Prepare to debug CUDA GPU code
You might need to add additional compiler command line options to enable GPU debugging.
For NVIDIA’s nvcc
compiler, kernels must be compiled with the -g -G
flags. This enables generation of information for debuggers in the
kernels, and also disables some optimizations that would hinder
debugging. To use memory debugging with CUDA, --cudart shared
must also be passed to nvcc
.
For other compilers, see NVIDIA GPU language support, and your vendor’s own documentation.
Note
OpenCL debugging of GPUs is not supported.