ParlayLib - A Toolkit for Programming Parallel Algorithms on Shared-Memory Multicore Machines
ParlayLib is a C++ library for developing efficient parallel algorithms and software on shared-memory multicore machines. It provides additional tools and primitives that go beyond what is available in the C++ standard library, and simplifies the task of programming provably efficient and scalable parallel algorithms. It consists of a sequence data type (analogous to std::vector), many parallel routines and algorithms, a work-stealing scheduler to support nested parallelism, and a scalable memory allocator. It has been developed over a period of seven years and used in a variety of software including the PBBS benchmark suite, the Ligra, Julienne, and Aspen graph processing frameworks, the Graph Based Benchmark Suite, and the PAM library for parallel balanced binary search trees, and an implementation of the TPC-H benchmark suite.
Parlay is designed to be reasonably portable by being built upon mostly standards-compliant modern C++. It builds on GCC and Clang on Linux, GCC and Apple Clang on OSX, and Microsoft Visual C++ (MSVC) and MinGW on Windows. It is also tested on GCC and Clang via Windows Subsystem for Linux (WSL) and Cygwin. Support beyond x86-64 has not yet been explored. We would warmly welcome contributions that seek to achieve this.
Documentation
You can find Parlay's full reference documentation here
Examples
Parlay comes with over 50 example applications that you can learn from. You can find them in the examples directory.
Developer documentation
If you are interested in contributing to Parlay, the following pages describe useful information about our testing and benchmarking setups. If you just want to use Parlay in your own projects, these links are not relevant to you.