[PATCH] D148185: Add more efficient bitwise vector reductions on AArch64

[Dave Green via Phabricator via llvm-commits]

dmgreen added a comment.

Thanks. Can you rebase again after 41549b535097 <https://reviews.llvm.org/rG41549b535097db923386a64ed769bf450943e6a3>?

I think the remaining regressions are some of the illegal types (which will be different if the value was already in a vector), some cases with Not where the sign bits are not calculated from a constant buildvector properly, and one case where I think it hits the Depth limit. Some of those are a bit of a shame but understandable, and we can fix the buildvector if we have a useful testcase for it. I think we can get this in and address that after.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D148185/new/

https://reviews.llvm.org/D148185