llvm-project

42a6e0da - [InstCombine] Fold canonicalized form of vecreduce_or(get_active_lane_mask) (#220085)

[InstCombine] Fold canonicalized form of vecreduce_or(get_active_lane_mask) (#220085) get_active_lane_mask is called with lower and upper bounds. We can just compare the bounds themselves. This change is intended to improve vecreduce_or codegen for fixed-length vectors. vscales deal with vecreduce_or directly. https://godbolt.org/z/84sjGGdWe alive2 doesn't seem to support get_active_lane_mask at the moment.