CMake: remove CMAKE_CUDA_ARCHITECTURES default (#553)

Allow CMake to automatically determine `CMAKE_CUDA_ARCHITECTURES` if it
is not set rather than defaulting to `70`

CHANGELOG.md

@@ -6,6 +6,12 @@
 
 ### New Features and Enhancements
 
+The default value of `CMAKE_CUDA_ARCHITECTURES` is no longer set to `70` and is
+now determined automatically by CMake. The previous default was only valid for
+Volta GPUs while the automatically selected value will vary across compilers and
+compiler versions. As such, users are encouraged to override this value with the
+architecture for their system.
+
 ### Bug Fixes
 
 Fixed the loading of ARKStep's default first order explicit method.

cmake/SundialsTPLOptions.cmake

@@ -46,10 +46,6 @@ sundials_option(ENABLE_PTHREAD BOOL "Enable Pthreads support" OFF)
 # -------------------------------------------------------------
 sundials_option(ENABLE_CUDA BOOL "Enable CUDA support" OFF)
 
-# CMake 3.18 adds this option.
-sundials_option(CMAKE_CUDA_ARCHITECTURES STRING "Target CUDA architecture" "70"
-                DEPENDS_ON ENABLE_CUDA)
-
 # -------------------------------------------------------------
 # Enable HIP support?
 # -------------------------------------------------------------

doc/shared/RecentChanges.rst

@@ -2,6 +2,12 @@
 
 **New Features and Enhancements**
 
+The default value of :cmakeop:`CMAKE_CUDA_ARCHITECTURES` is no longer set to
+``70`` and is now determined automatically by CMake. The previous default was
+only valid for Volta GPUs while the automatically selected value will vary
+across compilers and compiler versions. As such, users are encouraged to
+override this value with the architecture for their system.
+
 **Bug Fixes**
 
 Fixed the loading of ARKStep's default first order explicit method.

doc/shared/sundials/Install.rst

@@ -523,9 +523,19 @@ illustration only.
 
 .. cmakeoption:: CMAKE_CUDA_ARCHITECTURES
 
-   Specifies the CUDA architecture to compile for.
+   Specifies the CUDA architecture to compile for i.e., ``60`` for Pascal,
+   ``70`` Volta, ``80`` for Ampere, ``90`` for Hopper, etc. See the `GPU compute
+   capability tables <https://developer.nvidia.com/cuda-gpus>`_ on the NVIDIA
+   webpage and the `GPU Compilation <https://docs.nvidia.com/cuda/cuda-compiler-driver-nvcc/index.html#gpu-compilation>`_
+   section of the CUDA documentation for more information.
 
-   Default: ``sm_30``
+   Default: Determined automatically by CMake. Users are encouraged to override
+   this value with the architecture for their system as the default varies
+   across compilers and compiler versions.
+
+   .. versionchanged:: x.y.z
+
+      In prior versions ``CMAKE_CUDA_ARCHITECTURES`` defaulted to ``70``.
 
 .. cmakeoption:: EXAMPLES_ENABLE_C