cl_nv_compiler_options Extension Where is this documented?

I see this extension come up on printouts of device queries, but I cannot find what it does, or how to use it. I checked nVidia Programming Guide & Best Practices, but a search gets no hits. I checked Khronos OpenCL API Registry web page. Nothing there as well.

A web search only gets any hits in pages for CUDA. I guess that is where the options themselves are, but in any case it only mentions “cl_nv_compiler_options”, not how to use it.

I thought it might be a way to specify something like “-cl-nv-maxrregcount=32” inside a kernel. This might avoid getting an error on platforms like OSX when it is specified as a compiler option. Building kernel compiler options on the fly, based on the platform, is a pain. This is not going to do anybody much good unless someone can find how to use it.

There is a text file at download page of 3.0 sdk, it somehow describes extensions.

I would like to bring up this subject again… it seems there are additional -cl-nv-XXXXXX options, beyond what’s described in the CUDA 3.0 documentation. For example: The clcc project, discussed in this thread, uses -cl-nv-arch and -cl-nv-cstd.

Which other options are there, these days? Where is the document for CUDA 11.x, corresponding to that cl_nv extensions page for CUDA 3.0?