cudaError 209

Accelerated Computing CUDA CUDA Programming and Performance
, ,
1

image
image2416×730 210 KB

image
image1612×1098 118 KB

I cuda11.4 on Ubuntu 18.04 system, launch kernel returns error 209, where should I check it, some people say compute_sm do not match, but it seems to have no effect, anyone know how to deal with it? Looking forward to your reply!
Forgot to mention that the GPU is 960m.

2

Please don’t post pictures of text on these forums. Post the actual text, using the formatting tools that are at the top of the edit pane.

error 209 appears to be “no kernel image is available for execution on the device”

This indicates a mismatch between the arch specification that you compiled your code with, and the actual arch specification of the GPU you are running on.

Unfortunately I’m not able to tell which GPU you are running on from the nvidia-smi output.

A GTX 960m GPU is or should be a cc5.0 GPU. So -arch=compute_50; -code=sm_50 should be correct. I’m not a CMake expert, so its possible you have done something incorrectly syntax-wise in CMake. CMake is not a NVIDIA tool.

My suggestions to start would be:

  1. Run the deviceQuery sample code on your GPU. This will output its exact compute capability.
  2. Compile your code directly via nvcc (not via CMake) to ensure that you can compile correctly and run correctly with the arch specification that matches what you observed via deviceQuery
  3. Then go sort out CMake. Usually when people are having trouble with CMake, I think its a good idea to specify CMake verbose output, and make sure the actual compile commands that it issues match what you expect.
3

thank you very much for your reply and suggestions.I will try to do it.