Copperhead: Data Parallel Python
--------------------------------

Copperhead is a data parallel language embedded in Python,
along with a compiler which creates highly efficient parallel code.
Currently our compiler targets CUDA-enabled GPUs.

Copperhead is hosted at
    http://code.google.com/p/copperhead

Copperhead depends on:
  - CUDA.  We require version 4.1 or better.
  - boost::python.  Tested with version 1.48.  Important note: it is imperative
  that your boost::python library was built with the same compiler as your
  Python interpreter.
  - Codepy.  This will be installed automatically if you do not have it.
  - Thrust.  This comes along with CUDA and does not require special
  installation.
  
Copperhead runs on OS X and Linux.  It currently does not run on Windows,
although we plan on supporting Windows in the future.

To build Copperhead, you need:
  - g++ version 4.5, supported by nvcc.  CUDA 4.1 supports g++ 4.5, but not
  newer versions.
  - Scons.

For systems with a custom built boost::python, please create a file named
siteconf.py in the top directory, alongside setup.py.  In this file, describe
your boost installation.  For example:
#!/usr/bin/python
BOOST_INC_DIR = "/home/username/boost_1_48_0"
BOOST_LIB_DIR = "/home/username/boost_1_48_0/stage/lib"
BOOST_PYTHON_LIBNAME = "boost_python"

The BOOST_PYTHON_LIBNAME variable should hold the name of the library that
you would pass to the linker if you were linking. On Linux, for example,
if the filename of the library is libboost_python-gcc43.so, you would write
BOOST_PYTHON_LIBNAME = "boost_python-gcc43"

If siteconf.py does not exist, the build system will attempt to guess,
but may fail. 
    
More installation instructions and documentation are found on the website.