CMP0104

New in version 3.18.

Initialize CMAKE_CUDA_ARCHITECTURES when CMAKE_CUDA_COMPILER_ID is NVIDIA. Raise an error if CUDA_ARCHITECTURES is empty.

CMAKE_CUDA_ARCHITECTURES introduced in CMake 3.18 is used to initialize CUDA_ARCHITECTURES, which passes correct code generation flags to the CUDA compiler.

Previous to this users had to manually specify the code generation flags. This policy is for backwards compatibility with manually specifying code generation flags.

The OLD behavior for this policy is to not initialize CMAKE_CUDA_ARCHITECTURES when CMAKE_CUDA_COMPILER_ID is NVIDIA. Empty CUDA_ARCHITECTURES is allowed.

The NEW behavior of this policy is to initialize CMAKE_CUDA_ARCHITECTURES when CMAKE_CUDA_COMPILER_ID is NVIDIA and raise an error if CUDA_ARCHITECTURES is empty during generation.

If CUDA_ARCHITECTURES is set to a false value no architectures flags are passed to the compiler. This is intended to support packagers and the rare cases where full control over the passed flags is required.

This policy was introduced in CMake version 3.18. CMake version 3.20.6 warns when the policy is not set and uses OLD behavior. Use the cmake_policy() command to set it to OLD or NEW explicitly.

Note

The OLD behavior of a policy is deprecated by definition and may be removed in a future version of CMake.

Examples

set_property(TARGET tgt PROPERTY CUDA_ARCHITECTURES 35 50 72)

Generates code for real and virtual architectures 30, 50 and 72.

set_property(TARGET tgt PROPERTY CUDA_ARCHITECTURES 70-real 72-virtual)

Generates code for real architecture 70 and virtual architecture 72.

set_property(TARGET tgt PROPERTY CUDA_ARCHITECTURES OFF)

CMake will not pass any architecture flags to the compiler.