In the sixth part of his 12-part blog series about structured patterns as a basis for high-level parallel programming, Intel’s Michael McCool writesabout the partition pattern, which provides a simple way for the software developer to express both data locality and isolation. In this pattern, McCool explains, a collection of data is broken down into a set of non-overlapping regions each called a partition, and then a single task is assigned to operate on each partition. Since the partitions are non-overlapping, these tasks can freely operate in parallel on the data in each partition without interfering with one another.
To read the full blog post, click here.
http://software.intel.com/en-us/blogs/2009/07/14/parallel-pattern-6-partition/
