A complexity-based classification for multiprocessor synchronization
Author(s)
Ellen, Faith; Gelashvili, Rati; Shavit, Nir; Zhu, Leqi
Download446_2019_361_ReferencePDF.pdf (404.5Kb)
Open Access Policy
Open Access Policy
Creative Commons Attribution-Noncommercial-Share Alike
Terms of use
Metadata
Show full item recordAbstract
Abstract
For many years, Herlihy’s elegant computability-based Consensus Hierarchy has been our best explanation of the relative power of various objects. Since real multiprocessors allow the different instructions they support to be applied to any memory location, it makes sense to consider combining the instructions supported by different objects, rather than considering collections of different objects. Surprisingly, this causes Herlihy’s computability-based hierarchy to collapse. In this paper, we suggest an alternative: a complexity-based classification of the relative power of sets of multiprocessor synchronization instructions, captured by the minimum number of memory locations of unbounded size that are needed to solve obstruction-free consensus when using different sets of instructions.
Date issued
2019-09-04Department
Massachusetts Institute of Technology. Computer Science and Artificial Intelligence LaboratoryPublisher
Springer Berlin Heidelberg