I am new to CUDA development and have recently switched to using Nvidia's Nsight Eclipse IDE (version 7.0 on Ubuntu 14.04).
I have a CUDA program that compiles and works without issue when building using cmake.
However, when I compile the program within Nsight, the resulting binary runs very slowly: >17000ms / frame (Nsight) vs <20ms / frame (cmake).
What could be causing this slowdown and how do I go about resolving this problem?