too many kernels?

I had a perfectly fine working program and I added a new kernel to it, but as of now I don’t even call this new kernel. What is really baffling me is that now when I run the program, I get completely different results.

How does a kernel that doesn’t get called affect the execution of the program?

My hunch is that you changed something other than the kernel. Do you have source code control so you can roll back your changes? :)


My guess is that you are addressing out of array bounds in shared memory. There is a bug in 0.8 that shared memory is allocated for all kernels even though they cannot be called at the same time. Thus adding a kernel will change the smem location of all kernels following in the .cu file.