See [~psandoz] comment for scanAddI:
https://github.com/openjdk/jdk/pull/28639#discussion_r2738982413

"These are lane shifting operations, so another variant can use compress with the masks as input. Another could use slice, ideally the rearrange and slice variants would generate comparable code, or the compress and slice would."

I suspect that this approach could generate better code, but we should make sure that eventually the shuffles also become equally performant.