[PATCH] D11091: Actually support volatile memcpys in NVPTX lowering

[Jingyue Wu]

jingyue accepted this revision.
jingyue added a comment.
This revision is now accepted and ready to land.

LGTM

More of a feature request, llvm.memcpy now takes alignments as arguments. We should emit larger loads/stores (e.g. st.u32 or even vector store) when alignments permit. Worth a TODO?


Repository:
  rL LLVM

http://reviews.llvm.org/D11091