What high-level languages ​​can I use to speed up GPU computing?

Question

I think I'm interested in using today's GPU capacities (CUDA, etc.) to speed up the work of my scientific applications with a crunch. But I do not want to invest time in C / C ++ coding. I prefer Scala, but I will not mind using Erlang, Haskell, F #, C #, Java, Python or another language of this level. Is it possible?

Answer 1

ScalaCL is core to Scala. It works through a set of specialized collections that delay work on the GPU through OpenCL.

Then there is also the ScalaCL Plugin , part of ScalaCL. Which is a compiler plugin that will automatically rewrite part of your code to use OpenCL bindings for acceleration, without extra work!

Answer 2

And don't forget Matlab , Mathematica, and Fortran , all of which have CUDA support. Mathematica supports OpenCL too.

Answer 3

Java and Python have bindings for OpenCL.

You will need to write GPU code in CUDA or OpenCL. If you did not find a library that did basic parallelization of loops.

Answer 4

http://en.wikipedia.org/wiki/CUDA#Language_bindings lists bindings for Java (can be used with Scala, but they will need an adaptive level to adapt to them) .NET and Python.

See also https://github.com/ztellman/penumbra (Clojure) and http://blogs.msdn.com/b/satnam_singh/archive/2009/12/15/gpgpu-and-x64-multicore-programming- with-accelerator-from-f.aspx and http://tomasp.net/blog/accelerator-intro.aspx (F #)