Using translation layers
The Nvidia CUDA platform is one of the good reasons for the popularity of the company's 3D cards in the professional segment. There have been various developments for running CUDA applications on other GPUs for a long time, but now Nvidia has decided to start counteracting this. The company has reportedly banned CUDA-based software from running on other hardware platforms using translation layers.
A new clause has appeared in the Nvidia license agreement, however, so far only on the website, but not in the software itself.
You may not reverse engineer, decompile, or disassemble any part of the output generated using elements of the SDK for the purpose of translating such output artifacts to a non-Nvidia platform.
The initiative appears to be aimed at combating projects like ZLUDA, which allow CUDA applications to run on Intel and AMD GPUs. In addition, which may be more important for Nvidia, such an initiative should interfere with Chinese developments in this direction. For example, earlier the Chinese Denglin Technology stated that it was already developing processors with a computing architecture compatible with programming models such as CUDA/OpenCL.