Aggressive Transactional Boosting

Traditional software transactional memory systems implemented synchronization and recovery by tracking pertransaction memory access. In recent work, we introduced transactional boosting, a methodology whereby performance is improved by forgoing read/write sets and relying on data structure commutativity and abstract locks for synchronization. In this paper, we advocate a new aggressive form of transactional boosting, which enables concurrent execution of non-commuting transactions. Abstract locks are speculatively handed off from one transaction to the next, and dependencies are created, enforcing certain commit orders. We outline a preliminary algorithm, and discuss novel techniques for performing recovering lazily, and detecting cyclic dependencies.