Visible to Intel only — GUID: GUID-FE991F96-A4BF-43AA-9866-88E057BA11A0
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
Parallel Reduction for rvalues
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
Waiting for Single Messages in Flow Graph
Visible to Intel only — GUID: GUID-FE991F96-A4BF-43AA-9866-88E057BA11A0
Upgrade/Downgrade
It is possible to upgrade a reader lock to a writer lock, by using the method upgrade_to_writer. Here is an example.
std::vector<string> MyVector;
typedef spin_rw_mutex MyVectorMutexType;
MyVectorMutexType MyVectorMutex;
void AddKeyIfMissing( const string& key ) {
// Obtain a reader lock on MyVectorMutex
MyVectorMutexType::scoped_lock lock(MyVectorMutex,/*is_writer=*/false);
size_t n = MyVector.size();
for( size_t i=0; i<n; ++i )
if( MyVector[i]==key ) return;
if( !lock.upgrade_to_writer() )
// Check if key was added while lock was temporarily released
for( int i=n; i<MyVector.size(); ++i )
if(MyVector[i]==key ) return;
vector.push_back(key);
}Note that the vector must sometimes be searched again. This is necessary because upgrade_to_writer might have to temporarily release the lock before it can upgrade. Otherwise, deadlock might ensue, as discussed in Lock Pathologies. Method upgrade_to_writer returns a bool that is true if it successfully upgraded the lock without releasing it, and false if the lock was released temporarily. Thus when upgrade_to_writer returns false, the code must rerun the search to check that the key was not inserted by another writer. The example presumes that keys are always added to the end of the vector, and that keys are never removed. Because of these assumptions, it does not have to re-search the entire vector, but only the elements beyond those originally searched. The key point to remember is that when upgrade_to_writer returns false, any assumptions established while holding a reader lock may have been invalidated, and must be rechecked.
For symmetry, there is a corresponding method downgrade_to_reader, though in practice there are few reasons to use it.