The substantial effort of parallel programming is only justified if the resulting
codes are adequately efficient. In this sense, all types of performance tuning are
extremely important to parallel software development. With parallel programs,
performance improvements are much more difficult to achieve than with con
ventional (sequential) programs. One way to overcome this inherent difficulty
is to bring in graphical tools.
When trying to visualize parallel programs, one is faced with a vast amount
of relevant literature consisting of more than 100 articles in journals or confer-
ence proceedings and dozens of software systems.
This book pursues two major goals: first, to cover recent developments in
parallel program visualization techniques and tools, most of which have not yet
been dealt with by text books; second, to demonstrate the application of specific
visualization techniques and software tools to scientific parallel programs.
For this purpose, two prototypical problem areas have been chosen: (i) so-
lution of initial value problems of ordinary differential equations - a notoriously
difficult problem with respect to parallelization, and (ii) numerical integration -
a problem which is seemingly easy to parallelize. In these two fields the advan
tages of parallel program visualization are demonstrated. One representative
software system - ParaGraph - is used to show how visualization techniques
can contribute to experimental performance assessment and improvement.