I noticed that the loop vectorizer is not able to vectorize overflow instrinsics as they are not considered trivially vectorizable intrinsics, and loop vectorize is also not able to handle extractvalue instructions. Is this a low-hanging fruit for optimization? I think it's straightforward to expand these intrinsics into primitive arithmetic instructions, but it feels like bad design to un-canonicalize intrinsics. Is there a better alternative? v/r, Josh -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20140423/78094174/attachment.html>