Hi.
I am developing the cuda program in visual studio 2012 and cuda toolkit 6.0 installed environment.
Then, once i set the break point inside kernel function to debug by nsight visual studio edition(nsight->start cuda debugging), i can’t break there and never step into kernel function.
Strange point is that debug works well in sample project provided by toolkit.
Would you help me?
Probably the kernel is not getting executed properly. Add proper cuda error checking to your code, and run your code with cuda-memcheck.
I put the error checking function cudaGetLasstError.
It returns cudaSuccess after kernel and i could make sure that kernel is running by cuPrintf inside kernel function.
as i described, cuda sample has no problem, so to find out the difference between sample project and my project, compared two project’s settings.
There was only one difference, that is code generation setting, mine was compute_10, sm_10, but sample’s one was compute_10, sm_10;compute_20, sm_20;compute_50, sm_50, so i imitated sample’s setting.
but at this time, fatal error occured “ptx assembly aborted due to errors”.
Please help me.
Sorry, Its my fault.
My project was lay on the non uni-code path.