Visible to Intel only — GUID: GUID-A5AF1F2A-8753-495A-A42C-1CC0210C6C83
Package Contents
Parallelizing Simple Loops
Parallelizing Complex Loops
Parallelizing Data Flow and Dependence Graphs
Work Isolation
Exceptions and Cancellation
Containers
Mutual Exclusion
Timing
Memory Allocation
The Task Scheduler
Design Patterns
Migrating from Threading Building Blocks (TBB)
Constrained APIs
Invoke a Callable Object
Appendix A Costs of Time Slicing
Appendix B Mixing With Other Threading Packages
References
parallel_for_each Body semantics and requirements
parallel_sort ranges interface extension
TBB_malloc_replacement_log Function
Type-specified message keys for join_node
Scalable Memory Pools
Helper Functions for Expressing Graphs
concurrent_lru_cache
task_group extensions
The customizing mutex type for concurrent_hash_map
Visible to Intel only — GUID: GUID-A5AF1F2A-8753-495A-A42C-1CC0210C6C83
Always Use wait_for_all()
One of the most common mistakes made in flow graph programming is to forget to call wait_for_all. The function graph::wait_for_all blocks until all tasks spawned by the graph are complete. This is not only useful when you want to wait until the computation is done, but it is necessary to call wait_for_all before destroying the graph, or any of its nodes. For example, the following function will lead to a program failure:
void no_wait_for_all() {
graph g;
function_node< int, int > f( g, 1, []( int i ) -> int {
return spin_for(i);
} );
f.try_put(1);
// program will fail when f and g are destroyed at the
// end of the scope, since the body of f is not complete
}
In the function above, the graph g and its node f are destroyed at the end of the function’s scope. However, the task spawned to execute f’s body is still in flight. When the task completes, it will look for any successors connected to its node, but by then both the graph and the node have been deleted out from underneath it. Placing a g.wait_for_all() at the end of the function prevents the premature destruction of the graph and node.
If you use a flow graph and see mysterious behavior, check first to see that you have called wait_for_all.