Help. Visual Profiler doesn't create timeline

Accelerated Computing CUDA CUDA Programming and Performance
1

I installed Visual Profiler 5.5, but after that I’ve started to couldn’t create timeline. I was able to generate with Visual Profiler 5.0. Now, I can’t genarate with any version of visual profiler. I take that error:

Unable to read the entire session timeline. The displayed timeline may be empty or incomplete because the application aborted or failed to flush profile data before exiting. The application should call cudaDeviceReset() before exiting to ensure that all profile data is flushed.

Please help me

2

Have you tested it with any of the CUDA Samples?

3

I had the similar issue and the way i solved is by adding cudaDeviceReset(); API call somewhere in the application while exiting. The other reason could be if your application is not using GPU at all.

4

I was testing my app before. After installing cuda 5.5, I started to get error. Also I added cudaDeviceReset() function, but result is same

5

First of all, try to run Visual Profile with one of the CUDA examples, to check if that error is in your app or not. Sometimes, when we update the toolkit, some errors may appear in our app.

6

I tested example first, it works. And then I tried my app without any change, it works too. I didn’t understand what is the problem

7

When you say that the example and your app work, what do you mean? Do they both run without errors but both of them fail in Visual Profiler? Excuse me, but I don’t know yet if the problem is just your app or Visual Profiler with any app.

8

Hi All,

I have the same issue. I used nvprof.exe in cuda 5.0 with no problems. Now the nvprof giving me the following error:
Unable to read the entire session timeline. the displayed timeline may be empty or incomplete because the application aborted or failed to flush profile data before exiting. The application should call cudaDeviceReset() to ensure that all profile data is flushed.

Not sure why asking me to flush since I already have istat=cudaDeviceReset() at the end of my program.
also, this error generated for my fortran cuda program as well as nvidia cuda 5.5 release examples. It seems like the 5.5 nvidia profiler has issues need to be fixed.

9

Adding another thread for reference, it’s caused by “path” issue in some developers end.