Advanced Topics in Dataflow Computing and Multithreading

ISBN-13: 9780818665424 / Angielski / Twarda / 1995 / 464 str.

Guang R. Gao; Lubomir Bic; Jean-Luc Gaudiot

Opis

Examines recent advances in design, modeling, and implementation of dataflow and multithreaded computers. The text contains reports concerning many of the world's leading projects engaged in the continuing evolution and application of dataflow concepts. It covers the broad range of dataflow principles in program representation -- from language design to processor architecture -- and compiler optimization techniques.

The first section of the book delves into massively parallel distributed memory and multithreaded architecture design, synchronization and pipelined design, and superpipelined data-driven VLSI processors. The next section, on language and programming issues, discusses stream data types, the development of well-structured software, and coarse-grain dataflow programming.

Other parts of the text study parallelization of dataflow programs, an analytical model for the behavior of dataflow graphs, compare a centralized work distribution scheme with a distributed scheme, and present a comprehensive approach to understanding workload management schemes. Altogether, the book introduces the reader to dataflow concepts that show how functional programming ideas can be harnessed to exploit the power of parallel computing.

Spis treści

Foreword vii

Introduction ix

Processor Design

Design Principle of Massively Parallel Distributed–Memory Multiprocessor Architecture 1
M. Amamiya and T. Kawano

StarT the Next Generation: Integrating Global Caches and Dataflow Architecture 19
B.S. Ang, Arvind, and D. Chiou

Synchronization and Pipeline Design for a Multithreaded Massively Parallel Computer 55
S. Sakai

Superpipelined Dynamic Data–Driven VLSI Processors 75
H. Terada, M. Iwata, S. Miyata, and S. Komori

Language and Programming Issues

Stream Data Types for Signal Processing 87
J.B. Dennis

Multilateral Diagrammatical Specification Environment Based on Data–Driven Paradigm 103
M. Iwata and H. Terada

Coarse–Grain Dataflow Programming of Conventional Parallel Computers 113
R. Jagannathan

Distributed Data Structure in Thread–Based Programming for a Highly Parallel Dataflow Machine EM–4 131
M. Sato, Y. Kodama, S. Sakai, Y. Yamaguchi, and S. Sekiguchi

Programmability and Performance Issues of Multiprocessors on Hard Nonnumeric Problems 143
A. Sohn and J.–L. Gaudiot

Compiling

Exploiting Iteration–Level Parallelism in Dataflow Programs 167
L. Bic, J.M.A. Roy, and M. Nagel

Empirical Study of a Dataflow Language on the CM–5 187
D.E. Culler, S.C. Goldstein, K.E. Schauser, and T. von Eicken

Programming the ADAM Architecture with SISAL 211
S. Mitrovic

Can Dataflow Machines Be Programmed with an Imperative Language? 229
S.F. Wail and D. Abramson

Resource Management and Scheduling

The Token Flow Model 267
J. Buck and E.A. Lee

Distributed Task Management in SISAL 291
M. Haines and A.P.W. Bohm

Load Balancing and Resource Management in the ADAM Machine 307
O.C. Maquelin

Workload Management in Massively Parallel Computers: Some Dataflow Experiences 325
D.F. Snelling and J.R. Gurd

Studies on Optimal Task Granularity and Random Mapping 349
T. Sterling, J. Kuehn, M. Thistle, and T. Anastasio

The Effects of Resource Limitations on Program Parallelism 367
K.B. Theobald, G.R. Gao, and L.J. Hendren

Program Characteristics and Performance Studies

The Dataflow Parallelism of FFT 393
A.P.W. Bohm and R.E. Hiromoto

Locality in the Dataflow Paradigm 405
I. Gottlieb and L. Biran

Locality and Latency in Hybrid Dataflow 417
W.A. Najjar, W.M. Miller, and A.P.W. Bohm

Implementation of Manipulator Control Computation on Conventional and Dataflow Multiprocessor 435
S. Zeng and G.K. Egan

Biography 449

Blurb

Examines recent advances in design, modeling, and implementation of dataflow and multithreaded computers. The text contains reports concerning many of the world′s leading projects engaged in the continuing evolution and application of dataflow concepts. It covers the broad range of dataflow principles in program representation –– from language design to processor architecture –– and compiler optimization techniques.

The book includes papers on massively parallel distributed memory and multithreaded architecture design, synchronization and pipelined design, and superpipelined data–driven VLSI processors. Other sections discuss stream data types, the development of well–structured software, and parallelization of dataflow programs It also details and analytical model for the behavior of dataflow graphs, compare a centralized work distribution scheme with a distributed scheme, and present a comprehensive approach to understanding workload management schemes. Altogether, the text introduces the reader to dataflow concepts that show how functional programming ideas can be harnessed to exploit the power of parallel computing.