do find NVML. Requiring finding the include file and library also makes the error message at the end obsolete. Of course, it's up for discussion what the best error message is and I'm happy to change accordingly.
Finding CUDAToolkit allows me to configure without specifying any extra CMake arguments since the path is deduced from the nvcc location.
Also Variorum was linking against NVML_HEADER instead of NVML_LIBRARY resulting in link failures even when the library was found.
Type of change
[x] Bug fix (non-breaking change which fixes an issue)
[ ] New feature/architecture support (non-breaking change which adds functionality)
[ ] Breaking change (fix or feature that would cause existing functionality to not work as expected)
[ ] Documentation update
[ ] Build/CI update
How Has This Been Tested?
Configuring and building in a docker container based on top of nvidia/cuda:11.7.1-devel-ubuntu22.04 via
Description
Without this fix NVML can't be found via standard CMake variables. With this, I can do
do find
NVML
. Requiring finding the include file and library also makes the error message at the end obsolete. Of course, it's up for discussion what the best error message is and I'm happy to change accordingly.Finding
CUDAToolkit
allows me to configure without specifying any extraCMake
arguments since the path is deduced from thenvcc
location.Also
Variorum
was linking againstNVML_HEADER
instead ofNVML_LIBRARY
resulting in link failures even when the library was found.Type of change
How Has This Been Tested?
Configuring and building in a docker container based on top of
nvidia/cuda:11.7.1-devel-ubuntu22.04
viaTest before merge:
Checklist:
./scripts/check-code-format.sh
and confirm my code code follows the style guidelines of variorum-DENABLE_WARNINGS=ON
)