Visual Profiler 5.5 - No Timeline, No Resuts, No Errors

I am trying to use Visual Profiler 5.5 with nVidia 690, Win7 x64, i7, 8 gigs of ram.
I have left only one CPU thread that creates/destroy cuContexts in my app, I have checked “Start execution with profiling” and “Enable concurrent kernel” (since I am using streams). I also tried to do synch reads from the device.

I have setuped Visual Profiler with path to my .exe file. When I hit “Generate Timeline”, my app starts, it does what it does. I closed it - and nothing happens.

Furthermore - I tried every single possible test/profile/experiment from the Visual Profiler - they pass all (no errors, no warnings, green checks appear next to their title) - but nothing else.
Result fields are empty. No timeline. No nothing.

My app is loading the GPU code from .fatbin file. The GPU code is compiled with all optimizations, --source-in-ptx, fast_math, maxrregcount for 2 compute cap (2.0 and 3.0). It is quite big (in terms of line of code).

Any ideas what should I do ?

Okay, let me put it the other way around …

Can I get some dev support (free or paid) from nVidia ?

Furthermore, CUDA debugger seems to be kind-of-working for me (well, it is 10x slower, with some of the local vars and so on, but it is working …).

I will be quite happy if the Visual Profiler become kind-of-wokring too … Please …

Is the profiler working for other apps on your system?

When you say “I closed it” does that mean that you terminated your application? It is important that your app call cudaDeviceReset() before exiting so that all buffered profile data is flushed (see the Profiler User’s Guide for more information. It is available under the Help menu).

You can also try running nvprof on your application from the command-line to see what happens. Do something like:

$ nvprof