallel and Distributed Processing Symposium, IPDPS 2013

AU - Yilmazer, Ayse

AU - Kaeli, David

PY - 2013

Y1 - 2013

N2 - Modern GPUs rely on atomic operations to perform global communication. These atomic operations can be used to construct finer-grained locks to provide support for mutual exclusion. However, equipped with only these basic synchronization primitives to support mutual exclusion results in inefficient use of resources. In this paper, we propose a new hardware-based blocking synchronization mechanism which uses hierarchical queuing for scalability and efficiency. We evaluate our design using a set of GPU applications for stressing synchronization mechanisms. We perform detailed simulation utilizing the Multi2Sim heterogeneous simulation infrastructure. Our results indicate that we can reduce the number of instructions executed by a GPU application by as much as 84%, while improving execution performance by as much as 73%.

AB - Modern GPUs rely on atomic operations to perform global communication. These atomic operations can be used to construct finer-grained locks to provide support for mutual exclusion. However, equipped with only these basic synchronization primitives to support mutual exclusion results in inefficient use of resources. In this paper, we propose a new hardware-based blocking synchronization mechanism which uses hierarchical queuing for scalability and efficiency. We evaluate our design using a set of GPU applications for stressing synchronization mechanisms. We perform detailed simulation utilizing the Multi2Sim heterogeneous simulation infrastructure. Our results indicate that we can reduce the number of instructions executed by a GPU application by as much as 84%, while improving execution performance by as much as 73%.

KW - GPUs

KW - Mutual-exclusion

KW - synchronization

UR - http://www.scopus.com/inward/record.url?scp=84884871574&partnerID=8YFLogxK

U2 - 10.1109/IPDPS.2013.82

DO - 10.1109/IPDPS.2013.82

M3 - Paper

AN - SCOPUS:84884871574

SP - 475

EP - 486

Y2 - 20 May 2013 through 24 May 2013

ER -

HQL: A scalable synchronization mechanism for GPUs

Abstract

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this