Efficient and Practical Constructions of LL/SC Variables

Over the past decade, LL/SC have emerged as the most suitable synchronization instructions for the design of lock-free algorithms. However, current architectures do not support these instructions; instead, they support either CAS or RLL/RSC (e.g. POWER4, MIPS, SPARC, IA-64). To bridge this gap, this paper presents two efficient wait-free algorithms for implementing 64-bit LL/SC objects from 64-bit CAS or RLL/RSC objects. Our first algorithm is practical: it has a small, constant time complexity (of 4 for LL and 5 for SC) and a space overhead of only 4 words per process. This algorithm uses unbounded sequence numbers. For theoretical interest, we also present a more complex bounded algorithm that still guarantees constant time complexity and O(1) space overhead per process. The LL/SC primitive is free of the well-known ABA problem that afflicts CAS. By efficiently implementing LL/SC words from CAS words, this work presents an efficient general solution to the ABA problem