A distributed thread scheduler for dynamic multithreading on throughput processors

Ta Kan Yen; Hsien Kai Kuo; Bo-Cheng Lai

doi:10.1109/VLDI-DAT.2013.6533822

A distributed thread scheduler for dynamic multithreading on throughput processors

Ta Kan Yen, Hsien Kai Kuo, Bo-Cheng Lai

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

1 Scopus citations

Abstract

GPGPUs have emerged as one of the most widely used throughput processors. Deep multithreading and cache hierarchy are the two effective implementations to achieve high throughput computing in modern GPGPUs. However, these are two conflicting design options. Finding a proper design point between the two has become a significant performance factor to GPGPUs. This paper proposes a distributed thread scheduler for dynamic multithreading on GPGPUs. By demonstrating the trade-off issue between the multithreading and cache contention, the proposed scheduler dynamically adjusts the multithreading degree to achieve superior performance. With the proposed scheduler, the cache misses can be decreased by 20.6% and 37.9% on the L1 and L2 cache respectively. The overall performance can be enhanced by an average of 16.4%.

Original language	English
Title of host publication	2013 International Symposium on VLSI Design, Automation, and Test, VLSI-DAT 2013
DOIs	https://doi.org/10.1109/VLDI-DAT.2013.6533822
State	Published - 15 Aug 2013
Event	2013 International Symposium on VLSI Design, Automation, and Test, VLSI-DAT 2013 - Hsinchu, Taiwan Duration: 22 Apr 2013 → 24 Apr 2013

Publication series

Name	2013 International Symposium on VLSI Design, Automation, and Test, VLSI-DAT 2013

Conference

Conference	2013 International Symposium on VLSI Design, Automation, and Test, VLSI-DAT 2013
Country/Territory	Taiwan
City	Hsinchu
Period	22/04/13 → 24/04/13

Access to Document

10.1109/VLDI-DAT.2013.6533822

Cite this

@inproceedings{e0e096ff13a54e2bb8464802bcc81cb6,

title = "A distributed thread scheduler for dynamic multithreading on throughput processors",

abstract = "GPGPUs have emerged as one of the most widely used throughput processors. Deep multithreading and cache hierarchy are the two effective implementations to achieve high throughput computing in modern GPGPUs. However, these are two conflicting design options. Finding a proper design point between the two has become a significant performance factor to GPGPUs. This paper proposes a distributed thread scheduler for dynamic multithreading on GPGPUs. By demonstrating the trade-off issue between the multithreading and cache contention, the proposed scheduler dynamically adjusts the multithreading degree to achieve superior performance. With the proposed scheduler, the cache misses can be decreased by 20.6% and 37.9% on the L1 and L2 cache respectively. The overall performance can be enhanced by an average of 16.4%.",

author = "Yen, {Ta Kan} and Kuo, {Hsien Kai} and Bo-Cheng Lai",

year = "2013",

month = aug,

day = "15",

doi = "10.1109/VLDI-DAT.2013.6533822",

language = "English",

isbn = "9781467344357",

series = "2013 International Symposium on VLSI Design, Automation, and Test, VLSI-DAT 2013",

booktitle = "2013 International Symposium on VLSI Design, Automation, and Test, VLSI-DAT 2013",

note = "null ; Conference date: 22-04-2013 Through 24-04-2013",

}

Yen, TK, Kuo, HK & Lai, B-C 2013, A distributed thread scheduler for dynamic multithreading on throughput processors. in 2013 International Symposium on VLSI Design, Automation, and Test, VLSI-DAT 2013., 6533822, 2013 International Symposium on VLSI Design, Automation, and Test, VLSI-DAT 2013, 2013 International Symposium on VLSI Design, Automation, and Test, VLSI-DAT 2013, Hsinchu, Taiwan, 22/04/13. https://doi.org/10.1109/VLDI-DAT.2013.6533822

A distributed thread scheduler for dynamic multithreading on throughput processors. / Yen, Ta Kan; Kuo, Hsien Kai; Lai, Bo-Cheng.

2013 International Symposium on VLSI Design, Automation, and Test, VLSI-DAT 2013. 2013. 6533822 (2013 International Symposium on VLSI Design, Automation, and Test, VLSI-DAT 2013).

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

TY - GEN

T1 - A distributed thread scheduler for dynamic multithreading on throughput processors

AU - Yen, Ta Kan

AU - Kuo, Hsien Kai

AU - Lai, Bo-Cheng

PY - 2013/8/15

Y1 - 2013/8/15

N2 - GPGPUs have emerged as one of the most widely used throughput processors. Deep multithreading and cache hierarchy are the two effective implementations to achieve high throughput computing in modern GPGPUs. However, these are two conflicting design options. Finding a proper design point between the two has become a significant performance factor to GPGPUs. This paper proposes a distributed thread scheduler for dynamic multithreading on GPGPUs. By demonstrating the trade-off issue between the multithreading and cache contention, the proposed scheduler dynamically adjusts the multithreading degree to achieve superior performance. With the proposed scheduler, the cache misses can be decreased by 20.6% and 37.9% on the L1 and L2 cache respectively. The overall performance can be enhanced by an average of 16.4%.

AB - GPGPUs have emerged as one of the most widely used throughput processors. Deep multithreading and cache hierarchy are the two effective implementations to achieve high throughput computing in modern GPGPUs. However, these are two conflicting design options. Finding a proper design point between the two has become a significant performance factor to GPGPUs. This paper proposes a distributed thread scheduler for dynamic multithreading on GPGPUs. By demonstrating the trade-off issue between the multithreading and cache contention, the proposed scheduler dynamically adjusts the multithreading degree to achieve superior performance. With the proposed scheduler, the cache misses can be decreased by 20.6% and 37.9% on the L1 and L2 cache respectively. The overall performance can be enhanced by an average of 16.4%.

UR - http://www.scopus.com/inward/record.url?scp=84881347490&partnerID=8YFLogxK

U2 - 10.1109/VLDI-DAT.2013.6533822

DO - 10.1109/VLDI-DAT.2013.6533822

M3 - Conference contribution

AN - SCOPUS:84881347490

SN - 9781467344357

T3 - 2013 International Symposium on VLSI Design, Automation, and Test, VLSI-DAT 2013

BT - 2013 International Symposium on VLSI Design, Automation, and Test, VLSI-DAT 2013

Y2 - 22 April 2013 through 24 April 2013

ER -

A distributed thread scheduler for dynamic multithreading on throughput processors

Abstract

Publication series

Conference

Access to Document

Other files and links

Fingerprint

Cite this