SLcuda is a NVIDIA CUDA runtime implementation for SLang, for general
purpose programming of NVIDIA GPUs. It is broken into several modules,
the main module (cuda-module.cu) is the only one necessary to manage
data on the GPU, kernels can be dynamically created with functions in
cuda.sl. Other modules in here implement certain "canned" functions,
such as reductions (cuda-reduce.cu, cured.sl) and image smoothing
(cuda-image.cu, cuimg.sl). Random number generation is built into the
main module.

See INSTALL for installation instructions and pre-requisites.