INSTALL

1) Setup your CUDA environment. For better instructions, consult the "Getting Started" guide for your operating system at http://developer.nvidia.com/nvidia-gpu-computing-documentation
	a) Go to http://http://developer.nvidia.com/cuda-downloads
	b) Download and install the CUDA Toolkit (4.1 at the time of this writing)
	c) Download and install the CUDA Development Driver
	d) download and install the CUDA GPU Computing SDK

The remaining instructions are for OS X. Adjust the path of the "/Develoer/GPU Computing" directory to the location of your CUDA SDK folder. This is "~/NVIDIA_GPU_Computing_SDK" for Linux systems at the time of this reading. For Windows, consider using the Visual Studio solution found in the sources.

2) Check out the sources. As of March 2012, these are now available on github
    $ cd /Developer/GPU\ Computing/C/src

	Single Node Variant:
    $ git clone https://github.com/Corv/CUDA-GMM-MultiGPU.git
		
		or
	
	MPI Variant:
	$ git clone https://github.com/Corv/CUDA-GMM-MPI.git
	
	The MPI variant requires an MPI installation on the machine and adjustments to the Makefile to point to the MPI libraries. The current sources were configured for the 'Lincoln' cluster at NCSA in 2010.
	
3) Build the sources
	$ cd /Developer/GPU\ Computing/C/
	$ make
	
This will build the common libraries, all sample applications in the SDK, and the CUDA clustering application.

4) Run the application. It should be located at /Developer/GPU\ Computing/C/bin/darwin/release/gmm
Consult the program usage statement and the README for more information.