[Openmp-commits] [openmp] [OpenMP] Update out of date documentation (PR #142411)

[Johannes Doerfert via Openmp-commits]

================
@@ -92,104 +92,38 @@ For AMDGPU offload, please see :ref:`build_amdgpu_offload_capable_compiler`.
 
 Q: How to build an OpenMP Nvidia offload capable compiler?
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
-The Cuda SDK is required on the machine that will execute the openmp application.
-
-If your build machine is not the target machine or automatic detection of the
-available GPUs failed, you should also set:
-
-- ``LIBOMPTARGET_DEVICE_ARCHITECTURES='sm_<xy>;...'`` where ``<xy>`` is the numeric
-  compute capability of your GPU. For instance, set 
-  ``LIBOMPTARGET_DEVICE_ARCHITECTURES='sm_70;sm_80'`` to target the Nvidia Volta
-  and Ampere architectures. 
-
+The Cuda SDK is required on the machine that will build and execute the
+offloading application. Normally this is only required at runtime by dynamically
+opening the CUDA driver API. If this is disabled with
+``LIBOMPTARGET_DLOPEN_PLUGINS`` it will be directly linked at LLVM build time.
----------------
jdoerfert wrote:

Can you please elaborate more. "at runtime" and "*this* is disabled", are for example not clear. Maybe describe what is happening w/ and w/o the flag set.

https://github.com/llvm/llvm-project/pull/142411