Host Environment
- OS: Windows 10
- Compiler: Visual studio 19?
To Reproduce
Steps to reproduce the behavior:
install cuda (v11.0.221) with cudnn (v8.0.2) separately
./vcpkg install opencv[contrib,eigen,tiff,cuda,opengl,ipp,tbb]:x64-windows
Failure logs
Building package cudnn[core]:x64-windows...
-- Found CUDA V11.0.221
CMake Error at ports/cudnn/portfile.cmake:138 (message):
Please install manually cuDNN for your CUDA v11.0
Call Stack (most recent call first):
scripts/ports.cmake:142 (include)
Error: Building package cudnn:x64-windows failed with: BUILD_FAILED
Additional context
As stated above, I have CUDA version 11.0 and cuDNN version 8.0.2 installed as per the nvidia installation instructions. This includes a CUDA_PATH environment variable that points to the correct location. The vcpkg script however fails to find the cuDNN installation.
I already went into the cudnn portfile and it seems that all the path hints provided to the find_path and find_library calls are empty. As a workaround I moved lines 67 and 68 (that look for the cuda installation) before these calls and added ${CUDA_TOOLKIT_ROOT} as a hint to both. As a result, opencv is now happily building in the background.
Is there maybe something I missed on my part, or can you guys fix this?
Host Environment
To Reproduce
Steps to reproduce the behavior:
install cuda (v11.0.221) with cudnn (v8.0.2) separately
./vcpkg install opencv[contrib,eigen,tiff,cuda,opengl,ipp,tbb]:x64-windowsFailure logs
Building package cudnn[core]:x64-windows...
-- Found CUDA V11.0.221
CMake Error at ports/cudnn/portfile.cmake:138 (message):
Please install manually cuDNN for your CUDA v11.0
Call Stack (most recent call first):
scripts/ports.cmake:142 (include)
Error: Building package cudnn:x64-windows failed with: BUILD_FAILED
Additional context
As stated above, I have CUDA version 11.0 and cuDNN version 8.0.2 installed as per the nvidia installation instructions. This includes a CUDA_PATH environment variable that points to the correct location. The vcpkg script however fails to find the cuDNN installation.
I already went into the cudnn portfile and it seems that all the path hints provided to the find_path and find_library calls are empty. As a workaround I moved lines 67 and 68 (that look for the cuda installation) before these calls and added ${CUDA_TOOLKIT_ROOT} as a hint to both. As a result, opencv is now happily building in the background.
Is there maybe something I missed on my part, or can you guys fix this?