template<typename T = uintptr_t, template< class > class spinpolicy2 = spins_to_loop<125>::policy, template< class > class spinpolicy3 = spins_to_yield<250>::policy, template< class > class spinpolicy4 = spins_to_sleep::policy>
class quickcpplib::_xxx::configurable_spinlock::shared_spinlock< T, spinpolicy2, spinpolicy3, spinpolicy4 >
A non-FIFO policy configurable shared/exclusive spin lock meeting the SharedMutex concept.
- Template Parameters
-
| T | An unsigned integral type capable of atomic usage |
sizeof(shared_spinlock<T>) == sizeof(T). Suitable for usage in shared memory.
Implementing a fair shared spin lock with acceptable performance in just four bytes of storage is challenging, so this is a reader-biased shared lock which uses bit 0 as the exclusion bit, and the remainder of the bits to track how many readers are in the shared lock. Undefined behaviour will occur if the number of concurrent readers exceeds half the maximum value of a T.
Maximum performance on Intel for exclusive locks is the same as a spinlock<uintptr_t>. For shared locks it is roughly one third that of spinlock<uintptr_t> due to performing twice as many atomic read-modify-updates.
