Chroma only supports RHEL-derived distributions if they are based on version 6 or later. The packages included with RHEL 5 are too old to run Chroma.
As the root user, run the following commands:
yum groupinstall "Development tools"
yum install python-matplotlib python-devel uuid-devel lapack-devel atlas-devel \
mercurial git subversion mesa-libGLU-devel freeglut-devel SDL-devel gtk2-devel \
libXpm-devel libXft-devel libXext-devel libXlibX11-devel expat-devel bzip2-devel \
libXt-devel
yum install epel-release
yum install virtualenv pygame numpy
CUDA requires the use of the official NVIDIA graphics driver, rather than the open source Nouveau driver that is included with RHEL.
Edit /etc/grub.conf and add rdblacklist=nouveau to the end of the kernel line:
kernel /vmlinuz-2.6.32-279.11.1.el6.x86_64 ... rdblacklist=nouveau
Create the file /etc/modprobe.d/blacklist-nouveau.conf with the line:
blacklist nouveau
The NVIDIA driver can be installed by going to the CUDA Downloads and downloading the runfile corresponding to your operating system.
Drop to a console terminal by pressing CTRL+ALT+F1 and then enter:
su -
init 3
Login again at the prompt and enter the following commands:
su -
cd
cp /home/[your username]/Downloads/cuda_9.0.176_384.81_linux.run .
chmod +x cuda_9.0.176_384.81_linux.run
./cuda_9.0.176_384.81_linux.run
During the installation you can pick all the default options.
Once installed, you can ensure the CUDA compiler and libraries are in your path by adding the following lines to your bash login script (usually $HOME/.bashrc):
export PATH=/usr/local/cuda/bin:$PATH
export LD_LIBRARY_PATH=/usr/local/cuda/lib64/:$LD_LIBRARY_PATH
Warning
Non-bash shells will need to be adjusted appropriately. If you are using a 32-bit distribution, then lib64/ should be changed to lib/.
The rest of the installation process is described in Common Installation Guide.