operator new in CUDA is always a __host__ __device__ function

Question

I\'m trying to use tracy (https://github.com/wolfpld/tracy) with CUDA
With tracy, to measure memory usage you override operator new and operator delet