Your current filters are…
by Stephen Blair-Chappell
Not every attempt at making a program parallel is successful. In this session we present five case studies where parallelism did not achieve the expected results. Where programs actually failed, we examine the main causes of failure and suggest ways of overcoming them. Where programs were successfully parallelised but did not perform as well as expected, we look at the underlying reasons and show how to tune for better performance. Where the return on investment was considered to be too low, we analyse the reasons along with suggested estimates on programming effort for different kinds of parallel models. The examples here are all based on real projects. Finally we look at ways of spotting likely failures early in the software development cycle in order to avoid costly mistakes and wasted development effort creeping into a project.
This session assumes attendees have some experience in C programming.