Did Cuda8.0 Still Support NVCUVID?

Hi, Friends.

I’m using OpenCV 3.2 cuda video decode module, and rebuild opencv with CMake. But when I try to run the opencv cuda::video_writer sample, the program crash and throw_no_cuda.

Then I try to google this error, it says might be cause by without NVCUVID during rebuild. I find my CMakeLog says I successful configure with nvcuvid, and also have no bug during VS2015 build.

It seems CUDA8.0 no longer support NVCUVID, and only support NVENC. Is that true? What is the solution? My video card is GTX1070, so can I use CUDA7.5 instead?


According to the CUDA 8 release notes, the nvcuvid library is still available:


However, on newer hardware, the old NVCUVENC (cuda video encode) system is replaced by NVENC (hw video encode) and the newer NVDEC (hw video decode) should be used in place of NVCUVID (cuda video decode)

NVDEC and NVENC are separate systems and APIs, not part of CUDA.