Now, I realize there is already a "solution" to this problem, but that solution doesn't work for me.
My setup is very close to the one in this post : Can't debug CUDA: CUDA dynamic parallelism debugging is not supported in preemption mode . I'm also cognizant of this link : https://devtalk.nvidia.com/default/topic/536202/debugging-dynamic-parallelism-and-preemption-mode/
I'm on VS2012, Win 7 64bit, drivers are version 331.65, 2 GTX Titans (Device 0 driving display, Device 1 headless) and Nsight 3.2. I've followed the instructions in this post and turned off the forcing of SW preemption for Desktop & Headless GPUs. I've done a deviceQuery and both my Titans are showing up. Additionally, I've got my monitors plugged into the top Titan on the mobo, which I'm pretty sure is Device 0. Thus I've specified cudaSetDevice(1); in my code. I've disabled Windows Aero and...
...have no idea what else to do to prevent this from happening. I am toying with putting yet another GPU in my system, a GTX580 to drive the display, but I don't feel that should be necessary. I've tried changing the cudaSetDevice argument to 0 - same error, and 2 - can't find a CUDA device. Can anyone help me out here? I've got some beastly debugging to do.
Following the instructions listed in the link that I mentioned, I found that CUDA debugging would eventually work and work well. I don't really know what changed since I posted this question, but follow the previous solution's guidance and it should work.