System prerequisites to offload computations to the NVIDIA GPUs
To compile and link programs that contain code to be offloaded to the NVIDIA GPUs with IBM® XL Fortran for Linux, V16.1, such as using OpenMP directives or CUDA Fortran, you must ensure that your hardware, operating system, and software meet these requirements.
Hardware requirements
You can use any IBM Power Systems™ server that has one or more NVIDIA GPUs installed and is supported by your Linux operating system distribution and CUDA Toolkit. For example, you can use IBM POWER® System S822LC for high performance computing or IBM POWER System S824L. For a complete list of the IBM Power Systems servers, see Power® Servers.
Supported platforms
You must use a system that satisfies the installation requirements of the CUDA Toolkit. See the NVIDIA CUDA Toolkit website for more information.
Software requirements
- NVIDIA CUDA Toolkit 9.2, which you can download
from CUDA
Toolkit DownloadNote: To install the CUDA Toolkit, use the Package Manager installation. The Runfile installation is currently not supported on Power processors. For instructions about Package Manager installation, see the NVIDIA CUDA Installation Guide for Linux.