Detected only 1 CUDA by using openCV

This question was found in here to the following link:
https://devtalk.nvidia.com/default/topic/740390/embedded-systems/tegra-jetson-gpu-accleration-in-opencv-/

i have followed the step to reinstall the opencv4tegra to the L4T and try to run the gpu::getCudaEnableDeviceCount() function, but it shows the return value as 1, is that mean i can use only 1 cuda core out of 192 cores in openCV?

Can any one share some idea or experience about this problem?

Thanks.

hello,
I tried opencv cuda sample in TADP 3.0 and it won’t work.
The apk crashes exactly in gpu::getCudaEnableDeviceCount(), may I know what procedure you have done to make it work?

The other thing is also regarding to TADP CUDA_Sample that if I built boxfilter manually, it won’t work and crashes in .cu init() function. However the prebuilt library cab run and can see the result.

Is it because the libcuda.so?

thank you for your help.

I’m sorry that I didn’t declared clearly. For my case, I’m using jetson with the default ubuntu instead of the android platform. So I didn’t know if android base k1 opencv library would have .the GPU class or not, but as I know the opencvaandroid version doesn’t include the GPU class. At least I didn’t found it when I develop the opencv app for tegra3.

So I think you may have to see if there is any library in your system that is about cuda or gpu first.
Thanks.

thank you for your reply and all the libraries and samples are in TADP 3.0 so that makes me wonder how come the prebuilt library behaves differently from the manually build one.

Let me start a thread for my issue for better tracking.

thank you Samuel