This library should build without problems on Mac OS X (10.5+), Linux,

and Windows. If you encounter a build problem, please report it along

with a log of the build messages to abergeron@gmail.com.

Requirements:

- either an OpenCL runtime (with headers) or the CUDA toolkit

- CMake [ http://cmake.org ] (to build)

Run CMake on the CMakeList.txt file in src/ and build according to

your platform. Set CMAKE_INSTALL_PREFIX to your desired path if you

don't want the platform default.

('cmake -DCMAKE_INSTALL_PREFIX=<prefix>' for unix-like platforms)

Pay attention to the messages from CMake since the library might still

build if no backends are found, but it won't be very useful.

There are instruction for installation in the CMake file which make

'make install' work on unix-like platforms. I have no idea how to

install on Windows.

If you also want the python bindings, run 'python setup.py install'

after building and installing the library which will install pygpu.