Running CUDA samples: cudaErrorNotSupported


I’ve been following one of Drive PX2 Getting Started guides. Among other things there is a topic about installing, building and running
stock CUDA samples.

No issues with installing and compiling these on PX2 target but the samples wont run complaining about CUDA not being supported. As far as I can tell they fail on cudaGraphicsGLRegisterBuffer() call. Any ideas please?

nvidia@nvidia:~/NVIDIA_CUDA-9.0_Samples/5_Simulations/fluidsGLES$ ./fluidsGLES
fluidsGLES Starting…

NOTE: The CUDA Samples are not meant for performance measurements. Results may vary when GPU Boost is enabled.

nvrm_gpu: Bug 200215060 workaround enabled.
GPU Device 0: “Graphics Device” with compute capability 6.1

CUDA device [Graphics Device] has 9 Multi-Processors
Using OpenGL ES APIVersion: OpenGL ES 3.2 NVIDIA 390.00
Vendor: NVIDIA Corporation
Renderer: NVIDIA Tegra X2 (nvgpu)/integrated
CUDA error at fluidsGLES.cpp:333 code=71(cudaErrorNotSupported) “cudaGraphicsGLRegisterBuffer(&cuda_vbo_resource, vbo, cudaGraphicsMapFlagsNone)”

Dear williamzsf7z,

Could you please try to re-run the application with ./fluidsGLES CUDA_VISIBLE_DEVICES=0? Thanks.

Hi Steve

Just tried that. No difference. Same output exactly as above.

Dear williamzsf7z,

Sorry for the confusion.

Could you please try to re-run the app with below setting? Thanks.

This did the trick. Thank you Steve!