CUB conda instructions

Summary: We have already uploaded a conda package of the cub 1.10.0 sources to our channel on anaconda. Here is the code we used.

Reviewed By: nikhilaravi

Differential Revision: D25395358

fbshipit-source-id: b58fd473fcafd425c98c9d7a7c32574f651383a0

packaging/cub_conda/README.md

@@ -0,0 +1,26 @@
+## For building conda package for NVIDIA CUB
+
+CUB is required for building PyTorch3D so it makes sense
+to provide a conda package to make its header files available.
+This directory is used to do that, it is independent of the rest
+of this repo.
+
+Make sure you are in a conda environment with
+anaconda-client and conda-build installed.
+
+From this directory, build the package with the following.
+```
+mkdir -p ./out
+conda build --no-anaconda-upload --output-folder ./out cub
+```
+
+You can then upload the package with the following.
+```
+retry () {
+    # run a command, and try again if it fails
+    $*  || (echo && sleep 8 && echo retrying && $*)
+}
+
+file=out/linux-64/nvidiacub-1.10.0-0.tar.bz2
+retry anaconda --verbose -t ${TOKEN} upload -u pytorch3d --force ${file} --no-progress
+```

packaging/cub_conda/cub/meta.yaml

@@ -0,0 +1,12 @@
+package:
+  name: nvidiacub
+  version: 1.10.0
+source:
+  url: https://github.com/NVIDIA/cub/archive/1.10.0.tar.gz
+  folder: source
+build:
+  script: mkdir $PREFIX/include && cp -r source/cub $PREFIX/include/cub
+
+about:
+  home: https://github.com/NVIDIA/cub
+  summary: CUB provides state-of-the-art, reusable software components for every layer of the CUDA programming model.