On the power of hardware transactional memory to simplify memory management

Abstract

Dynamic memory management is a significant source of complexity in the design and implementation of practical concurrent data structures. We study how hardware transactional memory (HTM) can be used to simplify and streamline memory reclamation for such data structures. We propose and evaluate several new HTMbased algorithms for the "Dynamic Collect" problem that lies at the heart of many modern memory management algorithms. We demonstrate that HTM enables simpler and faster solutions, with better memory reclamation properties, than prior approaches. Despite recent theoretical arguments that HTM provides no worst-case advantages, our results support the claim that HTM can provide significantly better common-case performance, as well as reduced conceptual complexity.

ThreadScan

“…These techniques originated in the realm of garbage collection. Algorithms such as [11,17,44], smart pointers [39,41] and more recent improvements using hardware transactions [12], assign shared counters to objects, and use them to count the number of references to an object at any given time. Reference counting schemes are automatic: counters and counter modifications can be added to code at compile time.…”

“…Recent work [12,2] has explored the potential of Hardware Transactional Memory (HTM) to simplify memory reclamation. Dragojevic et al [12] showed that HTM can be used to speed up and significantly simplify prior reclamation schemes, while Alistarh et al [2] devised a technique which streamlines the tracking of node references, by ensuring that each data structure operation appears as having executed inside a single hardware transaction.…”

“…Dragojevic et al [12] showed that HTM can be used to speed up and significantly simplify prior reclamation schemes, while Alistarh et al [2] devised a technique which streamlines the tracking of node references, by ensuring that each data structure operation appears as having executed inside a single hardware transaction. While these techniques are promising, their performance fundamentally relies on HTM, whose availability is currently limited.…”

Asynchronized Concurrency

“…Memory reclamation is of key importance in CSDSs [6,14,16,28,45]. Various techniques have been proposed, such as quiescent state [13,23,24], epochs [20], reference counters [14,21,53], pointers [6,28,45], and, recently, hardware transactional memory [1,16].…”

“…Various techniques have been proposed, such as quiescent state [13,23,24], epochs [20], reference counters [14,21,53], pointers [6,28,45], and, recently, hardware transactional memory [1,16]. ASCYLIB uses an epoch-based allocator that reduces the performance overheads.…”