[llvm] [VPlan] Remove ILV::sinkScalarOperands. (PR #136023)

[Florian Hahn via llvm-commits]

================
@@ -92,6 +92,13 @@ define void @VF1-VPWidenCanonicalIVRecipeExe(ptr %ptr1) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE12:%.*]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 8
+; CHECK-NEXT:    [[TMP4:%.*]] = add i64 [[OFFSET_IDX]], 8
+; CHECK-NEXT:    [[TMP5:%.*]] = add i64 [[OFFSET_IDX]], 16
+; CHECK-NEXT:    [[TMP6:%.*]] = add i64 [[OFFSET_IDX]], 24
+; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[PTR1]], i64 [[OFFSET_IDX]]
+; CHECK-NEXT:    [[NEXT_GEP1:%.*]] = getelementptr i8, ptr [[PTR1]], i64 [[TMP4]]
+; CHECK-NEXT:    [[NEXT_GEP2:%.*]] = getelementptr i8, ptr [[PTR1]], i64 [[TMP5]]
+; CHECK-NEXT:    [[NEXT_GEP3:%.*]] = getelementptr i8, ptr [[PTR1]], i64 [[TMP6]]
----------------
fhahn wrote:

Cost of pointer inductions is assumed free at the moment IIRC

https://github.com/llvm/llvm-project/pull/136023