[llvm] [llvm][optimization] Missed optimization : masked memory operations (PR #225028)

[Luke Lau via llvm-commits]

https://github.com/lukel97 commented:

I don't think this an optimisation we should do generically in VectorCombine as it's very easy for this to be an unprofitable transform. @ElvisWang123 and @eas are working on doing this within the loop vectorizer, where it should have more information about control flow + profiling information etc.

https://github.com/llvm/llvm-project/pull/225028