[llvm] [LV] Use VPReductionRecipe for partial reductions (PR #147513)

[Sam Tebbs via llvm-commits]

================
@@ -410,8 +410,8 @@ define i32 @dotp_predicated(i64 %N, ptr %a, ptr %b) {
 ; CHECK-NEXT:    [[TMP178:%.*]] = sext <16 x i8> [[TMP177]] to <16 x i32>
 ; CHECK-NEXT:    [[TMP97:%.*]] = sext <16 x i8> [[TMP159]] to <16 x i32>
 ; CHECK-NEXT:    [[TMP179:%.*]] = mul nsw <16 x i32> [[TMP178]], [[TMP97]]
----------------
SamTebbs33 wrote:

They are stray updates, indeed, so I've reverted these unnecessary changes. Thanks!

https://github.com/llvm/llvm-project/pull/147513