Merge pull request #1262 from jaimergp/cuda-migration-docs

Add docs for extended CUDA build matrices

Author: isuruf

src/maintainer/knowledge_base.rst

@@ -1284,6 +1284,21 @@ For now, you will have to add ``nvcuda.dll`` to the ``missing_dso_whitelist``::
       - "*/nvcuda.dll"   # [win]
 
 
+My feedstock is not building old CUDA versions anymore
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+With the `addition of CUDA 11.1 and 11.2 <https://github.com/conda-forge/conda-forge-pinning-feedstock/pull/1162>`_,
+the default build matrix for CUDA versions was trimmed down to versions 10.2, 11.0, 11.1, 11.2.
+
+If you really need it, you can re-add support for 9.2, 10.0 and 10.1. However, this is not recommended.
+Adding more CUDA versions to the build matrix will dramatically increase the number of jobs and will place a large
+burden on our CI resources. Only proceed if there's a known use case for the extra packages.
+
+1. Download this `migration file <https://github.com/conda-forge/conda-forge-pinning-feedstock/blob/master/recipe/migrations/cuda92_100_101.yaml>`_.
+2. In your feedstock fork, create a new branch and place the migration file under ``.ci_support/migrations``.
+3. Open a PR and re-render. CUDA 9.2, 10.0 and 10.1 will appear in the CI checks now. Merge when ready!
+
+
 Adding support for a new CUDA version
 -------------------------------------