Optimal Time-Space Tradeoff for Shared Memory Leader Election

Yehuda Afek, Gideon Stupp

Research output: Contribution to journalArticlepeer-review


Though it is common practice to treat synchronization primitives for multiprocessors as abstract data types, they are in reality machine instructions on registers. A crucial theoretical question with practical implications is the relationship between the size of the register and its computational power. We wish to study this question and choose as a first target the popular compare & swap operation (which is the basis for many modern multiprocessor architectures). Our main results are: 1. We show that leader election among n processes can be solved using a compare & swap register that can hold ≈ log n/log log n values and n read/write registers of size 2 log n bits. That is, n = (k -1)! where k is the number of values in the compare & swap register, so the compare & swap register has only O(log log n) bits. 2. We prove a tight tradeoff between the space and time necessary to solve leader election with a compare & swap register. Specifically, we show that any algorithm for leader election among n processes with a k value compare & swap register, where k ≥ log n/log log n, must have a run that accesses the compare & swap register Θ(logk n) times. We conjecture that for k < log n/log log n there is no solution. The results of this paper suggest that a complexity hierarchy for multiprocessor synchronization operations should be based on the space complexity of synchronization registers and their type.

Original languageEnglish
Pages (from-to)95-117
Number of pages23
JournalJournal of Algorithms
Issue number1
StatePublished - Oct 1997


Dive into the research topics of 'Optimal Time-Space Tradeoff for Shared Memory Leader Election'. Together they form a unique fingerprint.

Cite this