[llvm] [VPlan] Expand VPWidenIntOrFpInductionRecipe into separate recipes (PR #118638)

[Luke Lau via llvm-commits]

================
@@ -109,8 +109,8 @@ define void @load_store_interleave_group_tc_2(ptr noalias %data) {
 ; VF4-NEXT:    store i64 [[TMP32]], ptr [[TMP31]], align 8
 ; VF4-NEXT:    br label %[[PRED_STORE_CONTINUE6]]
 ; VF4:       [[PRED_STORE_CONTINUE6]]:
-; VF4-NEXT:    [[VEC_IND_NEXT]] = add <4 x i8> [[VEC_IND]], splat (i8 4)
 ; VF4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; VF4-NEXT:    [[VEC_IND_NEXT]] = add <4 x i8> [[VEC_IND]], splat (i8 4)
----------------
lukel97 wrote:

I gave this a try by inserting it before the canonical IV increment but it ended up creating more test diffs. It looks like there's a lot of cases where the widened IV increment comes after the canonical IV, e.g. in test/Transforms/LoopVectorize/vector-geps.ll

https://github.com/llvm/llvm-project/pull/118638