A super-quick non-locking shareable mutex with ability to promote a lock from shared mode to exclusive mode and back.
Written in C++ with <atomic> library including memory ordering tweaking to achieve what is missing in both std and boost shared_mutex / shared_lock.
You can create shared_lock or exclusive_lock and then upgrade / downgrade it with upgrade_to_exclusive() and downgrade_to_shared() methods.
Basically, it implements multiple readers / multiple writers concept where any of the threads can be promoted to writer (and go back to reader) at any time.
Also, there are three waiting strategies: burn - cycles the cpu, wait - sleeps for some given time, yield - switches to another thread.
See the tests for examples.
gx::shared_mutex m;
// ...
// thread1
{
gx::shared_lock l(m);
// do shareable work
l.upgrade_to_exclusive();
// do exclusive work
l.downgrade_to_shared();
// do shareable work
}
// ...
// thread2
{
gx::exclusive_lock l(m);
// do exclusive work
l.downgrade_to_shared();
// do shareable work
l.upgrade_to_exclusive();
// do exclusive work
}
// ...
// thread3
{
gx::shareable_lock l(m, gx::shared);
// do shareable work
l.upgrade_to_exclusive();
// do exclusive work
}