Compiler projects using llvm
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt < %s -loop-vectorize -force-vector-width=4 -force-vector-interleave=2 -S | FileCheck %s --check-prefix=UNROLL-NO-IC
; RUN: opt < %s -loop-vectorize -force-vector-width=1 -force-vector-interleave=2 -S | FileCheck %s --check-prefix=UNROLL-NO-VF
; RUN: opt < %s -loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -S | FileCheck %s --check-prefix=SINK-AFTER

target datalayout = "e-m:e-i64:64-i128:128-n32:64-S128"

; void recurrence_1(int *a, int *b, int n) {
;   for(int i = 0; i < n; i++)
;     b[i] =  a[i] + a[i - 1]
; }
;
;
;
define void @recurrence_1(i32* readonly noalias %a, i32* noalias %b, i32 %n) {
; UNROLL-NO-IC-LABEL: @recurrence_1(
; UNROLL-NO-IC-NEXT:  entry:
; UNROLL-NO-IC-NEXT:    br label [[FOR_PREHEADER:%.*]]
; UNROLL-NO-IC:       for.preheader:
; UNROLL-NO-IC-NEXT:    [[ARRAYIDX_PHI_TRANS_INSERT:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 0
; UNROLL-NO-IC-NEXT:    [[PRE_LOAD:%.*]] = load i32, i32* [[ARRAYIDX_PHI_TRANS_INSERT]], align 4
; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i32 [[N:%.*]], -1
; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 8
; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC:       vector.ph:
; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 8
; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i32> poison, i32 [[PRE_LOAD]], i32 3
; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC:       vector.body:
; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD1:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add i64 [[INDEX]], 0
; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 4
; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = add nuw nsw i64 [[TMP3]], 1
; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = add nuw nsw i64 [[TMP4]], 1
; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP5]]
; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP6]]
; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[TMP7]], i32 0
; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = bitcast i32* [[TMP9]] to <4 x i32>*
; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, <4 x i32>* [[TMP10]], align 4
; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, i32* [[TMP7]], i32 4
; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = bitcast i32* [[TMP11]] to <4 x i32>*
; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD1]] = load <4 x i32>, <4 x i32>* [[TMP12]], align 4
; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = shufflevector <4 x i32> [[WIDE_LOAD]], <4 x i32> [[WIDE_LOAD1]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], i64 [[TMP3]]
; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[TMP4]]
; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = add <4 x i32> [[WIDE_LOAD]], [[TMP13]]
; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = add <4 x i32> [[WIDE_LOAD1]], [[TMP14]]
; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, i32* [[TMP15]], i32 0
; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = bitcast i32* [[TMP19]] to <4 x i32>*
; UNROLL-NO-IC-NEXT:    store <4 x i32> [[TMP17]], <4 x i32>* [[TMP20]], align 4
; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[TMP15]], i32 4
; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = bitcast i32* [[TMP21]] to <4 x i32>*
; UNROLL-NO-IC-NEXT:    store <4 x i32> [[TMP18]], <4 x i32>* [[TMP22]], align 4
; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-IC-NEXT:    br i1 [[TMP23]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; UNROLL-NO-IC:       middle.block:
; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[WIDE_LOAD1]], i32 3
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[WIDE_LOAD1]], i32 2
; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC:       scalar.ph:
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[PRE_LOAD]], [[FOR_PREHEADER]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_PREHEADER]] ]
; UNROLL-NO-IC-NEXT:    br label [[SCALAR_BODY:%.*]]
; UNROLL-NO-IC:       scalar.body:
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP24:%.*]], [[SCALAR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[SCALAR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; UNROLL-NO-IC-NEXT:    [[ARRAYIDX32:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV_NEXT]]
; UNROLL-NO-IC-NEXT:    [[TMP24]] = load i32, i32* [[ARRAYIDX32]], align 4
; UNROLL-NO-IC-NEXT:    [[ARRAYIDX34:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
; UNROLL-NO-IC-NEXT:    [[ADD35:%.*]] = add i32 [[TMP24]], [[SCALAR_RECUR]]
; UNROLL-NO-IC-NEXT:    store i32 [[ADD35]], i32* [[ARRAYIDX34]], align 4
; UNROLL-NO-IC-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
; UNROLL-NO-IC-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
; UNROLL-NO-IC-NEXT:    br i1 [[EXITCOND]], label [[FOR_EXIT]], label [[SCALAR_BODY]], !llvm.loop [[LOOP2:![0-9]+]]
; UNROLL-NO-IC:       for.exit:
; UNROLL-NO-IC-NEXT:    ret void
;
; UNROLL-NO-VF-LABEL: @recurrence_1(
; UNROLL-NO-VF-NEXT:  entry:
; UNROLL-NO-VF-NEXT:    br label [[FOR_PREHEADER:%.*]]
; UNROLL-NO-VF:       for.preheader:
; UNROLL-NO-VF-NEXT:    [[ARRAYIDX_PHI_TRANS_INSERT:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 0
; UNROLL-NO-VF-NEXT:    [[PRE_LOAD:%.*]] = load i32, i32* [[ARRAYIDX_PHI_TRANS_INSERT]], align 4
; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = add i32 [[N:%.*]], -1
; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
; UNROLL-NO-VF-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 2
; UNROLL-NO-VF-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-VF:       vector.ph:
; UNROLL-NO-VF-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 2
; UNROLL-NO-VF-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-VF:       vector.body:
; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi i32 [ [[PRE_LOAD]], [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[INDUCTION:%.*]] = add i64 [[INDEX]], 0
; UNROLL-NO-VF-NEXT:    [[INDUCTION1:%.*]] = add i64 [[INDEX]], 1
; UNROLL-NO-VF-NEXT:    [[TMP3:%.*]] = add nuw nsw i64 [[INDUCTION]], 1
; UNROLL-NO-VF-NEXT:    [[TMP4:%.*]] = add nuw nsw i64 [[INDUCTION1]], 1
; UNROLL-NO-VF-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP3]]
; UNROLL-NO-VF-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP4]]
; UNROLL-NO-VF-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP5]], align 4
; UNROLL-NO-VF-NEXT:    [[TMP8]] = load i32, i32* [[TMP6]], align 4
; UNROLL-NO-VF-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], i64 [[INDUCTION]]
; UNROLL-NO-VF-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDUCTION1]]
; UNROLL-NO-VF-NEXT:    [[TMP11:%.*]] = add i32 [[TMP7]], [[VECTOR_RECUR]]
; UNROLL-NO-VF-NEXT:    [[TMP12:%.*]] = add i32 [[TMP8]], [[TMP7]]
; UNROLL-NO-VF-NEXT:    store i32 [[TMP11]], i32* [[TMP9]], align 4
; UNROLL-NO-VF-NEXT:    store i32 [[TMP12]], i32* [[TMP10]], align 4
; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; UNROLL-NO-VF-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-VF-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; UNROLL-NO-VF:       middle.block:
; UNROLL-NO-VF-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; UNROLL-NO-VF-NEXT:    br i1 [[CMP_N]], label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-VF:       scalar.ph:
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[PRE_LOAD]], [[FOR_PREHEADER]] ], [ [[TMP8]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_PREHEADER]] ]
; UNROLL-NO-VF-NEXT:    br label [[SCALAR_BODY:%.*]]
; UNROLL-NO-VF:       scalar.body:
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP14:%.*]], [[SCALAR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[SCALAR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; UNROLL-NO-VF-NEXT:    [[ARRAYIDX32:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV_NEXT]]
; UNROLL-NO-VF-NEXT:    [[TMP14]] = load i32, i32* [[ARRAYIDX32]], align 4
; UNROLL-NO-VF-NEXT:    [[ARRAYIDX34:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
; UNROLL-NO-VF-NEXT:    [[ADD35:%.*]] = add i32 [[TMP14]], [[SCALAR_RECUR]]
; UNROLL-NO-VF-NEXT:    store i32 [[ADD35]], i32* [[ARRAYIDX34]], align 4
; UNROLL-NO-VF-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
; UNROLL-NO-VF-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
; UNROLL-NO-VF-NEXT:    br i1 [[EXITCOND]], label [[FOR_EXIT]], label [[SCALAR_BODY]], !llvm.loop [[LOOP2:![0-9]+]]
; UNROLL-NO-VF:       for.exit:
; UNROLL-NO-VF-NEXT:    ret void
;
; SINK-AFTER-LABEL: @recurrence_1(
; SINK-AFTER-NEXT:  entry:
; SINK-AFTER-NEXT:    br label [[FOR_PREHEADER:%.*]]
; SINK-AFTER:       for.preheader:
; SINK-AFTER-NEXT:    [[ARRAYIDX_PHI_TRANS_INSERT:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 0
; SINK-AFTER-NEXT:    [[PRE_LOAD:%.*]] = load i32, i32* [[ARRAYIDX_PHI_TRANS_INSERT]], align 4
; SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i32 [[N:%.*]], -1
; SINK-AFTER-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
; SINK-AFTER-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
; SINK-AFTER-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 4
; SINK-AFTER-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; SINK-AFTER:       vector.ph:
; SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 4
; SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i32> poison, i32 [[PRE_LOAD]], i32 3
; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
; SINK-AFTER:       vector.body:
; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[TMP3:%.*]] = add i64 [[INDEX]], 0
; SINK-AFTER-NEXT:    [[TMP4:%.*]] = add nuw nsw i64 [[TMP3]], 1
; SINK-AFTER-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP4]]
; SINK-AFTER-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[TMP5]], i32 0
; SINK-AFTER-NEXT:    [[TMP7:%.*]] = bitcast i32* [[TMP6]] to <4 x i32>*
; SINK-AFTER-NEXT:    [[WIDE_LOAD]] = load <4 x i32>, <4 x i32>* [[TMP7]], align 4
; SINK-AFTER-NEXT:    [[TMP8:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; SINK-AFTER-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], i64 [[TMP3]]
; SINK-AFTER-NEXT:    [[TMP10:%.*]] = add <4 x i32> [[WIDE_LOAD]], [[TMP8]]
; SINK-AFTER-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, i32* [[TMP9]], i32 0
; SINK-AFTER-NEXT:    [[TMP12:%.*]] = bitcast i32* [[TMP11]] to <4 x i32>*
; SINK-AFTER-NEXT:    store <4 x i32> [[TMP10]], <4 x i32>* [[TMP12]], align 4
; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; SINK-AFTER-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; SINK-AFTER-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; SINK-AFTER:       middle.block:
; SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i32 3
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i32 2
; SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
; SINK-AFTER:       scalar.ph:
; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[PRE_LOAD]], [[FOR_PREHEADER]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_PREHEADER]] ]
; SINK-AFTER-NEXT:    br label [[SCALAR_BODY:%.*]]
; SINK-AFTER:       scalar.body:
; SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP14:%.*]], [[SCALAR_BODY]] ]
; SINK-AFTER-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[SCALAR_BODY]] ]
; SINK-AFTER-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; SINK-AFTER-NEXT:    [[ARRAYIDX32:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV_NEXT]]
; SINK-AFTER-NEXT:    [[TMP14]] = load i32, i32* [[ARRAYIDX32]], align 4
; SINK-AFTER-NEXT:    [[ARRAYIDX34:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
; SINK-AFTER-NEXT:    [[ADD35:%.*]] = add i32 [[TMP14]], [[SCALAR_RECUR]]
; SINK-AFTER-NEXT:    store i32 [[ADD35]], i32* [[ARRAYIDX34]], align 4
; SINK-AFTER-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
; SINK-AFTER-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
; SINK-AFTER-NEXT:    br i1 [[EXITCOND]], label [[FOR_EXIT]], label [[SCALAR_BODY]], !llvm.loop [[LOOP2:![0-9]+]]
; SINK-AFTER:       for.exit:
; SINK-AFTER-NEXT:    ret void
;
entry:
  br label %for.preheader

for.preheader:
  %arrayidx.phi.trans.insert = getelementptr inbounds i32, i32* %a, i64 0
  %pre_load = load i32, i32* %arrayidx.phi.trans.insert
  br label %scalar.body

scalar.body:
  %0 = phi i32 [ %pre_load, %for.preheader ], [ %1, %scalar.body ]
  %indvars.iv = phi i64 [ 0, %for.preheader ], [ %indvars.iv.next, %scalar.body ]
  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
  %arrayidx32 = getelementptr inbounds i32, i32* %a, i64 %indvars.iv.next
  %1 = load i32, i32* %arrayidx32
  %arrayidx34 = getelementptr inbounds i32, i32* %b, i64 %indvars.iv
  %add35 = add i32 %1, %0
  store i32 %add35, i32* %arrayidx34
  %lftr.wideiv = trunc i64 %indvars.iv.next to i32
  %exitcond = icmp eq i32 %lftr.wideiv, %n
  br i1 %exitcond, label %for.exit, label %scalar.body

for.exit:
  ret void
}

; int recurrence_2(int *a, int n) {
;   int minmax;
;   for (int i = 0; i < n; ++i)
;     minmax = min(minmax, max(a[i] - a[i-1], 0));
;   return minmax;
; }
;
;
;
define i32 @recurrence_2(i32* nocapture readonly %a, i32 %n) {
; UNROLL-NO-IC-LABEL: @recurrence_2(
; UNROLL-NO-IC-NEXT:  entry:
; UNROLL-NO-IC-NEXT:    [[CMP27:%.*]] = icmp sgt i32 [[N:%.*]], 0
; UNROLL-NO-IC-NEXT:    br i1 [[CMP27]], label [[FOR_PREHEADER:%.*]], label [[FOR_COND_CLEANUP:%.*]]
; UNROLL-NO-IC:       for.preheader:
; UNROLL-NO-IC-NEXT:    [[ARRAYIDX2_PHI_TRANS_INSERT:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 -1
; UNROLL-NO-IC-NEXT:    [[DOTPRE:%.*]] = load i32, i32* [[ARRAYIDX2_PHI_TRANS_INSERT]], align 4
; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 8
; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC:       vector.ph:
; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 8
; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i32> poison, i32 [[DOTPRE]], i32 3
; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC:       vector.body:
; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD2:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ poison, [[VECTOR_PH]] ], [ [[TMP21:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[VEC_PHI1:%.*]] = phi <4 x i32> [ poison, [[VECTOR_PH]] ], [ [[TMP22:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add i64 [[INDEX]], 0
; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 4
; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP3]]
; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP4]]
; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, i32* [[TMP5]], i32 0
; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = bitcast i32* [[TMP7]] to <4 x i32>*
; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, <4 x i32>* [[TMP8]], align 4
; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[TMP5]], i32 4
; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = bitcast i32* [[TMP9]] to <4 x i32>*
; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD2]] = load <4 x i32>, <4 x i32>* [[TMP10]], align 4
; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = shufflevector <4 x i32> [[WIDE_LOAD]], <4 x i32> [[WIDE_LOAD2]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = sub nsw <4 x i32> [[WIDE_LOAD]], [[TMP11]]
; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = sub nsw <4 x i32> [[WIDE_LOAD2]], [[TMP12]]
; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = icmp sgt <4 x i32> [[TMP13]], zeroinitializer
; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = icmp sgt <4 x i32> [[TMP14]], zeroinitializer
; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = select <4 x i1> [[TMP15]], <4 x i32> [[TMP13]], <4 x i32> zeroinitializer
; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = select <4 x i1> [[TMP16]], <4 x i32> [[TMP14]], <4 x i32> zeroinitializer
; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = icmp slt <4 x i32> [[VEC_PHI]], [[TMP17]]
; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = icmp slt <4 x i32> [[VEC_PHI1]], [[TMP18]]
