Unfortunately I didn't have time to run any tests this weekend like I had hoped. Regardless here are the instructions for those who want to get started. Please note that I'm writing this assuming that you have cross-compiled for the CARMA kit before. If not I recommend this blog to get started: http://cudamusing.blogspot.com/
-Download and install "arm-linux-gnueabi-gfortran-4.5" using sudo apt-get install.
-Download and save armel versions of lapack and blas from launchpad.com (CARMA users: be sure to use libraries for the Natty distro). You will point to these in your make.inc file.
-CARMA USERS ONLY: Apparently SECO didn't add the cuda/src directory to their version of the CUDA toolkit so I just downloaded the contents of that folder from the normal NVIDIA version and saved it in my CUDA directory
-CARMA USERS ONLY: Open the Makefile.internal file and change the -Dmagma_devptr_t flag to equal "integer(kind=4)". In the same file on line 41 (NVOPTS +=...) add the following flags: "-arch sm_21 -target-cpu-arch ARM --compiler-bindir /usr/bin/arm-linux-gnueabi-gcc-4.5 -m32"
-CARMA USERS ONLY: Open the CMakeLists.txt file and change the -Dmagma_devptr_t flag to equal "integer(kind=4)".
And I think that's all that I changed... It was over a few days of trial and error so please let me know if there was anything that I missed for those that are trying to compile.
Attached is my make.inc file for everyone look at. Again, I haven't tested anything yet so there's no guarantee that everything will work, but it will compile.