1

I wan to debug with Nsight but the compiler do not stop in break points. I have enabled the option that generates debug information.

enter image description here

Here is the command line:

Driver API (NVCC Compilation Type is .cubin, .gpu, or .ptx)

set CUDAFE_FLAGS=--sdk_dir "C:\Program Files (x86)\Microsoft SDKs\Windows\v7.0A\" "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v5.5\bin\nvcc.exe" --use-local-env --cl-version 2010 -ccbin "c:\Program Files (x86)\Microsoft Visual Studio 10.0\VC\bin\x86_amd64" -G --keep-dir x64\Debug -maxrregcount=0 --machine 32 --compile -cudart static -o MP7.dir\Debug\%(Filename)%(Extension).obj "%(FullPath)"

Runtime API (NVCC Compilation Type is hybrid object or .c file)

set CUDAFE_FLAGS=--sdk_dir "C:\Program Files (x86)\Microsoft SDKs\Windows\v7.0A\" "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v5.5\bin\nvcc.exe" --use-local-env --cl-version 2010 -ccbin "c:\Program Files (x86)\Microsoft Visual Studio 10.0\VC\bin\x86_amd64" -G --keep-dir x64\Debug -maxrregcount=0 --machine 32 --compile -cudart static -g -Xcompiler "/EHsc /nologo /Zi " -o MP7.dir\Debug\%(Filename)%(Extension).obj "%(FullPath)"

When the Nsight debuger starts the break point changes and an exclamation mark appear with the following box info:

The break point will not be currently be hit No CUDA device code for source file c:\users\..\file.cu. Did you enable device information in your project?

Both .ilk and .pdb files are in the same folder with the .exe. Also, in the Nsight debugger as Working directory that folder that includes the aforementioned files was set.

enter image description here

Moreover, breaking points were set inside the kernel functions.

How can I find out what is wrong? I am in a deadlock.

Thank you!

PS: Win 7 64x, VS2010 64x solution, Nsight 3.2 64x, Cuda sdk 5.5 64x, Nsight is working fine in sdk samples, GeForce 9600 GT (compute capability 1.1)

Thoth
  • 993
  • 12
  • 36
  • Perhaps this discussion [Cuda-gdb not stopping at breakpoints inside kernel](http://stackoverflow.com/questions/21723664/cuda-gdb-not-stopping-at-breakpoints-inside-kernel/21791042#21791042) could be of interest for you. – Vitality Feb 16 '14 at 21:35
  • Thank you!. Also using `compute_20,sm_20` with cc. 1.1 corrupt my results. What should I do with that post? Is it considered as a duplicate? Should I erase it? – Thoth Feb 16 '14 at 22:40

1 Answers1

2

Your device is compute capability 1.1

But you are compiling for compute capability 2.0: "Code Generation compute_20,sm_20"

Therefore there is no device code corresponding to your compute capability 1.1 device, and so you cannot properly set or hit breakpoints.

Change code generation to compute_11,sm_11 to match your device.

Robert Crovella
  • 143,785
  • 11
  • 213
  • 257
  • I want to use atomicAdd in shared memory. Using the configuration,`compute_11,sm_11` returns `error : Ptx assembly aborted due to errors`. Shared memory `atomicAdds` are not working for `cc. 1.1`? I have to work in global memory? – Thoth Feb 17 '14 at 10:17
  • Perhaps you should read the programming guide section on [atomics](http://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#atomic-functions). Atomic add in shared memory requires cc 1.2 – Robert Crovella Feb 17 '14 at 13:47