I’ve been struggling for the last two days with Nsight debugger. What I really want is:
- I have the Windows application A (c#-managed) that starts the Windows application B (C++ native, OpenGL, CUDA) using the Process.Start() method.
- When process B is launched along with passed parameters from the application A, I want from within the Visual Studio to attach to process B to be able to debug the CUDA code using the nsight debugger.
My working configuration is:
a) CUDA 8.0 integrated in the VS 2012
b) GeForce Game Ready driver 382.52
c) NVIDIA Nsight Visual Studio Edition, Version 5.2, Build 18.104.22.16821
d) Windows 10
To debug the application I start nsight monitor as administrator (very important).
What I could observe until now is that if I start a single CUDA application (e.g. volumeFiltering from the CUDA samples), then I can see it in the “available processes” in the Visual Studio and I can attach to it. But if I start a CUDA application such that application A which is the managed code application starts the application B which is a CUDA application, then I am able to see both applications in the “available processes” with following conditions:
- Application A is enabled and is of type CUDA
- Application B is disabled and has no type
Initially, the application A is not of type CUDA which is definitelly correct. It gets that type assigned as soon as the process B is launched.
For the given constellation, someone can give advice on what I can do to get the application B enabled in the “available processes” in Visual Studio in order to attach to it and to debug it?