This is CUDA vector addition sample that can be used to learn about CUDA build process.
It was created after experiencing bug in glibc-2.17-323.el7_9 that lead to nvcc compilation failures (https://bugzilla.redhat.com/show_bug.cgi?id=1925204).
Good references about CUDA build process are
- https://docs.nvidia.com/cuda/cuda-compiler-driver-nvcc/index.html
- https://on-demand.gputechconf.com/gtc/2013/webinar/cuda-toolkit-as-build-tool.pdf
You can start the docker container with make run_nvidia.
Then, you can call make all to compile kernel.cu file.
All intermediate compilation files are preserved.
Or, you can run the compilation step by step by executing make targets step1 step2 ... and analyze the output of every step.
Inspect the Makefile for more details.