CUDA Toolkit CUDA is NVIDIA's parallel computing architecture. It
enables dramatic increases in computing performance by harnessing the
power of the GPU.

NOTE:
- the binary file to download is 5.5 GB
- the temporary package file will need 7 GB
- the final package will be over 4 GB (7 GB when installed).
Be sure to have enough free space before building this.

By default, the available nsight components are not included in
the final package. They may be enabled by setting the NSIGHT
environment variable to yes when building the package e.g.
    NSIGHT=yes bash cudatoolkit_12.SlackBuild

This is version 12, which should be installable together with
other cudatoolkits. This one may be activated by sourcing
/etc/profile.d/cuda-12.9.sh

NOTE:
This SlackBuild has been constructed primarily to enable compilation
of Blender and its libraries from source code. Blender 5.1.0 requires
gcc & g++ of at least 14.2.0 and this toolkit will be configured to
use that (based on the required gcc14 dependency). If different
compilers are required for whatever reason, they can be specified in
environment variables CUDA_CC and CUDA_CXX when running this SlackBuild
e.g.
    CUDA_CC=/usr/local/bin/gcc-12.3.5 CUDA_CXX=/usr/local/bin/g++-gcc-12.3.5 bash cudatoolkit_12.SlackBuild
If these are not set and gcc14 is not installed, cuda compilers will
be set to the default installed gcc & g++.