CUDA Sample Error: Operation Not Supported

Good Afternoon,

After a fresh install onto the PX2 with the 5.0.5.0 Linux SDK, there seems to be an issue when using CUDA. The simpleGLES sample compiles, but when running it gives the following error:

simpleGLES (VBO) starting...

nvrm_gpu: Bug 200215060 workaround enabled.
Using OpenGL ES APIVersion: OpenGL ES 3.2 NVIDIA 390.00
Vendor: NVIDIA Corporation
Renderer: NVIDIA Tegra X2 (nvgpu)/integrated
Line 451: CUDA Error: operation not supported

When running through cuda-gdb, it shows that the problematic function is cudaGraphicsGLRegisterBuffer. Is this function actually unsupported on the PX2? If so, what workarounds exist to achieve similar functionality?

Thanks.

OpenGL is not supported on Tegra + dGPU. Many samples have been changed to choose iGPU over dGPU. But it appears that simpleGLES is not one of them.

You can run simpleGLES on iGPU with this command:

CUDA_VISIBLE_DEVICES=1 ./simpleGLES

[Problem solved, it wasn’t related to this problem]

@kathy_nv I am getting the same bug, I am using a very simple use case(with int8 precision on d-GPU)as described here, https://docs.nvidia.com/deeplearning/sdk/tensorrt-developer-guide/index.html#tensorflow_python_workflow

nvrm_gpu: Bug 200215060 workaround enabled.

@dhingratul, i am also getting the same bug, can you tell me whether the bug is resolved for you, if yes please help me out