greg7mdp/parallel-hashmap

What is it?

Key features

• 01Header-only, with no build required beyond the include path.
• 02Drop-in replacements for std::unordered_map, std::unordered_set, std::map and std::set.
• 03Parallel hash maps split into submaps, with optional internal locking for concurrent access.
• 04Flat tables can be dumped to disk and restored as a single array for trivially copyable types.
• 05Supports heterogeneous lookup and C++11 through C++17 APIs such as try_emplace.

Use cases

• Replacing std::unordered_map in performance-sensitive C++ code.
• Storing very large numbers of entries where lower memory use matters.
• Building multithreaded services that share a hash table across threads with internal locking.
• Persisting large lookup tables quickly by dumping them to disk as a binary array.

Quick start

1. 1

   Copy the parallel_hashmap directory into your project and add it to the include path

2. 2

   Build and run the tests and examples with CMake

   cmake -DPHMAP_BUILD_TESTS=ON -DPHMAP_BUILD_EXAMPLES=ON -B build && cmake --build build && ctest --test-dir build

Things to know

• !The README recommends migrating to the gtl project, which requires C++20, and future feature work moves there.
• !Flat hash maps move values in memory, so pointers and iterators into them can be invalidated on mutation.
• !It is a C++11 header-only library with no packaged install; you must copy the directory manually.

Topics

FAQ

What is parallel-hashmap?

A header-only C++ library of fast, memory-friendly hash maps, sets, and B-tree containers that can replace the std unordered and ordered containers. Parallel Hashmap (phmap) is a C++ library of hash map and set implementations, plus B-tree based ordered containers that serve as alternatives to std::map and std::set. It is header-only: you copy the parallel_hashmap directory into a project and include it, with no build step. The hash tables are drop-in replacements for std::unordered_map and std::unordered_set, and the B-trees for std::map and std::set.

Is parallel-hashmap free to use? What license does it have?

parallel-hashmap is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.

Is parallel-hashmap still maintained?

Updated within 6 months. The last commit to parallel-hashmap was on Sep 10, 2026.

What can I use parallel-hashmap for?

Replacing std::unordered_map in performance-sensitive C++ code. Storing very large numbers of entries where lower memory use matters. Building multithreaded services that share a hash table across threads with internal locking. Persisting large lookup tables quickly by dumping them to disk as a binary array.

How do I install parallel-hashmap?

Copy the parallel_hashmap directory into your project and add it to the include path Build and run the tests and examples with CMake: cmake -DPHMAP_BUILD_TESTS=ON -DPHMAP_BUILD_EXAMPLES=ON -B build && cmake --build build && ctest --test-dir build

Where can I get parallel-hashmap?

The source code and installation instructions are on GitHub: https://github.com/greg7mdp/parallel-hashmap