One way of looking at parallelization is by examining the critical paths in the application. A critical path is the set of steps that determine the minimum time that the task can
complete in. A serial program might complete tasks A, B, C, and D. Not all of the tasks need to have dependencies. B might depend on the results of A, and D might depend on the results of B and C, but C might not depend on any previous results. This kind of data can be displayed on a graph such as the one in Figure 3.24.
It is relatively straightforward to identify the critical path in a process once the dependencies and durations have been identified. From this graph, it is apparent that task C could be performed in parallel with tasks A and B. Given timing data, it would be possible to estimate the expected performance of this parallelization strategy.
Copyright © 2018-2020 BrainKart.com; All Rights Reserved. Developed by Therithal info, Chennai.