[go: up one dir, main page]

Menu

[r353]: / release-3 / INSTALL  Maximize  Restore  History

Download this file

43 lines (27 with data), 1.2 kB

To install cocolib, you need to first run the configure script:

> ./configure.sh

If that was successful, you can build the library and the examples
via

> make

The library requires some pre-requisites. On Ubuntu (or derivatives),
you can install all required packages available on apt via the script

> ./install_dependencies_ubuntu.sh

You will need the newest nVidia driver. To install that one via a
ppa repository, use

> ./install_nvidia_driver_ubuntu.sh

What you need to install manually is the nVidia CUDA toolkit.
You need version 4.2 or higher in order to have it work with
the newest Linux distributions. For earlier toolkit versions, you
need to downgrade the gcc compiler to 4.4 or an even more
previous version.

You can get the toolkit on www.nvidia.com.


If you managed to successfully build cocolib,
I suggest you change directory to "examples" and run
the test scripts for the library to make sure it was built
correctly.

> cd examples
> ./test_cocolib_quick.sh

This will test only the fast algorithms, and should complete in
1-2 minutes. To also test the more sophisticated algorithms, run

> ./test_cocolib.sh

This will take about 15 minutes.