The inherent complexity of transactional memory and what to do about it

Abstract

This talk overviews some of the lower bounds on the complexity of implementing software transactional memory, and explains their underlying assumptions. The talk will discuss how these lower bounds align with experimental results and design choices made in existing implementations to indicate that the transactional approach for concurrent programming must compromise either programming simplicity or scalability. The talk will describe several contemporary research avenues that address the challenge of concurrent programming. For example, optimizing coarse-grained techniques, and concurrent programming with mini-transactions - simple atomic operations on a small number of locations. This extended abstract outlines some of the ideas that will be presented in the talk