DO NOT REVIEW; For reference only.
This patch contains Arm's downstream changes to LLVM for SVE/SVE2. This supports SVE/SVE2 CodeGen, scalable auto-vectorization and the ACLE (C/C++ intrinsics interface).
While the patches are still based on LLVM 7, it should be a good indication of the changes we've made to support scalable vectors. Note that these patches are meant for reference and are not intended to be committed. This patch may help clarify some of the design choices we've made when implementing scalable vectors for SVE, and it allows others to experiment with our scalable-vector implementation.
The patches apply cleanly to the LLVM 7 release branch.