whilst debugging, i occasionally run into a cuda memory copy that generally stalls the program, with the following sequence of calls:
cuMemGetAttribute_v2()
cudbgGetAPIVersion()
cudbgApiDetach()
cudbgReportDriverInternalError()
it is not always the same memory copy, and mostly occur ‘deeper’ into the program - later on in the program
i am not entirely sure how to interpret this - i suppose it is generally unwanted to have a cudbgReportDriverInternalError(), but i am not sure how critical this is
i have by now also come across a cuda-gdb internal error, which i reported under the gdb forum
i am not sure whether the gdb internal error and driver internal error are related, and which causes which then