; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --version 5
; RUN: opt -passes=loop-vectorize -enable-vplan-native-path -S %s | FileCheck %s

@A = common global [1024 x i64] zeroinitializer, align 16
@B = common global [1024 x i64] zeroinitializer, align 16

define void @inner_latch_header_first_successor(i64 %N, i32 %c, i64 %M) {
; CHECK-LABEL: define void @inner_latch_header_first_successor(
; CHECK-SAME: i64 [[N:%.*]], i32 [[C:%.*]], i64 [[M:%.*]]) {
; CHECK-NEXT:  [[ENTRY:.*]]:
; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[M]], 4
; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK:       [[VECTOR_PH]]:
; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[M]], 4
; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[M]], [[N_MOD_VF]]
; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[C]], i64 0
; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
; CHECK-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <4 x i64> poison, i64 [[N]], i64 0
; CHECK-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT1]], <4 x i64> poison, <4 x i32> zeroinitializer
; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK:       [[VECTOR_BODY]]:
; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_LATCH:.*]] ]
; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_LATCH]] ]
; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1024 x i64], ptr @A, i64 0, <4 x i64> [[VEC_IND]]
; CHECK-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <4 x i64> @llvm.masked.gather.v4i64.v4p0(<4 x ptr> [[TMP0]], i32 4, <4 x i1> splat (i1 true), <4 x i64> poison)
; CHECK-NEXT:    call void @llvm.masked.scatter.v4i32.v4p0(<4 x i32> [[BROADCAST_SPLAT]], <4 x ptr> [[TMP0]], i32 4, <4 x i1> splat (i1 true))
; CHECK-NEXT:    br label %[[INNER3:.*]]
; CHECK:       [[INNER3]]:
; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i64> [ zeroinitializer, %[[VECTOR_BODY]] ], [ [[TMP4:%.*]], %[[INNER3]] ]
; CHECK-NEXT:    [[VEC_PHI4:%.*]] = phi <4 x i64> [ [[WIDE_MASKED_GATHER]], %[[VECTOR_BODY]] ], [ [[TMP3:%.*]], %[[INNER3]] ]
; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1024 x i64], ptr @B, i64 0, <4 x i64> [[VEC_PHI]]
; CHECK-NEXT:    [[WIDE_MASKED_GATHER5:%.*]] = call <4 x i64> @llvm.masked.gather.v4i64.v4p0(<4 x ptr> [[TMP1]], i32 4, <4 x i1> splat (i1 true), <4 x i64> poison)
; CHECK-NEXT:    [[TMP2:%.*]] = add nsw <4 x i64> [[WIDE_MASKED_GATHER5]], [[VEC_PHI4]]
; CHECK-NEXT:    [[TMP3]] = add nsw <4 x i64> [[TMP2]], [[VEC_PHI4]]
; CHECK-NEXT:    [[TMP4]] = add nuw nsw <4 x i64> [[VEC_PHI]], splat (i64 1)
; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq <4 x i64> [[TMP4]], [[BROADCAST_SPLAT2]]
; CHECK-NEXT:    [[TMP9:%.*]] = extractelement <4 x i1> [[TMP6]], i32 0
; CHECK-NEXT:    br i1 [[TMP9]], label %[[VECTOR_LATCH]], label %[[INNER3]]
; CHECK:       [[VECTOR_LATCH]]:
; CHECK-NEXT:    [[VEC_PHI6:%.*]] = phi <4 x i64> [ [[TMP3]], %[[INNER3]] ]
; CHECK-NEXT:    call void @llvm.masked.scatter.v4i64.v4p0(<4 x i64> [[VEC_PHI6]], <4 x ptr> [[TMP0]], i32 4, <4 x i1> splat (i1 true))
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT:    br i1 [[TMP7]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; CHECK:       [[MIDDLE_BLOCK]]:
; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[M]], [[N_VEC]]
; CHECK-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
; CHECK:       [[SCALAR_PH]]:
; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-NEXT:    br label %[[OUTER_HEADER:.*]]
; CHECK:       [[OUTER_HEADER]]:
; CHECK-NEXT:    [[IV_OUTER:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_OUTER_NEXT:%.*]], %[[OUTER_LATCH:.*]] ]
; CHECK-NEXT:    [[GEP_A:%.*]] = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 [[IV_OUTER]]
; CHECK-NEXT:    [[RED_START:%.*]] = load i64, ptr [[GEP_A]], align 4
; CHECK-NEXT:    store i32 [[C]], ptr [[GEP_A]], align 4
; CHECK-NEXT:    br label %[[INNER:.*]]
; CHECK:       [[INNER]]:
; CHECK-NEXT:    [[IV_INNER:%.*]] = phi i64 [ 0, %[[OUTER_HEADER]] ], [ [[IV_INNER_NEXT:%.*]], %[[INNER]] ]
; CHECK-NEXT:    [[RED:%.*]] = phi i64 [ [[RED_START]], %[[OUTER_HEADER]] ], [ [[RED_NEXT:%.*]], %[[INNER]] ]
; CHECK-NEXT:    [[GEP_B:%.*]] = getelementptr inbounds [1024 x i64], ptr @B, i64 0, i64 [[IV_INNER]]
; CHECK-NEXT:    [[TMP8:%.*]] = load i64, ptr [[GEP_B]], align 4
; CHECK-NEXT:    [[ADD:%.*]] = add nsw i64 [[TMP8]], [[RED]]
; CHECK-NEXT:    [[RED_NEXT]] = add nsw i64 [[ADD]], [[RED]]
; CHECK-NEXT:    [[IV_INNER_NEXT]] = add nuw nsw i64 [[IV_INNER]], 1
; CHECK-NEXT:    [[INNER_EC:%.*]] = icmp ne i64 [[IV_INNER_NEXT]], [[N]]
; CHECK-NEXT:    br i1 [[INNER_EC]], label %[[INNER]], label %[[OUTER_LATCH]]
; CHECK:       [[OUTER_LATCH]]:
; CHECK-NEXT:    [[RED_NEXT_LCSSA:%.*]] = phi i64 [ [[RED_NEXT]], %[[INNER]] ]
; CHECK-NEXT:    store i64 [[RED_NEXT_LCSSA]], ptr [[GEP_A]], align 4
; CHECK-NEXT:    [[IV_OUTER_NEXT]] = add nuw nsw i64 [[IV_OUTER]], 1
; CHECK-NEXT:    [[OUTER_EC:%.*]] = icmp eq i64 [[IV_OUTER_NEXT]], [[M]]
; CHECK-NEXT:    br i1 [[OUTER_EC]], label %[[EXIT]], label %[[OUTER_HEADER]], !llvm.loop [[LOOP3:![0-9]+]]
; CHECK:       [[EXIT]]:
; CHECK-NEXT:    ret void
;
entry:
  br label %outer.header

