Cannot hit breakpoints inside kernel using Nsight on a Turing GPU

0
votes

My computer's setup is:

OS: Windows 10

IDE: Visual Studio 2019 (and 2015)

GPU: Quadro 4000 RTX

NVIDIA driver package: 441.22 Drivers for use with the CUDA Toolkit 10.2, including Nsight 2019.4

I opened a CUDA sample project called "matrixMul", and set breakpoints inside the kernel 

template <int BLOCK_SIZE> __global__ void MatrixMulCUDA(float *C, float *A,
                                                        float *B, int wA,
                                                        int wB)

However, if I "Start CUDA Debbugging (Next-Gen)", I get the error message "Could not initialize driver for the specified CUDA Debugger. Debugging has been automatically stopped."

enter image description here

enter image description here

If I "Start CUDA Debbugging (Legacy)", I get the error message "A CUDA context was created on a GPU that is not supported by the legacy CUDA debugger. Breakpoints will be disabled. See output window for details. ".

enter image description here

I am pretty confused at the moment. I never had problem like this before when using Legacy debugging with Nsight 6.5.

Please advise.

1
visual-studiocudansight
Unless you have the GPU in TCC mode, debugging isn't supported - talonmies
@talonmies Care to explain a little more? So should I set my GPU on TCC mode? Thanks. - Nick X Tsui
I thought that was pretty self-explanatory. If you don't have the card in TCC mode, you can't use next gen debugging, and your GPU doesn't support legacy debugging - talonmies
OK. Thanks a lot. I added another display card, and now it is working. - Nick X Tsui

1 Answers

2
votes

As pointed out in comments, if this card is not set to TCC mode, then it cannot be used for CUDA debugging in Windows using next generation debugging (which is all the Turing cards support).

My solution is that I added another NVIDIA card to my computer to dedicate on display, so my Quadro 4000 RTX can be focused on computation (TCC mode). It works perfect now.