My current configuration : Visual Studio 2010, CUDA 5.5, Nsight version 3.1, Graphics card driver version 327.23 (latest)
The problem is that breakpoints in __global__ kernels are not hit when starting "CUDA Debugging"
Anyone with the same problem? It was possible for me to have kernel breakpoints on some older version of Nsight & CUDA, but I wouldn't want to downgrade to previous versions. Thanks.