; UNROLL-NO-IC-NEXT:    [[TMP21]] = select <4 x i1> [[TMP19]], <4 x i32> [[VEC_PHI]], <4 x i32> [[TMP17]]
; UNROLL-NO-IC-NEXT:    [[TMP22]] = select <4 x i1> [[TMP20]], <4 x i32> [[VEC_PHI1]], <4 x i32> [[TMP18]]
; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-IC-NEXT:    br i1 [[TMP23]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
; UNROLL-NO-IC:       middle.block:
; UNROLL-NO-IC-NEXT:    [[RDX_MINMAX_CMP:%.*]] = icmp slt <4 x i32> [[TMP21]], [[TMP22]]
; UNROLL-NO-IC-NEXT:    [[RDX_MINMAX_SELECT:%.*]] = select <4 x i1> [[RDX_MINMAX_CMP]], <4 x i32> [[TMP21]], <4 x i32> [[TMP22]]
; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = call i32 @llvm.vector.reduce.smin.v4i32(<4 x i32> [[RDX_MINMAX_SELECT]])
; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[WIDE_LOAD2]], i32 3
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[WIDE_LOAD2]], i32 2
; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC:       scalar.ph:
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[DOTPRE]], [[FOR_PREHEADER]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_PREHEADER]] ]
; UNROLL-NO-IC-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ poison, [[FOR_PREHEADER]] ], [ [[TMP24]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT:    br label [[SCALAR_BODY:%.*]]
; UNROLL-NO-IC:       for.cond.cleanup.loopexit:
; UNROLL-NO-IC-NEXT:    [[MINMAX_0_COND_LCSSA:%.*]] = phi i32 [ [[MINMAX_0_COND:%.*]], [[SCALAR_BODY]] ], [ [[TMP24]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT:    br label [[FOR_COND_CLEANUP]]
; UNROLL-NO-IC:       for.cond.cleanup:
; UNROLL-NO-IC-NEXT:    [[MINMAX_0_LCSSA:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[MINMAX_0_COND_LCSSA]], [[FOR_COND_CLEANUP_LOOPEXIT]] ]
; UNROLL-NO-IC-NEXT:    ret i32 [[MINMAX_0_LCSSA]]
; UNROLL-NO-IC:       scalar.body:
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP25:%.*]], [[SCALAR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[SCALAR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[MINMAX_028:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[MINMAX_0_COND]], [[SCALAR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV]]
; UNROLL-NO-IC-NEXT:    [[TMP25]] = load i32, i32* [[ARRAYIDX]], align 4
; UNROLL-NO-IC-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[TMP25]], [[SCALAR_RECUR]]
; UNROLL-NO-IC-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[SUB3]], 0
; UNROLL-NO-IC-NEXT:    [[COND:%.*]] = select i1 [[CMP4]], i32 [[SUB3]], i32 0
; UNROLL-NO-IC-NEXT:    [[CMP5:%.*]] = icmp slt i32 [[MINMAX_028]], [[COND]]
; UNROLL-NO-IC-NEXT:    [[MINMAX_0_COND]] = select i1 [[CMP5]], i32 [[MINMAX_028]], i32 [[COND]]
; UNROLL-NO-IC-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; UNROLL-NO-IC-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
; UNROLL-NO-IC-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
; UNROLL-NO-IC-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP_LOOPEXIT]], label [[SCALAR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
;
; UNROLL-NO-VF-LABEL: @recurrence_2(
; UNROLL-NO-VF-NEXT:  entry:
; UNROLL-NO-VF-NEXT:    [[CMP27:%.*]] = icmp sgt i32 [[N:%.*]], 0
; UNROLL-NO-VF-NEXT:    br i1 [[CMP27]], label [[FOR_PREHEADER:%.*]], label [[FOR_COND_CLEANUP:%.*]]
; UNROLL-NO-VF:       for.preheader:
; UNROLL-NO-VF-NEXT:    [[ARRAYIDX2_PHI_TRANS_INSERT:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 -1
; UNROLL-NO-VF-NEXT:    [[DOTPRE:%.*]] = load i32, i32* [[ARRAYIDX2_PHI_TRANS_INSERT]], align 4
; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
; UNROLL-NO-VF-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 2
; UNROLL-NO-VF-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-VF:       vector.ph:
; UNROLL-NO-VF-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 2
; UNROLL-NO-VF-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-VF:       vector.body:
; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi i32 [ [[DOTPRE]], [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ poison, [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[VEC_PHI1:%.*]] = phi i32 [ poison, [[VECTOR_PH]] ], [ [[TMP16:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[INDUCTION:%.*]] = add i64 [[INDEX]], 0
; UNROLL-NO-VF-NEXT:    [[INDUCTION2:%.*]] = add i64 [[INDEX]], 1
; UNROLL-NO-VF-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDUCTION]]
; UNROLL-NO-VF-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDUCTION2]]
; UNROLL-NO-VF-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP3]], align 4
; UNROLL-NO-VF-NEXT:    [[TMP6]] = load i32, i32* [[TMP4]], align 4
; UNROLL-NO-VF-NEXT:    [[TMP7:%.*]] = sub nsw i32 [[TMP5]], [[VECTOR_RECUR]]
; UNROLL-NO-VF-NEXT:    [[TMP8:%.*]] = sub nsw i32 [[TMP6]], [[TMP5]]
; UNROLL-NO-VF-NEXT:    [[TMP9:%.*]] = icmp sgt i32 [[TMP7]], 0
; UNROLL-NO-VF-NEXT:    [[TMP10:%.*]] = icmp sgt i32 [[TMP8]], 0
; UNROLL-NO-VF-NEXT:    [[TMP11:%.*]] = select i1 [[TMP9]], i32 [[TMP7]], i32 0
; UNROLL-NO-VF-NEXT:    [[TMP12:%.*]] = select i1 [[TMP10]], i32 [[TMP8]], i32 0
; UNROLL-NO-VF-NEXT:    [[TMP13:%.*]] = icmp slt i32 [[VEC_PHI]], [[TMP11]]
; UNROLL-NO-VF-NEXT:    [[TMP14:%.*]] = icmp slt i32 [[VEC_PHI1]], [[TMP12]]
; UNROLL-NO-VF-NEXT:    [[TMP15]] = select i1 [[TMP13]], i32 [[VEC_PHI]], i32 [[TMP11]]
