Harlan is a new research programming language focused around taking advantage of modern GPUs. The Harlan language syntax is derived from Scheme while the language itself currently compiles to OpenCL.
Harlan is largely the work out of a researcher at Indiana University for tapping the potential offered by modern graphics processors. Harlan tries to make it easier and more productive to tap GPUs while not having as much of a learning hurdle for the developer as explicitly writing code for OpenCL or CUDA. Harlan itself compiles to OpenCL so that it can then be used by Intel's OpenCL SDK, NVIDIA's CUDA Toolkit, or AMD's APP SDK.
More details on the Harlan GPU computing initiative can be found via its GitHub project page