[llvm] [LoopVectorize] Generate wide active lane masks (PR #147535)

[Kerry McLaughlin via llvm-commits]

================
@@ -0,0 +1,129 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt -S --passes="default<O3>" -scalable-vectorization=off -prefer-predicate-over-epilogue=predicate-dont-vectorize -enable-wide-lane-mask \
----------------
kmclaughlin-arm wrote:

I've changed the tests to only use `-passes=loop-vectorize` & regenerated them with the `--filter-out-after` flag to reduce the size instead.
In the updated tests, the extractelement always uses the first `@llvm.vector.extract` so I will investigate the possible missing instcombine separately.

https://github.com/llvm/llvm-project/pull/147535