A major deficiency of static scheduling techniques is their inability to handle data-dependent looping and conditional branching which cause non-determinismic task execution times, in effect, dynamic alterations of the task graph.This paper
presents a heuristic, hybrid dynamic/static task-scheduling strategy which attempts to handlenon-deterministic task execution times. The goal of this strategy is to achieve an improvement in execution time over static scheduling without being overwhelmed by overhead costs. In the proposed approach, the tasks are initially scheduled statically using estimates of task completion times. During run time, tasks aremigrated following heuristic rules to take advantage of idle periods on processors. Results from simulationsindicate that this technique can produce a performance improvement over the static approach when operating in certain problem domains. The proposed strategy is compared with ``ideal'' static scheduling in order to characterize its efficiency.
Copyright 1996 by Ka Kay Ho and Bruno R. Preiss.
Full text.
Copyright 2002 by Bruno R. Preiss, P.Eng. All rights reserved.
Tue Jan 1 13:41:25 EST 2002
This book comprises the source code to the(pronounced ``why not?'') logic simlator. It is my first serious foray into the realm of ``literate programming.'' Literate programming was pioneered by Donald Knuth during the development of the TeX typesetting software.
The essential idea is that a literate program is the combination of documentation and source code in a form suitable for reading by human beings. The TeX source to this book and the C++ source code it documents are combined in common files. A literate programming tool which I wrote, called loma ,
weaves the the files into the TeX source that produced this book and into the C++ source programs which can be compiled to produce the executable program.The source to this book is not in the public domain. However, the C++ source code for the
simulator is available. The C++ source code is cross-referenced with this book, so that you can easily find the page in this which defines and describes the code. The
source code can be found on the world-wide web at http://www.pads.uwaterloo.ca/Bruno.Preiss/sources.html .
Copyright 1995 by Bruno R. Preiss.
Full text.
BibTeX entry.
Copyright 2002 by Bruno R. Preiss, P.Eng. All rights reserved.
Tue Jan 1 13:41:25 EST 2002
A complete set of solutions to the exercises in [44].
Copyright 1999 by John Wiley & Sons.
Full text.
BibTeX entry.
Copyright 2002 by Bruno R. Preiss, P.Eng. All rights reserved.
Tue Jan 1 13:41:25 EST 2002