Futhark is a functional data parallel array programming language originally developed at UCPH Department of Computer Science (DIKU) as part of the HIPERFIT project.^[2] It focuses on enabling data parallel programs written in a functional style to be executed with high performance on massively parallel hardware, in particular on graphics processing units (GPUs). Futhark is strongly inspired by NESL, and its implementation uses a variant of the flattening transformation, but imposes constraints on how parallelism can be expressed in order to enable more aggressive compiler optimisations. In particular, irregular nested data parallelism is not supported.^[3]

Overview

Futhark is a language in the ML family, with an indentation-insensitive syntax derived from OCaml, Standard ML, and Haskell. The type system is based on Hindley-Milner with a variety of extensions, such as uniqueness types and size-dependent types. Futhark is not intended as a general-purpose programming language for writing full applications, but is instead focused on writing computational "kernels" (not necessarily the same as a GPU kernel) which are then invoked from applications written in conventional languages.^[4]

Examples

def dotprod xs ys = f64.sum (map2 (*) xs ys))

def dotprod [n] (xs: [n]f64) (ys: [n]f64) : f64 = f64.sum (map2 (*) xs ys))

def matmul [n][m][p] (A: [n][m]f64) (B: [m][p]f64) : [n][p]f64 =
  map (\A_row ->
         map (\B_col -> dotprod A_row B_col)
             (transpose B))
      A

References