How to design a parallel program
The design of algorithms that exploit parallelism is based on a series of operations, which must necessarily be carried out for the program to perform the job correctly without producing partial or erroneous results. The macro operations that must be carried out for a correct parallelization of an algorithm are:
- Task decomposition
- Task assignment
- Agglomeration
- Mapping
Task decomposition
In this first phase, the software program is split into tasks or a set of instructions that can then be executed on different processors to implement parallelism. To do this subdivision, there are two methods that are used:
- Domain decomposition: Here, the data of the problems is decomposed; the application is common to all the processors ...
Get Python Parallel Programming Cookbook now with the O’Reilly learning platform.
O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.
