CUDA Nsight - find out what CPU is doing between kernel/runtime api calls

Question

I'm wondering if there's any way to discover in Nsight what CPU function calls are made between two runtime API/driver or kernel calls (an image's worth a thousand words):

Is it possible?

Answer 1

Using the NVIDA Tools Extension API you can get a variety of information about CPU activity (in addition to GPU activity) in the Nsight VSE profiler display.

There are a variety of examples given in the Nsight VSE 3.0 documentation.

The same API works with the standalone visual profiler -- nvvp.