HyFi

Architecture-independent parallelism on networks of multiprocessors

David K Lowenthal, Ragavanis Subramanian

Research output: Contribution to journalArticle

Abstract

A network of parallel workstations promises cost-effective parallel computing. This article presents the HyFi (Hybrid Filaments) package, which can be used to create architecture-independent parallel programs that is, programs that are portable and efficient across different parallel machines. HyFi integrates Shared Filaments (SF), which provides parallelism on shared-memory multiprocessors, and Distributed Filaments (DF), which extracts parallelism from networks of uniprocessors, This enables parallelism on any architecture, including homogeneous networks of multiprocessors. HyFi uses fine-grain parallelism and implicit shared-variable communication to provide a uniform programming model. HyFi adopts the same basic execution model as SF and DF; this work discusses the modifications necessary to develop the hybrid system. In particular, HyFi modifies the signal-thread model as well as the software distributed shared memory of DF. It also unifies the SF and DF reduction operations as well as the dynamic load-balancing mechanism of fork-join filaments. Application programs written using the HyFi API can run unchanged on any architecture. Performance is encouraging on fork/join applications, where excellent speedup is achieved. Also, the fine-grain model of HyFi allows up to a 14.5% improvement due to overlap of communication and computation. Unfortunately, we find that iterative applications do not speed up well due to the inability of the Pentium Xeon architecture to efficiently support concurrent memory accesses.

Original languageEnglish (US)
Pages (from-to)272-282
Number of pages11
JournalInternational Journal of Computers and Applications
Volume25
Issue number4
StatePublished - 2003
Externally publishedYes

Fingerprint

Data storage equipment
Communication
Parallel processing systems
Dynamic loads
Hybrid systems
Application programming interfaces (API)
Application programs
Resource allocation
Costs

Keywords

  • Architecture independence
  • Distributed shared memory
  • Fine-grain parallelism

ASJC Scopus subject areas

  • Hardware and Architecture

Cite this

HyFi : Architecture-independent parallelism on networks of multiprocessors. / Lowenthal, David K; Subramanian, Ragavanis.

In: International Journal of Computers and Applications, Vol. 25, No. 4, 2003, p. 272-282.

Research output: Contribution to journalArticle

@article{150e6fcaaf544f40b84086f6a498bde3,
title = "HyFi: Architecture-independent parallelism on networks of multiprocessors",
abstract = "A network of parallel workstations promises cost-effective parallel computing. This article presents the HyFi (Hybrid Filaments) package, which can be used to create architecture-independent parallel programs that is, programs that are portable and efficient across different parallel machines. HyFi integrates Shared Filaments (SF), which provides parallelism on shared-memory multiprocessors, and Distributed Filaments (DF), which extracts parallelism from networks of uniprocessors, This enables parallelism on any architecture, including homogeneous networks of multiprocessors. HyFi uses fine-grain parallelism and implicit shared-variable communication to provide a uniform programming model. HyFi adopts the same basic execution model as SF and DF; this work discusses the modifications necessary to develop the hybrid system. In particular, HyFi modifies the signal-thread model as well as the software distributed shared memory of DF. It also unifies the SF and DF reduction operations as well as the dynamic load-balancing mechanism of fork-join filaments. Application programs written using the HyFi API can run unchanged on any architecture. Performance is encouraging on fork/join applications, where excellent speedup is achieved. Also, the fine-grain model of HyFi allows up to a 14.5{\%} improvement due to overlap of communication and computation. Unfortunately, we find that iterative applications do not speed up well due to the inability of the Pentium Xeon architecture to efficiently support concurrent memory accesses.",
keywords = "Architecture independence, Distributed shared memory, Fine-grain parallelism",
author = "Lowenthal, {David K} and Ragavanis Subramanian",
year = "2003",
language = "English (US)",
volume = "25",
pages = "272--282",
journal = "International Journal of Computers and Applications",
issn = "1206-212X",
publisher = "ACTA Press",
number = "4",

}

TY - JOUR

T1 - HyFi

T2 - Architecture-independent parallelism on networks of multiprocessors

AU - Lowenthal, David K

AU - Subramanian, Ragavanis

PY - 2003

Y1 - 2003

N2 - A network of parallel workstations promises cost-effective parallel computing. This article presents the HyFi (Hybrid Filaments) package, which can be used to create architecture-independent parallel programs that is, programs that are portable and efficient across different parallel machines. HyFi integrates Shared Filaments (SF), which provides parallelism on shared-memory multiprocessors, and Distributed Filaments (DF), which extracts parallelism from networks of uniprocessors, This enables parallelism on any architecture, including homogeneous networks of multiprocessors. HyFi uses fine-grain parallelism and implicit shared-variable communication to provide a uniform programming model. HyFi adopts the same basic execution model as SF and DF; this work discusses the modifications necessary to develop the hybrid system. In particular, HyFi modifies the signal-thread model as well as the software distributed shared memory of DF. It also unifies the SF and DF reduction operations as well as the dynamic load-balancing mechanism of fork-join filaments. Application programs written using the HyFi API can run unchanged on any architecture. Performance is encouraging on fork/join applications, where excellent speedup is achieved. Also, the fine-grain model of HyFi allows up to a 14.5% improvement due to overlap of communication and computation. Unfortunately, we find that iterative applications do not speed up well due to the inability of the Pentium Xeon architecture to efficiently support concurrent memory accesses.

AB - A network of parallel workstations promises cost-effective parallel computing. This article presents the HyFi (Hybrid Filaments) package, which can be used to create architecture-independent parallel programs that is, programs that are portable and efficient across different parallel machines. HyFi integrates Shared Filaments (SF), which provides parallelism on shared-memory multiprocessors, and Distributed Filaments (DF), which extracts parallelism from networks of uniprocessors, This enables parallelism on any architecture, including homogeneous networks of multiprocessors. HyFi uses fine-grain parallelism and implicit shared-variable communication to provide a uniform programming model. HyFi adopts the same basic execution model as SF and DF; this work discusses the modifications necessary to develop the hybrid system. In particular, HyFi modifies the signal-thread model as well as the software distributed shared memory of DF. It also unifies the SF and DF reduction operations as well as the dynamic load-balancing mechanism of fork-join filaments. Application programs written using the HyFi API can run unchanged on any architecture. Performance is encouraging on fork/join applications, where excellent speedup is achieved. Also, the fine-grain model of HyFi allows up to a 14.5% improvement due to overlap of communication and computation. Unfortunately, we find that iterative applications do not speed up well due to the inability of the Pentium Xeon architecture to efficiently support concurrent memory accesses.

KW - Architecture independence

KW - Distributed shared memory

KW - Fine-grain parallelism

UR - http://www.scopus.com/inward/record.url?scp=0042570472&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0042570472&partnerID=8YFLogxK

M3 - Article

VL - 25

SP - 272

EP - 282

JO - International Journal of Computers and Applications

JF - International Journal of Computers and Applications

SN - 1206-212X

IS - 4

ER -