Lowering conflicts of high contention software transactional memory

Abstract

Two concurrent transactions are said to conflict based on linearizability semantics if they access the same shared data and at least one of them modifies that data. In many applications enforcing the strict linearizability semantics over the entire read-set of a transaction can lead to a large number of unnecessary aborts. Concurrent threads executing transactions with high contentions can result a high transaction abort ratio which degrade the overall performance. In this paper, we investigate the use of Causal Consistency with weaker semantics that provides a good trade-off between strong semantics and good practical performance even for long transactions. Experimental results show that the implementation is simple, and substantially reduce the number of Conflicts. © 2008 IEEE.