outer.header:
  %iv.outer = phi i64 [ 0, %entry ], [ %iv.outer.next, %outer.latch ]
  %gep.A = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.outer
  %red.start = load i64 , ptr %gep.A, align 4
  store i32 %c, ptr %gep.A, align 4
  br label %inner

inner:
  %iv.inner = phi i64 [ 0, %outer.header ], [ %iv.inner.next, %inner ]
  %red = phi i64 [ %red.start, %outer.header ], [ %red.next, %inner ]
  %gep.B = getelementptr inbounds [1024 x i64], ptr @B, i64 0, i64 %iv.inner
  %2 = load i64, ptr %gep.B, align 4
  %add = add nsw i64 %2, %red
  %red.next = add nsw i64 %add, %red
  %iv.inner.next = add nuw nsw i64 %iv.inner, 1
  %inner.ec = icmp ne i64 %iv.inner.next, %N
  br i1 %inner.ec, label %inner, label %outer.latch

outer.latch:
  store i64 %red.next, ptr %gep.A, align 4
  %iv.outer.next = add nuw nsw i64 %iv.outer, 1
  %outer.ec = icmp eq i64 %iv.outer.next, %M
  br i1 %outer.ec, label %exit, label %outer.header, !llvm.loop !1

exit:
  ret void
}


define void @inner_latch_header_second_successor(i64 %N, i32 %c, i64 %M) {
; CHECK-LABEL: define void @inner_latch_header_second_successor(
; CHECK-SAME: i64 [[N:%.*]], i32 [[C:%.*]], i64 [[M:%.*]]) {
; CHECK-NEXT:  [[ENTRY:.*]]:
; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[M]], 4
; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK:       [[VECTOR_PH]]:
; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[M]], 4
; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[M]], [[N_MOD_VF]]
; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[C]], i64 0
; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
; CHECK-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <4 x i64> poison, i64 [[N]], i64 0
; CHECK-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT1]], <4 x i64> poison, <4 x i32> zeroinitializer
; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK:       [[VECTOR_BODY]]:
; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_LATCH:.*]] ]
; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_LATCH]] ]
; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [1024 x i64], ptr @A, i64 0, <4 x i64> [[VEC_IND]]
; CHECK-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <4 x i64> @llvm.masked.gather.v4i64.v4p0(<4 x ptr> [[TMP0]], i32 4, <4 x i1> splat (i1 true), <4 x i64> poison)
; CHECK-NEXT:    call void @llvm.masked.scatter.v4i32.v4p0(<4 x i32> [[BROADCAST_SPLAT]], <4 x ptr> [[TMP0]], i32 4, <4 x i1> splat (i1 true))
; CHECK-NEXT:    br label %[[INNER3:.*]]
; CHECK:       [[INNER3]]:
; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i64> [ zeroinitializer, %[[VECTOR_BODY]] ], [ [[TMP4:%.*]], %[[INNER3]] ]
; CHECK-NEXT:    [[VEC_PHI4:%.*]] = phi <4 x i64> [ [[WIDE_MASKED_GATHER]], %[[VECTOR_BODY]] ], [ [[TMP3:%.*]], %[[INNER3]] ]
; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1024 x i64], ptr @B, i64 0, <4 x i64> [[VEC_PHI]]
