CUDA: Add abstraction for cuda runtime selection

Fixes #17559
Replace our hard-coded default of cudart=static with a first-class abstraction to select the runtime library from an enumeration of logical names.

1 files changed, 24 insertions, 0 deletions

diff --git a/Help/variable/CMAKE_CUDA_RUNTIME_LIBRARY.rst b/Help/variable/CMAKE_CUDA_RUNTIME_LIBRARY.rst
new file mode 100644
index 0000000..ea1c1b8
--- /dev/null
+++ b/Help/variable/CMAKE_CUDA_RUNTIME_LIBRARY.rst
@@ -0,0 +1,24 @@
+CMAKE_CUDA_RUNTIME_LIBRARY
+--------------------------
+
+Select the CUDA runtime library for use by compilers targeting the MSVC ABI.
+This variable is used to initialize the :prop_tgt:`CUDA_RUNTIME_LIBRARY`
+property on all targets as they are created.
+
+The allowed case insensitive values are:
+
+.. include:: ../prop_tgt/CUDA_RUNTIME_LIBRARY-VALUES.txt
+
+Contents of ``CMAKE_CUDA_RUNTIME_LIBRARY`` may use
+:manual:`generator expressions <cmake-generator-expressions(7)>`.
+
+If this variable is not set then the :prop_tgt:`CUDA_RUNTIME_LIBRARY` target
+property will not be set automatically.  If that property is not set then
+CMake uses the default value ``Static`` to select the CUDA runtime library.
+
+.. note::
+
+  This property has effect only when the ``CUDA`` language is enabled. To
+  control the CUDA runtime linking when only using the CUDA SDK with the
+  ``C`` or ``C++`` language we recommend using the :module:`FindCUDAToolkit`
+  module.