reference manual. Its probably best/essential to take a careful look at
the Release Notes as well.
Software Development Kit
Install the software development kit similarly: "sh
NVIDIA_CUDA_sdk_2.0beta2_linux.run" which will install the example
projects into your home space by default to the directory
$HOME/NVIDIA_CUDA_SDK. If you then go to that directory and
execute "make" and the CUDA Toolkit is installed properly, all the
example projects should compile. You can then test out those examples
by running them in the NVIDIA_CUDA_SDK/bin/linux/release
executables directory.
Visual Profiler
The Visual Profiler is just a tarball at this stage. Unpack it somewhere
with
"tar xfz CudaVisualProfiler_linux_1.0_13June08.tar.gz -C somewhere".
Then add the CudaVisualProfiler/bin directory to your
LD_LIBRARY_PATH:
export LD_LIBRARY_PATH=(blah blah):/usr/local/cuda/lib:(somewhere)
/CudaVisualProfiler/bin
The profiler binary needs a couple of libraries in that directory.
Installation and configuration: The mex
environment
To compile a mex file for CUDA, you will need four files or scripts
available:
A Makefile.1.
A nvopts.sh configuration file.2.
A mex source file with the *.cu extension.3.
The nvmex executable.4.
You can find the Makefile, the nvopts.sh, and the nvmex files in the