Implementing parallel and concurrent tree structures

Abstract

As one of the most important data structures used in algorithm design and programming, balanced search trees are widely used in real-world applications for organizing data. Answering the challenges thrown up by modern largevolume and ever-changing data, it is important to consider parallelism, concurrency, and persistence. This tutorial will introduce techniques for supporting functionalities on trees, including various parallel algorithms, concurrency, multiversioning, etc. In particular, this tutorial will focus on an algorithmic framework for parallel balanced binary trees, which works for multiple balancing schemes, including AVL trees, red-black trees, weight-based trees, and treaps. This framework allows for theoretically-efficient algorithms. The corresponding implementation is available as a library, which demonstrates good performance both sequentially and in parallel in various use scenarios. This tutorial will focus on the following topics: 1) the algorithms and techniques used in the PAM library; 2) the interface of the library and a hands-on introduction to the download/installation of the library; 3) examples of applying the library to various applications and 4) introduction about other useful techniques for parallel tree structures and performance comparisons with PAM. CCS Concepts • Theory of computation → Sorting and searching; Shared memory algorithms; • Computing methodologies → Shared memory algorithms.

Optimal Parallel Algorithms in the Binary-Forking Model

“…Early PRAM and BSP algorithms had explored parallel priority queues in a variety of approaches [10,32,40,41,72,73], and heavily rely on synchronization-based techniques such as pipelining. These algorithms do not have better bounds than recent batchdynamic search trees [18,20,[83][84][85] when mapping to the fork-join model. Other previous papers considered the concurrent, externalmemory, and other settings [6, 16, 30, 56, 62, 63, 74-77, 86, 94].…”