Hello,
I have an application where a kernel is 500 steps, packed with 5 blocks and 10 threads. It turns out that when run the app in Nsight, no error is found. But when run the app normally drive the video problem, and the application terminates, but does not accuse any error exception.
The only thing I can think is that as the debugger Nsight, the speed is much smaller than in the normal way, and this could be giving problem in normal form of application. Is there any relationship? Anyone know give me a superficial explanation?