I am facing a significant issue in my project. I have been trying to enable CUDA Language support in my project for cross-compilation (host Linux 64 bits target Orin Nano aarch), but after numerous unsuccessful attempts, I have reached a dead end.
The SDK I am using only supports CUDA 11.4 with library support , and not with CUDA language support.
Also the SDK I am using includes version 2.35 of GLIBC, while the CUDA version provided by the SDK is 11.4. Unfortunately, CUDA 11.4 support language is not compatible with GLIBC 2.35, as the latest GLIBC version supported by CUDA 11.4 support language is 2.27.
Since I am not permitted to modify any part of the SDK, I have had to abandon this approach. I would like to know if it is possible to utilize argmax and other acceleration features without enabling CUDA Language support? also if someone has an approach to enable CUDA support language in my project, it is welcome.
I am capable to use this:
find_package(CUDAToolkit REQUIRED)
but not this one:
project(MyProject LANGUAGES CXX CUDA)
Example issue got trying to enable CUDA support Language downloading CUDA 11.4 support language in a local path and use C/C++ compilers from SDK
