[llvm] [LV] Vectorize predicated loop-invariant loads via masked load + broadcast (PR #200361)

[Florian Hahn via llvm-commits]

================

----------------
fhahn wrote:

the horizontal reduction could be expensive, should this compare cost of any-of, masked-load, broadcast vs cost of replicated load?

https://github.com/llvm/llvm-project/pull/200361