I’m finding that the beta release of the cuda-gdb does not halt at any breakpoint set at a kernel function that is launched within a stream, but I do have nominal execution/breakpoint control if the host launches the same kernel functions synchronously. My platform is RHEL 5.2, Ge Force 8600 GTS, CUDA 2.1.
Does anyone know whether cuda-gdb does (or when if will) support debugging of asynchronous kernel functions?
Thank you in advance for your help…