Techniques to improve performance in requester-wins hardware transactional memory

Armejach, Adrià; Titos-Gil, Ruben; Negi, Anurag; Unsal, Osman S.; Cristal, Adrián

doi:10.1145/2555289.2555299

Cited by 1 publication

(1 citation statement)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Several applied papers, e.g. [16,9,23,11] discuss the trade-offs involved in implementing transactional protocols in hardware, and the performance pathologies of such systems. Reference [23] explicitly discusses the possibility of adding delays to hardware transactions, and implements and compares heuristics for tuning these delays.…”

Section: Related Workmentioning

confidence: 99%

The Transactional Conflict Problem

Alistarh

Haider

Kübler

et al. 2018

Proceedings of the 30th on Symposium on Parallelism in Algorithms and Architectures

View full text Add to dashboard Cite

The transactional conflict problem arises in transactional systems whenever two or more concurrent transactions clash on a data item. While the standard solution to such conflicts is to immediately abort one of the transactions, some practical systems consider the alternative of delaying conflict resolution for a short interval, which may allow one of the transactions to commit. The challenge in the transactional conflict problem is to choose the optimal length of this delay interval so as to minimize the overall running time penalty for the conflicting transactions.In this paper, we propose a family of optimal online algorithms for the transactional conflict problem. Specifically, we consider variants of this problem which arise in different implementations of transactional systems, namely "requestor wins" and "requestor aborts" implementations: in the former, the recipient of a coherence request is aborted, whereas in the latter, it is the requestor which has to abort. Both strategies are implemented by real systems. We show that the requestor aborts case can be reduced to a classic instance of the ski rental problem, while the requestor wins case leads to a new version of this classical problem, for which we derive optimal deterministic and randomized algorithms. Moreover, we prove that, under a simplified adversarial model, our algorithms are constant-competitive with the offline optimum in terms of throughput. We validate our algorithmic results empirically through a hardware simulation of hardware transactional memory (HTM), showing that our algorithms can lead to non-trivial performance improvements for classic concurrent data structures. arXiv:1804.00947v1 [cs.DC] 3 Apr 2018• We show that, under an adversarial transaction scheduling model, our algorithms are globally constant competitive in terms of the sum of running times of the transactions, i.e. inverse of throughput. (Section 6) Further, we discuss strategies to add probabilistic progress guarantees to our algorithms (Section 7).• We validate our strategy on a hardware simulator implementation of HTM [18] and in synthetic tests, showing that it can yield throughput improvements of up to 4×, and that it is competitive with a finely-tuned manual approach (Section 8).The Transactional Conflict Problem. In more detail, the transactional conflict problem presents the following trade-off. (Please see Figure 1 for an

show abstract