mchakravarty/haskell-cuda repository overview

Docker image supporting Haskell GPU programming:

• accelerate-cuda-0.15.0.0 and
• accelerate-fft-0.15.0.0.

The current library versions are based on LTS Haskell 1.4 with GHC 7.8.4 and CUDA 6.5.

Available from the public Docker registry in two flavours:

• mchakravarty/haskell-cuda:lts1.4-cuda6.5: plain image executing ghci by default
• mchakravarty/haskell-cuda:lts1.4-cuda6.5-onbuild: build image to containerise Cabal executables

⁠Usage of the build image

Create a Dockerfile that just contains

FROM mchakravarty/haskell-cuda:lts1.4-cuda6.5-onbuild

Place this Dockerfile into the application directory containing a Cabal file or into a directory together with a Cabal source distribution tarball (as generated by cabal sdist).

Build your application container image, including your pre-compiled application, with

docker build -t app .

By default, the application container will execute cabal run, passing any run arguments on.

⁠Nota Bene

This image requires a host OS with NVIDIA's CUDA driver (version 6.5) to access any CUDA device.