Adaptive locks: Combining transactions and locks for efficient concurrency

Abstract

Transactional memory is being advanced as an alternative to traditional lock-based synchronization for concurrent programming. Transactional memory simplifies the programming model and maximizes concurrency. At the same time, transactions can suffer from interference that causes them to often abort, from heavy overheads for memory accesses, and from expressiveness limitations (e.g., for I/O operations). In this paper we propose an adaptive locking technique that dynamically observes whether a critical section would be best executed transactionally or while holding a mutex lock.The critical new elements of our approach include the adaptivity logic and cost-benefit analysis, a low-overhead implementation of statistics collection and adaptive locking in a full C compiler, and an exposition of the effects on the programming model. In experiments with both micro-and macro-benchmarks we found adaptive locks to consistently match or outperform the better of the two component mechanisms (mutexes or transactions). Compared to either mechanism alone, adaptive locks often provide 3-to-10x speedups. Additionally, adaptive locks simplify the programming model by reducing the need for fine-grained locking: with adaptive locks, the programmer can specify coarse-grained locking annotations and often achieve fine-grained locking performance due to the transactional memory mechanisms.

Weak atomicity for the x86 memory consistency model

“…We have also discussed a practical application (and the primary motivator) for the STM design: software lock elision for x86 machine code. Other solutions for software lock elision are either limited to Java [23], C++ [20] or impose a locking construct with entirely new semantics [21] on programs. The only other software transactional memory designed for direct use with x86 machine code [16] does not support the x86 memory consistency model and therefore fails simple tests such as privatisation.…”

SAW: Java Synchronization Selection from Lock or Software Transactional Memory

“…• It does not switch synchronization mechanisms during the application's execution as Adaptive Locks [19] does. • It does not allow the programmer to specify different synchronization mechanisms for different critical sections in the same program.…”

“…Adaptive Locks [19] is a synchronization mechanism that combines locks and transactions in C. It dynamically switches the mechanism used to whichever is more appropriate, a lock and an STM. Its critical section is described by using an atomic block, which is extended syntax of C, with a special lock variable.…”

Towards an efficient approach for resource sharing in real-time multiprocessor systems

“…The same principle has been used recently by Di Natale et al in [20] where an optimization problem is formulated to decrease the time delay and the memory requirement for tasks executed in a single processor. Finally an adaptive mechanism has been proposed by Usui et al in [21] that combine both lock-based and lock-free approaches. During runtime, the mechanism decides which approach should be used (lock-free or lock-based) to increase the efficiency of the system, however, this work does not provide any real-time guarantees.…”