; CHECK-NEXT:    [[WIDE_MASKED_GATHER5:%.*]] = call <4 x i64> @llvm.masked.gather.v4i64.v4p0(<4 x ptr> [[TMP1]], i32 4, <4 x i1> splat (i1 true), <4 x i64> poison)
; CHECK-NEXT:    [[TMP2:%.*]] = add nsw <4 x i64> [[WIDE_MASKED_GATHER5]], [[VEC_PHI4]]
; CHECK-NEXT:    [[TMP3]] = add nsw <4 x i64> [[TMP2]], [[VEC_PHI4]]
; CHECK-NEXT:    [[TMP4]] = add nuw nsw <4 x i64> [[VEC_PHI]], splat (i64 1)
; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq <4 x i64> [[TMP4]], [[BROADCAST_SPLAT2]]
; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <4 x i1> [[TMP5]], i32 0
; CHECK-NEXT:    br i1 [[TMP6]], label %[[VECTOR_LATCH]], label %[[INNER3]]
; CHECK:       [[VECTOR_LATCH]]:
; CHECK-NEXT:    [[VEC_PHI6:%.*]] = phi <4 x i64> [ [[TMP3]], %[[INNER3]] ]
; CHECK-NEXT:    call void @llvm.masked.scatter.v4i64.v4p0(<4 x i64> [[VEC_PHI6]], <4 x ptr> [[TMP0]], i32 4, <4 x i1> splat (i1 true))
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT:    br i1 [[TMP7]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
; CHECK:       [[MIDDLE_BLOCK]]:
; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[M]], [[N_VEC]]
; CHECK-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
; CHECK:       [[SCALAR_PH]]:
; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-NEXT:    br label %[[OUTER_HEADER:.*]]
; CHECK:       [[OUTER_HEADER]]:
; CHECK-NEXT:    [[IV_OUTER:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_OUTER_NEXT:%.*]], %[[OUTER_LATCH:.*]] ]
; CHECK-NEXT:    [[GEP_A:%.*]] = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 [[IV_OUTER]]
; CHECK-NEXT:    [[RED_START:%.*]] = load i64, ptr [[GEP_A]], align 4
; CHECK-NEXT:    store i32 [[C]], ptr [[GEP_A]], align 4
; CHECK-NEXT:    br label %[[INNER:.*]]
; CHECK:       [[INNER]]:
; CHECK-NEXT:    [[IV_INNER:%.*]] = phi i64 [ 0, %[[OUTER_HEADER]] ], [ [[IV_INNER_NEXT:%.*]], %[[INNER]] ]
; CHECK-NEXT:    [[RED:%.*]] = phi i64 [ [[RED_START]], %[[OUTER_HEADER]] ], [ [[RED_NEXT:%.*]], %[[INNER]] ]
; CHECK-NEXT:    [[GEP_B:%.*]] = getelementptr inbounds [1024 x i64], ptr @B, i64 0, i64 [[IV_INNER]]
; CHECK-NEXT:    [[TMP8:%.*]] = load i64, ptr [[GEP_B]], align 4
; CHECK-NEXT:    [[ADD:%.*]] = add nsw i64 [[TMP8]], [[RED]]
; CHECK-NEXT:    [[RED_NEXT]] = add nsw i64 [[ADD]], [[RED]]
; CHECK-NEXT:    [[IV_INNER_NEXT]] = add nuw nsw i64 [[IV_INNER]], 1
; CHECK-NEXT:    [[INNER_EC:%.*]] = icmp eq i64 [[IV_INNER_NEXT]], [[N]]
; CHECK-NEXT:    br i1 [[INNER_EC]], label %[[OUTER_LATCH]], label %[[INNER]]
; CHECK:       [[OUTER_LATCH]]:
; CHECK-NEXT:    [[RED_NEXT_LCSSA:%.*]] = phi i64 [ [[RED_NEXT]], %[[INNER]] ]
; CHECK-NEXT:    store i64 [[RED_NEXT_LCSSA]], ptr [[GEP_A]], align 4
; CHECK-NEXT:    [[IV_OUTER_NEXT]] = add nuw nsw i64 [[IV_OUTER]], 1
; CHECK-NEXT:    [[OUTER_EC:%.*]] = icmp eq i64 [[IV_OUTER_NEXT]], [[M]]
; CHECK-NEXT:    br i1 [[OUTER_EC]], label %[[EXIT]], label %[[OUTER_HEADER]], !llvm.loop [[LOOP5:![0-9]+]]
; CHECK:       [[EXIT]]:
; CHECK-NEXT:    ret void
;
entry:
  br label %outer.header

