Two-handed emulation

Abstract

This paper partly addresses the question of whether, in principle, there is any point in adding richer hardware synchronization primitives when the existing set is "universal", and therefore sufficient to synchronize any data structure in a non-blocking manner. The context of this paper is the ongoing investigation of the utility of adding a DCAS instruction to modern processors to aid the design and performance of non-blocking algorithms. We add one more piece of evidence in support of this instruction.In particular, we demonstrate that DCAS is sufficient to enable a technique called "two-handed emulation" which yields efficient and understandable implementations of a class of data structures. We present a non-blocking implementation of a doubly-linked list to show the basic technique. We describe a non-blocking implementation of a dynamically resizable hash-table to illustrate how the technique is amenable to optimizations that improve performance and increase concurrency. -485-1/02/0007...$5.00.crease total system throughput.Nevertheless, non-blocking algorithms have not been widely deployed in practice outside of a select group of data structures (for example FIFO queues [18], deques [3], counters, stacks [20], and others). Two impediments block the widespread use of NBS: conceptual complexity and poor performance in the common case.

Efficient randomized DCAS

“…. , }, by writing ⊥ to them (lines [11][12][13][14][15]. The mechanism in lines 13-15 ensures that no array entry [ ] [ ] gets erased anymore, once the value of has changed, i.e., once the object got locked by some other process.…”

“…Massalin and Pu [25] employed DCAS to build a lock-free operating system kernel, as well as various lock-free data structures (such as stacks, lists, queues, etc.). Greenwald [15] provided MCAS implementations using DCAS. Other examples of algorithms using DCAS include priority queues [23], double ended queues [2,6,15], and general -read-modify-write primitives [9].…”

“…Greenwald [15] provided MCAS implementations using DCAS. Other examples of algorithms using DCAS include priority queues [23], double ended queues [2,6,15], and general -read-modify-write primitives [9]. Attiya and Hillel [8] presented a software transactional memory system using DCAS, and state that their algorithm shows that "DCAS is, in some sense, a 'silver bullet' for highly-concurrent implementations, allowing to achieve locality properties that are difficult, maybe even impossible, to obtain using only CAS. "…”