[llvm] [VPlan] Simplify add x, (sub 0, y) -> sub x, y (PR #187228)

[Ramkumar Ramachandra via llvm-commits]

================
@@ -41,14 +41,14 @@ define i64 @select_decreasing_induction_icmp_const_start(ptr %a) {
 ; IC4VF4-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; IC4VF4:       [[VECTOR_BODY]]:
 ; IC4VF4-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
-; IC4VF4-NEXT:    [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 19999, i64 19998, i64 19997, i64 19996>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; IC4VF4-NEXT:    [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 19999, i64 19998, i64 19997, i64 19996>, %[[VECTOR_PH]] ], [ [[TMP19:%.*]], %[[VECTOR_BODY]] ]
 ; IC4VF4-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i64> [ splat (i64 9223372036854775807), %[[VECTOR_PH]] ], [ [[TMP13:%.*]], %[[VECTOR_BODY]] ]
 ; IC4VF4-NEXT:    [[VEC_PHI1:%.*]] = phi <4 x i64> [ splat (i64 9223372036854775807), %[[VECTOR_PH]] ], [ [[TMP14:%.*]], %[[VECTOR_BODY]] ]
 ; IC4VF4-NEXT:    [[VEC_PHI2:%.*]] = phi <4 x i64> [ splat (i64 9223372036854775807), %[[VECTOR_PH]] ], [ [[TMP15:%.*]], %[[VECTOR_BODY]] ]
 ; IC4VF4-NEXT:    [[VEC_PHI3:%.*]] = phi <4 x i64> [ splat (i64 9223372036854775807), %[[VECTOR_PH]] ], [ [[TMP16:%.*]], %[[VECTOR_BODY]] ]
-; IC4VF4-NEXT:    [[STEP_ADD:%.*]] = add nsw <4 x i64> [[VEC_IND]], splat (i64 -4)
-; IC4VF4-NEXT:    [[STEP_ADD_2:%.*]] = add nsw <4 x i64> [[STEP_ADD]], splat (i64 -4)
-; IC4VF4-NEXT:    [[STEP_ADD_3:%.*]] = add nsw <4 x i64> [[STEP_ADD_2]], splat (i64 -4)
+; IC4VF4-NEXT:    [[STEP_ADD:%.*]] = sub <4 x i64> [[VEC_IND]], splat (i64 4)
----------------
artagnon wrote:

Well, it's impossible to preserve nsw because https://github.com/llvm/llvm-project/pull/191284 is incorrect.

https://github.com/llvm/llvm-project/pull/187228