Hi, some questions on installation.
from the instruction it says CUDA is needed for GDRCopy.
CUDA=<cuda-install-top-dir> ./build-rpm-packages.sh
I'd like to know if only the library needs CUDA and we can install driver and kernel without CUDA?
Here is our use case:
On cluster/machine level, we only install Nvidia Driver and run the GDR installation, CUDA will not present
$ sudo yum groupinstall 'Development Tools'
$ sudo yum install dkms rpm-build make check check-devel subunit subunit-devel
$ cd packages
$ CUDA=<cuda-install-top-dir> ./build-rpm-packages.sh
$ sudo rpm -Uvh gdrcopy-kmod-<version>dkms.noarch.<platform>.rpm
$ sudo rpm -Uvh gdrcopy-<version>.<arch>.<platform>.rpm
$ sudo rpm -Uvh gdrcopy-devel-<version>.noarch.<platform>.rpm
On docker, we install CUDA, cudnn, NCCL, and run the same GDR installation again.
$ sudo yum groupinstall 'Development Tools'
$ sudo yum install dkms rpm-build make check check-devel subunit subunit-devel
$ cd packages
$ CUDA=<cuda-install-top-dir> ./build-rpm-packages.sh
$ sudo rpm -Uvh gdrcopy-kmod-<version>dkms.noarch.<platform>.rpm
$ sudo rpm -Uvh gdrcopy-<version>.<arch>.<platform>.rpm
$ sudo rpm -Uvh gdrcopy-devel-<version>.noarch.<platform>.rpm
Will above steps work? Thanks!
Hi, some questions on installation.
from the instruction it says CUDA is needed for GDRCopy.
CUDA=<cuda-install-top-dir> ./build-rpm-packages.shI'd like to know if only the library needs CUDA and we can install driver and kernel without CUDA?
Here is our use case:
On cluster/machine level, we only install Nvidia Driver and run the GDR installation, CUDA will not present
On docker, we install CUDA, cudnn, NCCL, and run the same GDR installation again.
Will above steps work? Thanks!