Pure functional programming to the GPU [closed]

Question

I've been wanting to play around with functional code, and thought what fun it would be to do some interactive real-time ray-tracing of some randomly composed functions. Does anyone know of any compiler or converter that can take code from a functional language (or any language, with high order functions) and convert it into something that can be used in CUDA or OpenCL?

Answer 1

There are several embedded languages in Haskell that target CUDA and/or OpenCL.

The most high level is accelerate, a subset of Haskell for array processing, embedded in Haskell. It works with CUDA.

Conal Elliott has a shader language embedded in Haskell, for graphics programming

Answer 2

There are several embedded languages in Haskell that target CUDA and/or OpenCL.

The most high level is accelerate, a subset of Haskell for array processing, embedded in Haskell. It works with CUDA.

Conal Elliott has a shader language embedded in Haskell, for graphics programming

Answer 3

There are some attempts, such as Obsidian (Haskell) and SaC (Single Assignment C).

More about this and some other attempts are described in this paper: Functional Programming for High-Performance Computing on Heterogeneous Architectures