I'm happy to announce that version 0.2 of Boost.Compute has been tagged and released!
Boost.Compute is a header-only C++ library for GPGPU and parallel-computing based on OpenCL. It is available on GitHub and instructions for getting started can be found in the documentation.
Since version 0.1 (released almost two months ago), there have been 81 new commits created by 4 different authors. A number of bugs have been fixed including bug #105 which caused crashes when using OpenCL v1.1 devices with code compiled against OpenCL v1.2 headers. Also, new algorithms including unique(), search() and find_end() have been added by Roshan Raghupathy as part of his ongoing work for the Boost.Compute GSoC project.
Many new examples have also been added. The mandelbrot example shows how to calculate the mandelbrot fractal and display it directly on the GPU using the OpenCL-OpenGL sharing extension. The new k_means example shows how to quickly calculate clusters for a set of 2D points on the GPU. Also, let me know (or submit an issue) if you have any ideas/requests for additional examples.
And, as always, I look forward to any and all feedback. You can submit an issue through the issue tracker or contact me directly (email can be found in my profile).
P.S. I'll be at the C++Now conference in Aspen all next week. Let me know if you'd like to meet up.