Because it’s nicer than writing CUDA or OpenCL by hand!
Futhark is a small programming language designed to be compiled to efficient parallel code. It is a statically typed, data-parallel, and purely functional array language in the ML family, and comes with a heavily optimising ahead-of-time compiler that presently generates GPU code via CUDA and OpenCL, although the language itself is hardware-agnostic and can also run on multicore CPUs. As a simple example, this function computes the average of an array of 64-bit floating-point numbers:
let average (xs: f64) = reduce (+) 0.0 xs / r64 (length xs)
Futhark is not designed for graphics programming, but instead uses the compute power of the GPU to accelerate data-parallel array computations. The language supports regular *nested* data-parallelism, as well as a form of imperative-style in-place modification of arrays, while still preserving the purity of the language via the use of a uniqueness type system.
While the Futhark language and compiler is an ongoing research project, it is quite usable for real programming, and can compile nontrivial programs which then run on real GPUs at high speed.
Futhark is a simple language and is designed to be easy to learn, although it omits some common features in order to generate high-performance parallel code. Nevertheless, Futhark can already be used for nontrivial programs. We are actively looking for more potential applications as well as people who are interested in contributing to the language design.
Futhark is not intended to replace existing general-purpose languages. The intended use case is that Futhark is only used for relatively small but compute-intensive parts of an application. The Futhark compiler generates code that can be easily integrated with non-Futhark code. For example, you can compile a Futhark program to a Python module that internally uses PyOpenCL to execute code on the GPU, yet looks like any other Python module from the outside (more on this here). The Futhark compiler will also generate more conventional C code, which can be accessed from any language with a basic FFI (an example here).
For more information, you can look at code examples, details on performance, our devblog, or maybe the docs, which also contains a list of our publications. You can of course also visit our main repository on Github, or our repository of benchmarks.