outer.header:
  %iv.outer = phi i64 [ 0, %entry ], [ %iv.outer.next, %outer.latch ]
  %gep.A = getelementptr inbounds [1024 x i64], ptr @A, i64 0, i64 %iv.outer
  %red.start = load i64 , ptr %gep.A, align 4
  store i32 %c, ptr %gep.A, align 4
  br label %inner

inner:
  %iv.inner = phi i64 [ 0, %outer.header ], [ %iv.inner.next, %inner ]
  %red = phi i64 [ %red.start, %outer.header ], [ %red.next, %inner ]
  %gep.B = getelementptr inbounds [1024 x i64], ptr @B, i64 0, i64 %iv.inner
  %2 = load i64, ptr %gep.B, align 4
  %add = add nsw i64 %2, %red
  %red.next = add nsw i64 %add, %red
  %iv.inner.next = add nuw nsw i64 %iv.inner, 1
  %inner.ec = icmp eq i64 %iv.inner.next, %N
  br i1 %inner.ec, label %outer.latch, label %inner

outer.latch:
  store i64 %red.next, ptr %gep.A, align 4
  %iv.outer.next = add nuw nsw i64 %iv.outer, 1
  %outer.ec = icmp eq i64 %iv.outer.next, %M
  br i1 %outer.ec, label %exit, label %outer.header, !llvm.loop !1

exit:
  ret void
}

!1 = distinct !{!1, !2, !3}
!2 = !{!"llvm.loop.vectorize.width", i32 4}
!3 = !{!"llvm.loop.vectorize.enable", i1 true}