# auto_partitioner¶

[algorithms.auto_partitioner]

Specify that a parallel loop should optimize its range subdivision based on work-stealing events.

A loop template with an auto_partitioner attempts to minimize range splitting while providing ample opportunities for work-stealing.

The range subdivision is initially limited to S subranges, where S is proportional to the number of threads specified by the global_contol or task_arena. Each of these subranges is not divided further unless it is stolen by an idle thread. If stolen, it is further subdivided to create additional subranges. Thus a loop template with an auto_partitioner creates additional subranges only when necessary to balance a load.

Performs sufficient splitting to balance load, not necessarily splitting as finely as Range::is_divisible permits. When used with classes such as blocked_range, the selection of an appropriate grain size is less important, and often acceptable performance can be achieved with the default grain size of 1.

The auto_partitioner class satisfies the CopyConstructibe requirement from ISO C++ [utility.arg.requirements] section.

Tip

When using auto_partitioner and a blocked_range for a parallel loop, the body may receive a subrange larger than the grain size of the blocked_range. Therefore do not assume that the grain size is an upper bound on the size of a subrange. Use simple_partitioner if an upper bound is required.

// Defined in header <tbb/partitioner.h>

namespace tbb {

class auto_partitioner {
public:
auto_partitioner() = default;
~auto_partitioner() = default;
};

}