I am trying to use nvvp to profile my code.
I am unable to run nvvp at all since it returns “Unable to profile application. The application being profiled returned a non-zero return code.” This happens with every cuda program I attempt to profile. It happens with the examples provided as part of CUDA SDK 6.5.
I have found one forum post relating to this problem (Out of memory during profiling - Nsight Eclipse Edition - NVIDIA Developer Forums), of which I have tried everything suggested to no end (as was the poster’s conclusion). The post didn’t indicate if this problem was resolved or not and if so how.
In short: nvvp can’t even profile the CUDA examples provided and there seems to be no valid resolution to this problem online.
Any suggestions?
P.S. I am able to use nsight within visual studio without any problems. I need to be able to perform similar profiling on other machines potentially without a development environment installed.