Install futhark with Homebrew

Project links

• Homepage
• Repository

Commands and aliases

• futhark

Project history

Futhark is developed at DIKU and presented by its maintainers as an ongoing research project that is also usable for real programming. Its official publications page traces the language and compiler research line from early work on fusion and size inference through papers on redomap, nested parallelism, incremental flattening, memory optimization, automatic differentiation, and size-dependent types.

The project's design goal is not to replace general-purpose languages. The homepage says Futhark is intended for relatively small compute-intensive parts of larger applications, with the compiler generating C or Python modules that can be integrated through ordinary foreign-function or Python module boundaries.

Adoption history

The examples page says much Futhark code is still in its own tests and benchmarks, but it also maintains a list of projects using Futhark, including visual simulations, games, ray tracers, a webcam filter stack, Neptune GPU code for Filecoin's Poseidon hash implementation, lattice Boltzmann work, and a compiler implemented partly in Futhark.

The docs page lists bridges for Haskell, J, Python, Rust, Standard ML, and OCaml, along with editor and syntax-highlighting support. That ecosystem shows adoption beyond compiler research: users package Futhark as a language toolchain plus integration generators and editor support.

How it is used

Users write Futhark array programs and compile them ahead of time, commonly targeting GPU backends or C/Python integration. The documentation points beginners to the online book, user guide, language reference, installation guide, generated prelude docs, examples, packages, and benchmarks.

The project presents Futhark as suitable for compute kernels embedded in larger systems, where developers want high-level functional array code but still need generated code that links into C, Python, or other host languages.

Why package nerds care

Futhark is significant as a package because it is not just a compiler binary; it is an experimental programming-language distribution with a standard library, generated documentation, package index, language server and editor ecosystem, benchmark suite, and code-generation targets. Package maintainers have to think about GPU backends, C toolchains, generated docs, and language integration at the same time.

Timeline

• 2013: Official publications list early fusion work that feeds into the Futhark research line.
• 2014: Papers on size inference and bounds checking appear in the Futhark publication list.
• 2016: Papers document redomap and APL-on-GPU work in Futhark.
• 2017: PLDI paper presents Futhark as purely functional GPU programming with nested parallelism and in-place array updates.
• 2019-2026: Official publications continue across compiler implementation, memory optimization, automatic differentiation, type systems, and applications.

Related projects

• The official docs list bridges and tools including futhask, futhark-pycffi, genfut, futhark-bindgen, cargo-futhark, futhark-server-sml, smlfut, futhark-mode, futhark-vim, and futhark-vscode.
• The homepage also points to the futhark-benchmarks repository and to the Futhark package collection.

Sources

• https://futhark-lang.org/
• https://futhark-lang.org/docs.html
• https://futhark-lang.org/examples.html
• https://futhark-lang.org/publications.html
• https://github.com/diku-dk/futhark