On seemingly random occassions a kernel under cuda-gdb gets stuck. It just never finishes.
I ctrl+c to see what happens. It’s always stuck on a line that does ++.
I’ve tried moving out a++ from b[a++] to a see that it’s actually the incrementation that fails, and it appears to be so.
Is this a common or otherwise known PEBKAC, or a bug in cuda-gdb?
I’ve had no luck using cuda-gdb whatsoever, it always freezes when using “next” or “step”. AFAIK it’s only officially supported for Red Hat on linux, but I hear the 2.3 SDK has cuda-gdb support for all major linux distros.
I bump this in the hope of getting any comment from nvidia :)
Should I expect the debugger not to work reasonably well under anything else than red hat?