Predictive log-synchronization

Shalev, Ori; Shavit, Nir

doi:10.1145/1217935.1217965

Cited by 13 publications

(6 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It uses a single Fetch&Add object in addition to an LL/SC object, thus proving that the lower bound in [22] can be beaten, if we use just a single Fetch&Add object in addition to an LL/SC object. Sim exploits the well-known technique [16,26,27,30] of having a thread executing an operation helping other already announced operations. The high-level idea of the algorithm is as simple as follows.…”

Section: Introductionmentioning

confidence: 99%

A highly-efficient wait-free universal construction

Fatourou

Kallimanis

2011

Proceedings of the Twenty-Third Annual ACM Symposium on Parallelism in Algorithms and Architectures

103

113

View full text Add to dashboard Cite

We present a new simple wait-free universal construction, called Sim, that uses just a Fetch&Add and an LL/SC object and performs a constant number of shared memory accesses. We have implemented Sim in a real shared-memory machine. In theory terms, our practical version of Sim, called P-Sim, has worse complexity than its theoretical analog; in practice though, we experimentally show that P-Sim outperforms several state-of-the-art lock-based and lock-free techniques, and this given that it is wait-free, i.e., that it satisfies a stronger progress condition than all the algorithms it outperforms.We have used P-Sim to get highly-efficient wait-free implementations of stacks and queues. Our experiments show that our implementations outperform the currently stateof-the-art shared stack and queue implementations which ensure only weaker progress properties than wait-freedom.

show abstract

Section: Introductionmentioning

confidence: 99%

A highly-efficient wait-free universal construction

Fatourou

Kallimanis

2011

Proceedings of the Twenty-Third Annual ACM Symposium on Parallelism in Algorithms and Architectures

103

113

View full text Add to dashboard Cite

show abstract

“…In the full version of the paper [9] we show that our transformation preserves the external behavior (e.g., linearizability, serializability, and deadlock-freedom) of the original lock-based code; Moreover, our transformation preserves disjoint access parallelism [34], (the property that threads that access disjoint data objects do not contend on low level shared memory locations), as it refrains from introducing a shared global clock (as some STM systems do [45]) or other sources of contention.…”

Section: Introductionmentioning

confidence: 86%

“…The advantage of our approach is in its generic nature, which allows us to parallelize existing code without requiring experts to perform manual optimizations. Other generic approaches we are familiar with are domination locking [26] and STM [45], both of which perform worse than our transformed code in our experiments. Further discussion of related work appears in Section 4.…”

Section: Introductionmentioning

confidence: 92%

“…In contrast, the code generated by our transformation is scalable, and its throughput continues to grow with the number of threads. We further use the Synchrobench framework [27] to compare our synthesized code to data structures that were recently hand-crafted by experts in the field [20,11,1,22,15], as well as a state-of-the-art STM [45]. Our results show that the implementations we have generated perform comparably to custom-tailored solutions.…”

Section: Introductionmentioning

confidence: 95%

See 1 more Smart Citation

Towards Automatic Lock Removal for Scalable Synchronization

Arbel

Golan-Gueta²,

Hillel³

et al. 2015

Lecture Notes in Computer Science

View full text Add to dashboard Cite

We present a code transformation for concurrent data structures, which increases their scalability without sacrificing correctness. Our transformation takes lock-based code and replaces some of the locking steps therein with optimistic synchronization in order to reduce contention. The main idea is to have each operation perform an optimistic traversal of the data structure as long as no shared memory locations are updated, and then proceed with pessimistic code. The transformed code inherits essential properties of the original one, including linearizability, serializability, and deadlock freedom. Our work complements existing pessimistic transformations that make sequential code thread-safe by adding locks. In essence, we provide a way to optimize such transformations by reducing synchronization bottlenecks (for example, locking the root of a tree). The resulting code scales well and significantly outperforms pessimistic approaches. We further compare our synthesized code to state-of-the-art data structures implemented by experts. We find that its performance is comparable to that achieved by the custom-tailored implementations. Our work thus shows the promise that automated approaches bear for overcoming the difficulty involved in manually hand-crafting concurrent data structures.

show abstract

“…Sim is an efficient wait-free implementation of the well-known combining technique [29,34,37,52,54,56,60]. Most of the previous implementations of this technique, including the algorithm presented in [52] (which we will call OyamaAlg from now on) and flatcombining [34], employ locks and therefore they are blocking (i.e.…”

Section: Chapter 1 Introductionmentioning

confidence: 99%

Efficient synchronization techniques for shared memory systems

Kallimanis¹,

Καλλιμάνης²

View full text Add to dashboard Cite

First, I would like to sincerely thank my supervisor Panagiota Fatourou for motivating and encouraging me during the entire period that I was conducting my PhD. I would also like to thank the members of my advisory committee, Vassilios Dimakopoulos and Leonidas Palios for their support. Many thanks to Dimitris Nikolopoulos for arranging the provision of access to some of the multi-core machines of the Department of Computer Science at Virginia Tech where I ran some of the experiments of this dissertation, and to Michael Scott for providing me access to the Rochester's Niagara 2 machine. I would also like to thank the rest of the members of my examination committee, Kostas Magoutis, Evaggelia Pitoura, and Apostolos Zarras. Special thanks go to my friends Spiros Agathos, Eytychia Datsika, Vasilis Kagias, Kostas Lillis, Thanos Mpiliris, Odysseas Petrocheilos, and Kostas Ramantas for supporting and encouraging me.Finally, I would like to thank Empirikion Foundation for the moral and financial support.

show abstract

Predictive log-synchronization

Cited by 13 publications

References 14 publications

A highly-efficient wait-free universal construction

A highly-efficient wait-free universal construction

Towards Automatic Lock Removal for Scalable Synchronization

Efficient synchronization techniques for shared memory systems

Contact Info

Product

Resources

About