Message-pool-based parallel operating system for the Kyushu University Reconfigurable Parallel Processor - parallel creation of multiple threads

Kunihiko Tsunedomi, Akira Fukuda, Kazuaki Murakami, Shinji Tomita

Research output: Contribution to journalConference article

Abstract

The Kyushu University Reconfigurable Parallel Processor system under development is a MIMD-type multiprocessor which consists of 128 processing elements, interconnected by a full (128×128) crossbar network. Reconfigurable memory architecture employed by the system allows the system to be configured as either a shared-memory TCMP (tightly coupled multiprocessor), a message-passing LCMP (loosely coupled multiprocessor), or a hybrid of the two. A parallel operating system under development is for the shared-memory TCMP, and aims at extracting various kinds of parallelism of the operating system itself to provide high-performance. To exploit the parallelism, the operating system is constructed by using a message-pool mechanism. A typical example of the parallelism is the parallel creation of multiple threads. In this paper, we propose four schemes for the parallel creation; the simple parallel scheme, the parallel template scheme, the chunk scheme, and the combination scheme. Simulation results show that the chunk scheme is the most desirable among the schemes.

Original languageEnglish
Pages (from-to)423-432
Number of pages10
JournalJournal of Information Processing
Volume14
Issue number4
Publication statusPublished - Dec 1 1991
Externally publishedYes
Event3rd Computer System Symposium - Tokyo, Jpn
Duration: Mar 26 1991Mar 27 1991

Fingerprint

Data storage equipment
Memory architecture
Computer operating systems
Message passing
Computer systems
Processing

All Science Journal Classification (ASJC) codes

  • Computer Science(all)

Cite this

Message-pool-based parallel operating system for the Kyushu University Reconfigurable Parallel Processor - parallel creation of multiple threads. / Tsunedomi, Kunihiko; Fukuda, Akira; Murakami, Kazuaki; Tomita, Shinji.

In: Journal of Information Processing, Vol. 14, No. 4, 01.12.1991, p. 423-432.

Research output: Contribution to journalConference article

@article{f98a2f74aafd43b98be20bc3bc335f28,
title = "Message-pool-based parallel operating system for the Kyushu University Reconfigurable Parallel Processor - parallel creation of multiple threads",
abstract = "The Kyushu University Reconfigurable Parallel Processor system under development is a MIMD-type multiprocessor which consists of 128 processing elements, interconnected by a full (128×128) crossbar network. Reconfigurable memory architecture employed by the system allows the system to be configured as either a shared-memory TCMP (tightly coupled multiprocessor), a message-passing LCMP (loosely coupled multiprocessor), or a hybrid of the two. A parallel operating system under development is for the shared-memory TCMP, and aims at extracting various kinds of parallelism of the operating system itself to provide high-performance. To exploit the parallelism, the operating system is constructed by using a message-pool mechanism. A typical example of the parallelism is the parallel creation of multiple threads. In this paper, we propose four schemes for the parallel creation; the simple parallel scheme, the parallel template scheme, the chunk scheme, and the combination scheme. Simulation results show that the chunk scheme is the most desirable among the schemes.",
author = "Kunihiko Tsunedomi and Akira Fukuda and Kazuaki Murakami and Shinji Tomita",
year = "1991",
month = "12",
day = "1",
language = "English",
volume = "14",
pages = "423--432",
journal = "Journal of Information Processing",
issn = "0387-6101",
publisher = "Information Processing Society of Japan",
number = "4",

}

TY - JOUR

T1 - Message-pool-based parallel operating system for the Kyushu University Reconfigurable Parallel Processor - parallel creation of multiple threads

AU - Tsunedomi, Kunihiko

AU - Fukuda, Akira

AU - Murakami, Kazuaki

AU - Tomita, Shinji

PY - 1991/12/1

Y1 - 1991/12/1

N2 - The Kyushu University Reconfigurable Parallel Processor system under development is a MIMD-type multiprocessor which consists of 128 processing elements, interconnected by a full (128×128) crossbar network. Reconfigurable memory architecture employed by the system allows the system to be configured as either a shared-memory TCMP (tightly coupled multiprocessor), a message-passing LCMP (loosely coupled multiprocessor), or a hybrid of the two. A parallel operating system under development is for the shared-memory TCMP, and aims at extracting various kinds of parallelism of the operating system itself to provide high-performance. To exploit the parallelism, the operating system is constructed by using a message-pool mechanism. A typical example of the parallelism is the parallel creation of multiple threads. In this paper, we propose four schemes for the parallel creation; the simple parallel scheme, the parallel template scheme, the chunk scheme, and the combination scheme. Simulation results show that the chunk scheme is the most desirable among the schemes.

AB - The Kyushu University Reconfigurable Parallel Processor system under development is a MIMD-type multiprocessor which consists of 128 processing elements, interconnected by a full (128×128) crossbar network. Reconfigurable memory architecture employed by the system allows the system to be configured as either a shared-memory TCMP (tightly coupled multiprocessor), a message-passing LCMP (loosely coupled multiprocessor), or a hybrid of the two. A parallel operating system under development is for the shared-memory TCMP, and aims at extracting various kinds of parallelism of the operating system itself to provide high-performance. To exploit the parallelism, the operating system is constructed by using a message-pool mechanism. A typical example of the parallelism is the parallel creation of multiple threads. In this paper, we propose four schemes for the parallel creation; the simple parallel scheme, the parallel template scheme, the chunk scheme, and the combination scheme. Simulation results show that the chunk scheme is the most desirable among the schemes.

UR - http://www.scopus.com/inward/record.url?scp=0026309095&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0026309095&partnerID=8YFLogxK

M3 - Conference article

AN - SCOPUS:0026309095

VL - 14

SP - 423

EP - 432

JO - Journal of Information Processing

JF - Journal of Information Processing

SN - 0387-6101

IS - 4

ER -