I have not touched CUDA for almost 2 years. This weekend, I installed the latest CUDA 4.2 and tried to compile some of my old programs. Interestingly, NVCC 4.2 still displays a heap of the advisory warning message “Cannot tell what pointer points to, assuming global memory space”, just like it did years ago. Is there any workaround to turn off those annoying warnings? I know they would disappear if I compile for Fermi only, but I do need to keep support for Tesla as well.