CUDA: Initialize CMAKE_CUDA_ARCHITECTURES using $ENV{CUDAARCHS}

NVCC's default architecture may be newer than the one supported by the
machine's GPU.
In such cases it's useful to have an environment variable for initializing
CMAKE_CUDA_ARCHITECTURES to avoid specifying it for every invocation.

1 files changed, 13 insertions, 0 deletions

diff --git a/Help/envvar/CUDAARCHS.rst b/Help/envvar/CUDAARCHS.rst
new file mode 100644
index 0000000..82369cd
--- /dev/null
+++ b/Help/envvar/CUDAARCHS.rst
@@ -0,0 +1,13 @@
+CUDAARCHS
+---------
+
+.. versionadded:: 3.20
+
+.. include:: ENV_VAR.txt
+
+Value used to initialize :variable:`CMAKE_CUDA_ARCHITECTURES` on the first
+configuration if it's not already defined. Subsequent runs will use the value
+stored in the cache.
+
+This is a semicolon-separated list of architectures as described in
+:prop_tgt:`CUDA_ARCHITECTURES`.