The Autotools project

About

This project was created as a simplistic template to whomever wants to create a fully GPL project that would compile under any machine with CUDA under the usual commands:

$ ./configure
$ make
$ sudo make install

Normally, everyone that installs the CUDA library will install the SDK examples and you can use their templates however, what if you don't know if the guy compiling your program has the examples? What if you started using cutilCheckError() and the user don't have it? Thinking on this and browsing a little the web I managed to create a fully GNU Autotools project to work as a starting point to any CUDA project, specially the one that doesn't related with the packeges found under the common/ directory of the SDK examples.

How to get it?

Simply checkout the svn code. The versioning used in the filename is pretty easy and it evolves as the CUDA evolves, the name of the project is always autotools-X.Y.Z where X.Y is the CUDA version (started with CUDA 3.1) and Z is my own incremental versioning.
So, just checkout the correct code (3.1 on upwards): For the latest development:

$ svn co http://svn.geekvault.org/cuda/trunk/autotools

For the 3.1 version:

$ svn co http://svn.geekvault.org/cuda/branch/autotools-3.1

And use it as a template, just changing and customizing it.

Why do I separated it like this?

Basically because the changes from CUDA 2.X to CUDA 3.0 and 3.1, on the 3.0 they started to drop out the debugging features, and on 3.1 they completely removed it, requesting a new version of the template, so for the first public release I'll be starting on the 3.1 version and I'll keep it going like this, always adding new features as they are found (and I have time to add them).