; UNROLL-NO-VF-NEXT:    [[TMP16]] = select i1 [[TMP14]], i32 [[VEC_PHI1]], i32 [[TMP12]]
; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; UNROLL-NO-VF-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-VF-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
; UNROLL-NO-VF:       middle.block:
; UNROLL-NO-VF-NEXT:    [[RDX_MINMAX_CMP:%.*]] = icmp slt i32 [[TMP15]], [[TMP16]]
; UNROLL-NO-VF-NEXT:    [[RDX_MINMAX_SELECT:%.*]] = select i1 [[RDX_MINMAX_CMP]], i32 [[TMP15]], i32 [[TMP16]]
; UNROLL-NO-VF-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; UNROLL-NO-VF-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-VF:       scalar.ph:
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[DOTPRE]], [[FOR_PREHEADER]] ], [ [[TMP6]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_PREHEADER]] ]
; UNROLL-NO-VF-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ poison, [[FOR_PREHEADER]] ], [ [[RDX_MINMAX_SELECT]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-VF-NEXT:    br label [[SCALAR_BODY:%.*]]
; UNROLL-NO-VF:       for.cond.cleanup.loopexit:
; UNROLL-NO-VF-NEXT:    [[MINMAX_0_COND_LCSSA:%.*]] = phi i32 [ [[MINMAX_0_COND:%.*]], [[SCALAR_BODY]] ], [ [[RDX_MINMAX_SELECT]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-VF-NEXT:    br label [[FOR_COND_CLEANUP]]
; UNROLL-NO-VF:       for.cond.cleanup:
; UNROLL-NO-VF-NEXT:    [[MINMAX_0_LCSSA:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[MINMAX_0_COND_LCSSA]], [[FOR_COND_CLEANUP_LOOPEXIT]] ]
; UNROLL-NO-VF-NEXT:    ret i32 [[MINMAX_0_LCSSA]]
; UNROLL-NO-VF:       scalar.body:
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP18:%.*]], [[SCALAR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[SCALAR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[MINMAX_028:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[MINMAX_0_COND]], [[SCALAR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV]]
; UNROLL-NO-VF-NEXT:    [[TMP18]] = load i32, i32* [[ARRAYIDX]], align 4
; UNROLL-NO-VF-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[TMP18]], [[SCALAR_RECUR]]
; UNROLL-NO-VF-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[SUB3]], 0
; UNROLL-NO-VF-NEXT:    [[COND:%.*]] = select i1 [[CMP4]], i32 [[SUB3]], i32 0
; UNROLL-NO-VF-NEXT:    [[CMP5:%.*]] = icmp slt i32 [[MINMAX_028]], [[COND]]
; UNROLL-NO-VF-NEXT:    [[MINMAX_0_COND]] = select i1 [[CMP5]], i32 [[MINMAX_028]], i32 [[COND]]
; UNROLL-NO-VF-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; UNROLL-NO-VF-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
; UNROLL-NO-VF-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
; UNROLL-NO-VF-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP_LOOPEXIT]], label [[SCALAR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
;
; SINK-AFTER-LABEL: @recurrence_2(
; SINK-AFTER-NEXT:  entry:
; SINK-AFTER-NEXT:    [[CMP27:%.*]] = icmp sgt i32 [[N:%.*]], 0
; SINK-AFTER-NEXT:    br i1 [[CMP27]], label [[FOR_PREHEADER:%.*]], label [[FOR_COND_CLEANUP:%.*]]
; SINK-AFTER:       for.preheader:
; SINK-AFTER-NEXT:    [[ARRAYIDX2_PHI_TRANS_INSERT:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 -1
; SINK-AFTER-NEXT:    [[DOTPRE:%.*]] = load i32, i32* [[ARRAYIDX2_PHI_TRANS_INSERT]], align 4
; SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
; SINK-AFTER-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
; SINK-AFTER-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
; SINK-AFTER-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 4
; SINK-AFTER-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; SINK-AFTER:       vector.ph:
; SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 4
; SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i32> poison, i32 [[DOTPRE]], i32 3
; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
; SINK-AFTER:       vector.body:
; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ poison, [[VECTOR_PH]] ], [ [[TMP12:%.*]], [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[TMP3:%.*]] = add i64 [[INDEX]], 0
; SINK-AFTER-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP3]]
; SINK-AFTER-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[TMP4]], i32 0
; SINK-AFTER-NEXT:    [[TMP6:%.*]] = bitcast i32* [[TMP5]] to <4 x i32>*
; SINK-AFTER-NEXT:    [[WIDE_LOAD]] = load <4 x i32>, <4 x i32>* [[TMP6]], align 4
; SINK-AFTER-NEXT:    [[TMP7:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; SINK-AFTER-NEXT:    [[TMP8:%.*]] = sub nsw <4 x i32> [[WIDE_LOAD]], [[TMP7]]
; SINK-AFTER-NEXT:    [[TMP9:%.*]] = icmp sgt <4 x i32> [[TMP8]], zeroinitializer
; SINK-AFTER-NEXT:    [[TMP10:%.*]] = select <4 x i1> [[TMP9]], <4 x i32> [[TMP8]], <4 x i32> zeroinitializer
; SINK-AFTER-NEXT:    [[TMP11:%.*]] = icmp slt <4 x i32> [[VEC_PHI]], [[TMP10]]
; SINK-AFTER-NEXT:    [[TMP12]] = select <4 x i1> [[TMP11]], <4 x i32> [[VEC_PHI]], <4 x i32> [[TMP10]]
; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; SINK-AFTER-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; SINK-AFTER-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
; SINK-AFTER:       middle.block:
; SINK-AFTER-NEXT:    [[TMP14:%.*]] = call i32 @llvm.vector.reduce.smin.v4i32(<4 x i32> [[TMP12]])
; SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i32 3
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i32 2
; SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[SCALAR_PH]]
; SINK-AFTER:       scalar.ph:
; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[DOTPRE]], [[FOR_PREHEADER]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_PREHEADER]] ]
; SINK-AFTER-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ poison, [[FOR_PREHEADER]] ], [ [[TMP14]], [[MIDDLE_BLOCK]] ]
; SINK-AFTER-NEXT:    br label [[SCALAR_BODY:%.*]]
; SINK-AFTER:       for.cond.cleanup.loopexit:
; SINK-AFTER-NEXT:    [[MINMAX_0_COND_LCSSA:%.*]] = phi i32 [ [[MINMAX_0_COND:%.*]], [[SCALAR_BODY]] ], [ [[TMP14]], [[MIDDLE_BLOCK]] ]
; SINK-AFTER-NEXT:    br label [[FOR_COND_CLEANUP]]
; SINK-AFTER:       for.cond.cleanup:
; SINK-AFTER-NEXT:    [[MINMAX_0_LCSSA:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[MINMAX_0_COND_LCSSA]], [[FOR_COND_CLEANUP_LOOPEXIT]] ]
; SINK-AFTER-NEXT:    ret i32 [[MINMAX_0_LCSSA]]
; SINK-AFTER:       scalar.body:
; SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP15:%.*]], [[SCALAR_BODY]] ]
; SINK-AFTER-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[SCALAR_BODY]] ]
; SINK-AFTER-NEXT:    [[MINMAX_028:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[MINMAX_0_COND]], [[SCALAR_BODY]] ]
; SINK-AFTER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[INDVARS_IV]]
; SINK-AFTER-NEXT:    [[TMP15]] = load i32, i32* [[ARRAYIDX]], align 4
; SINK-AFTER-NEXT:    [[SUB3:%.*]] = sub nsw i32 [[TMP15]], [[SCALAR_RECUR]]
; SINK-AFTER-NEXT:    [[CMP4:%.*]] = icmp sgt i32 [[SUB3]], 0
; SINK-AFTER-NEXT:    [[COND:%.*]] = select i1 [[CMP4]], i32 [[SUB3]], i32 0
; SINK-AFTER-NEXT:    [[CMP5:%.*]] = icmp slt i32 [[MINMAX_028]], [[COND]]
; SINK-AFTER-NEXT:    [[MINMAX_0_COND]] = select i1 [[CMP5]], i32 [[MINMAX_028]], i32 [[COND]]
; SINK-AFTER-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; SINK-AFTER-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32
; SINK-AFTER-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
; SINK-AFTER-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP_LOOPEXIT]], label [[SCALAR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
;
entry:
  %cmp27 = icmp sgt i32 %n, 0
  br i1 %cmp27, label %for.preheader, label %for.cond.cleanup

for.preheader:
  %arrayidx2.phi.trans.insert = getelementptr inbounds i32, i32* %a, i64 -1
  %.pre = load i32, i32* %arrayidx2.phi.trans.insert, align 4
  br label %scalar.body

for.cond.cleanup.loopexit:
  %minmax.0.cond.lcssa = phi i32 [ %minmax.0.cond, %scalar.body ]
  br label %for.cond.cleanup

for.cond.cleanup:
  %minmax.0.lcssa = phi i32 [ poison, %entry ], [ %minmax.0.cond.lcssa, %for.cond.cleanup.loopexit ]
  ret i32 %minmax.0.lcssa

scalar.body:
  %0 = phi i32 [ %.pre, %for.preheader ], [ %1, %scalar.body ]
  %indvars.iv = phi i64 [ 0, %for.preheader ], [ %indvars.iv.next, %scalar.body ]
  %minmax.028 = phi i32 [ poison, %for.preheader ], [ %minmax.0.cond, %scalar.body ]
  %arrayidx = getelementptr inbounds i32, i32* %a, i64 %indvars.iv
  %1 = load i32, i32* %arrayidx, align 4
  %sub3 = sub nsw i32 %1, %0
  %cmp4 = icmp sgt i32 %sub3, 0
  %cond = select i1 %cmp4, i32 %sub3, i32 0
  %cmp5 = icmp slt i32 %minmax.028, %cond
  %minmax.0.cond = select i1 %cmp5, i32 %minmax.028, i32 %cond
  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
  %lftr.wideiv = trunc i64 %indvars.iv.next to i32
  %exitcond = icmp eq i32 %lftr.wideiv, %n
  br i1 %exitcond, label %for.cond.cleanup.loopexit, label %scalar.body
}

; void recurrence_3(short *a, double *b, int n, float f, short p) {
;   b[0] = (double)a[0] - f * (double)p;
;   for (int i = 1; i < n; i++)
;     b[i] = (double)a[i] - f * (double)a[i - 1];
; }
;
; Check also that the casts were not moved needlessly.
;
;
define void @recurrence_3(i16* readonly noalias %a, double* noalias %b, i32 %n, float %f, i16 %p) {
; UNROLL-NO-IC-LABEL: @recurrence_3(
; UNROLL-NO-IC-NEXT:  entry:
; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A:%.*]], align 2
; UNROLL-NO-IC-NEXT:    [[CONV:%.*]] = sitofp i16 [[TMP0]] to double
; UNROLL-NO-IC-NEXT:    [[CONV1:%.*]] = fpext float [[F:%.*]] to double
; UNROLL-NO-IC-NEXT:    [[CONV2:%.*]] = sitofp i16 [[P:%.*]] to double
; UNROLL-NO-IC-NEXT:    [[MUL:%.*]] = fmul fast double [[CONV2]], [[CONV1]]
; UNROLL-NO-IC-NEXT:    [[SUB:%.*]] = fsub fast double [[CONV]], [[MUL]]
; UNROLL-NO-IC-NEXT:    store double [[SUB]], double* [[B:%.*]], align 8
; UNROLL-NO-IC-NEXT:    [[CMP25:%.*]] = icmp sgt i32 [[N:%.*]], 1
; UNROLL-NO-IC-NEXT:    br i1 [[CMP25]], label [[FOR_PREHEADER:%.*]], label [[FOR_END:%.*]]
; UNROLL-NO-IC:       for.preheader:
; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = add i32 [[N]], -2
; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add nuw nsw i64 [[TMP2]], 1
; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP3]], 8
; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC:       vector.ph:
; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP3]], 8
; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP3]], [[N_MOD_VF]]
; UNROLL-NO-IC-NEXT:    [[IND_END:%.*]] = add i64 1, [[N_VEC]]
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i16> poison, i16 [[TMP0]], i32 3
; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x double> poison, double [[CONV1]], i32 0
; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x double> [[BROADCAST_SPLATINSERT]], <4 x double> poison, <4 x i32> zeroinitializer
; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <4 x double> poison, double [[CONV1]], i32 0
; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLAT3:%.*]] = shufflevector <4 x double> [[BROADCAST_SPLATINSERT2]], <4 x double> poison, <4 x i32> zeroinitializer
; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC:       vector.body:
; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD1:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = add i64 [[OFFSET_IDX]], 0
; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = add i64 [[OFFSET_IDX]], 4
; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP4]]
; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP5]]
; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i16, i16* [[TMP6]], i32 0
; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = bitcast i16* [[TMP8]] to <4 x i16>*
; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, <4 x i16>* [[TMP9]], align 2
; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i16, i16* [[TMP6]], i32 4
; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = bitcast i16* [[TMP10]] to <4 x i16>*
; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD1]] = load <4 x i16>, <4 x i16>* [[TMP11]], align 2
; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = shufflevector <4 x i16> [[WIDE_LOAD]], <4 x i16> [[WIDE_LOAD1]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = sitofp <4 x i16> [[WIDE_LOAD]] to <4 x double>
; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = sitofp <4 x i16> [[WIDE_LOAD1]] to <4 x double>
; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = sitofp <4 x i16> [[TMP12]] to <4 x double>
; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = sitofp <4 x i16> [[TMP13]] to <4 x double>
; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = fmul fast <4 x double> [[TMP16]], [[BROADCAST_SPLAT]]
; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = fmul fast <4 x double> [[TMP17]], [[BROADCAST_SPLAT3]]
; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = fsub fast <4 x double> [[TMP14]], [[TMP18]]
; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = fsub fast <4 x double> [[TMP15]], [[TMP19]]
; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = getelementptr inbounds double, double* [[B]], i64 [[TMP4]]
; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = getelementptr inbounds double, double* [[B]], i64 [[TMP5]]
; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = getelementptr inbounds double, double* [[TMP22]], i32 0
; UNROLL-NO-IC-NEXT:    [[TMP25:%.*]] = bitcast double* [[TMP24]] to <4 x double>*
; UNROLL-NO-IC-NEXT:    store <4 x double> [[TMP20]], <4 x double>* [[TMP25]], align 8
; UNROLL-NO-IC-NEXT:    [[TMP26:%.*]] = getelementptr inbounds double, double* [[TMP22]], i32 4
; UNROLL-NO-IC-NEXT:    [[TMP27:%.*]] = bitcast double* [[TMP26]] to <4 x double>*
; UNROLL-NO-IC-NEXT:    store <4 x double> [[TMP21]], <4 x double>* [[TMP27]], align 8
; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
; UNROLL-NO-IC-NEXT:    [[TMP28:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-IC-NEXT:    br i1 [[TMP28]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
; UNROLL-NO-IC:       middle.block:
; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP3]], [[N_VEC]]
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[WIDE_LOAD1]], i32 3
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[WIDE_LOAD1]], i32 2
; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC:       scalar.ph:
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[TMP0]], [[FOR_PREHEADER]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 1, [[FOR_PREHEADER]] ]
; UNROLL-NO-IC-NEXT:    br label [[SCALAR_BODY:%.*]]
; UNROLL-NO-IC:       scalar.body:
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP29:%.*]], [[SCALAR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[ADVARS_IV:%.*]] = phi i64 [ [[ADVARS_IV_NEXT:%.*]], [[SCALAR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; UNROLL-NO-IC-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[ADVARS_IV]]
; UNROLL-NO-IC-NEXT:    [[TMP29]] = load i16, i16* [[ARRAYIDX5]], align 2
; UNROLL-NO-IC-NEXT:    [[CONV6:%.*]] = sitofp i16 [[TMP29]] to double
; UNROLL-NO-IC-NEXT:    [[CONV11:%.*]] = sitofp i16 [[SCALAR_RECUR]] to double
; UNROLL-NO-IC-NEXT:    [[MUL12:%.*]] = fmul fast double [[CONV11]], [[CONV1]]
; UNROLL-NO-IC-NEXT:    [[SUB13:%.*]] = fsub fast double [[CONV6]], [[MUL12]]
; UNROLL-NO-IC-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[B]], i64 [[ADVARS_IV]]
; UNROLL-NO-IC-NEXT:    store double [[SUB13]], double* [[ARRAYIDX15]], align 8
; UNROLL-NO-IC-NEXT:    [[ADVARS_IV_NEXT]] = add nuw nsw i64 [[ADVARS_IV]], 1
; UNROLL-NO-IC-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[ADVARS_IV_NEXT]] to i32
; UNROLL-NO-IC-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
; UNROLL-NO-IC-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[SCALAR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
; UNROLL-NO-IC:       for.end.loopexit:
; UNROLL-NO-IC-NEXT:    br label [[FOR_END]]
; UNROLL-NO-IC:       for.end:
; UNROLL-NO-IC-NEXT:    ret void
;
; UNROLL-NO-VF-LABEL: @recurrence_3(
; UNROLL-NO-VF-NEXT:  entry:
; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A:%.*]], align 2
; UNROLL-NO-VF-NEXT:    [[CONV:%.*]] = sitofp i16 [[TMP0]] to double
; UNROLL-NO-VF-NEXT:    [[CONV1:%.*]] = fpext float [[F:%.*]] to double
; UNROLL-NO-VF-NEXT:    [[CONV2:%.*]] = sitofp i16 [[P:%.*]] to double
; UNROLL-NO-VF-NEXT:    [[MUL:%.*]] = fmul fast double [[CONV2]], [[CONV1]]
; UNROLL-NO-VF-NEXT:    [[SUB:%.*]] = fsub fast double [[CONV]], [[MUL]]
; UNROLL-NO-VF-NEXT:    store double [[SUB]], double* [[B:%.*]], align 8
; UNROLL-NO-VF-NEXT:    [[CMP25:%.*]] = icmp sgt i32 [[N:%.*]], 1
; UNROLL-NO-VF-NEXT:    br i1 [[CMP25]], label [[FOR_PREHEADER:%.*]], label [[FOR_END:%.*]]
; UNROLL-NO-VF:       for.preheader:
; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = add i32 [[N]], -2
; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
; UNROLL-NO-VF-NEXT:    [[TMP3:%.*]] = add nuw nsw i64 [[TMP2]], 1
; UNROLL-NO-VF-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP3]], 2
; UNROLL-NO-VF-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-VF:       vector.ph:
; UNROLL-NO-VF-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP3]], 2
; UNROLL-NO-VF-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP3]], [[N_MOD_VF]]
; UNROLL-NO-VF-NEXT:    [[IND_END:%.*]] = add i64 1, [[N_VEC]]
; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-VF:       vector.body:
; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi i16 [ [[TMP0]], [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
; UNROLL-NO-VF-NEXT:    [[INDUCTION:%.*]] = add i64 [[OFFSET_IDX]], 0
; UNROLL-NO-VF-NEXT:    [[INDUCTION1:%.*]] = add i64 [[OFFSET_IDX]], 1
; UNROLL-NO-VF-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[INDUCTION]]
; UNROLL-NO-VF-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[INDUCTION1]]
; UNROLL-NO-VF-NEXT:    [[TMP6:%.*]] = load i16, i16* [[TMP4]], align 2
; UNROLL-NO-VF-NEXT:    [[TMP7]] = load i16, i16* [[TMP5]], align 2
; UNROLL-NO-VF-NEXT:    [[TMP8:%.*]] = sitofp i16 [[TMP6]] to double
; UNROLL-NO-VF-NEXT:    [[TMP9:%.*]] = sitofp i16 [[TMP7]] to double
; UNROLL-NO-VF-NEXT:    [[TMP10:%.*]] = sitofp i16 [[VECTOR_RECUR]] to double
; UNROLL-NO-VF-NEXT:    [[TMP11:%.*]] = sitofp i16 [[TMP6]] to double
; UNROLL-NO-VF-NEXT:    [[TMP12:%.*]] = fmul fast double [[TMP10]], [[CONV1]]
; UNROLL-NO-VF-NEXT:    [[TMP13:%.*]] = fmul fast double [[TMP11]], [[CONV1]]
; UNROLL-NO-VF-NEXT:    [[TMP14:%.*]] = fsub fast double [[TMP8]], [[TMP12]]
; UNROLL-NO-VF-NEXT:    [[TMP15:%.*]] = fsub fast double [[TMP9]], [[TMP13]]
; UNROLL-NO-VF-NEXT:    [[TMP16:%.*]] = getelementptr inbounds double, double* [[B]], i64 [[INDUCTION]]
; UNROLL-NO-VF-NEXT:    [[TMP17:%.*]] = getelementptr inbounds double, double* [[B]], i64 [[INDUCTION1]]
; UNROLL-NO-VF-NEXT:    store double [[TMP14]], double* [[TMP16]], align 8
; UNROLL-NO-VF-NEXT:    store double [[TMP15]], double* [[TMP17]], align 8
; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; UNROLL-NO-VF-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-VF-NEXT:    br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
; UNROLL-NO-VF:       middle.block:
; UNROLL-NO-VF-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP3]], [[N_VEC]]
; UNROLL-NO-VF-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-VF:       scalar.ph:
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[TMP0]], [[FOR_PREHEADER]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 1, [[FOR_PREHEADER]] ]
; UNROLL-NO-VF-NEXT:    br label [[SCALAR_BODY:%.*]]
; UNROLL-NO-VF:       scalar.body:
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP19:%.*]], [[SCALAR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[ADVARS_IV:%.*]] = phi i64 [ [[ADVARS_IV_NEXT:%.*]], [[SCALAR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; UNROLL-NO-VF-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[ADVARS_IV]]
; UNROLL-NO-VF-NEXT:    [[TMP19]] = load i16, i16* [[ARRAYIDX5]], align 2
; UNROLL-NO-VF-NEXT:    [[CONV6:%.*]] = sitofp i16 [[TMP19]] to double
; UNROLL-NO-VF-NEXT:    [[CONV11:%.*]] = sitofp i16 [[SCALAR_RECUR]] to double
; UNROLL-NO-VF-NEXT:    [[MUL12:%.*]] = fmul fast double [[CONV11]], [[CONV1]]
; UNROLL-NO-VF-NEXT:    [[SUB13:%.*]] = fsub fast double [[CONV6]], [[MUL12]]
; UNROLL-NO-VF-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[B]], i64 [[ADVARS_IV]]
; UNROLL-NO-VF-NEXT:    store double [[SUB13]], double* [[ARRAYIDX15]], align 8
; UNROLL-NO-VF-NEXT:    [[ADVARS_IV_NEXT]] = add nuw nsw i64 [[ADVARS_IV]], 1
; UNROLL-NO-VF-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[ADVARS_IV_NEXT]] to i32
; UNROLL-NO-VF-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
; UNROLL-NO-VF-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[SCALAR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
; UNROLL-NO-VF:       for.end.loopexit:
; UNROLL-NO-VF-NEXT:    br label [[FOR_END]]
; UNROLL-NO-VF:       for.end:
; UNROLL-NO-VF-NEXT:    ret void
;
; SINK-AFTER-LABEL: @recurrence_3(
; SINK-AFTER-NEXT:  entry:
; SINK-AFTER-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A:%.*]], align 2
; SINK-AFTER-NEXT:    [[CONV:%.*]] = sitofp i16 [[TMP0]] to double
; SINK-AFTER-NEXT:    [[CONV1:%.*]] = fpext float [[F:%.*]] to double
; SINK-AFTER-NEXT:    [[CONV2:%.*]] = sitofp i16 [[P:%.*]] to double
; SINK-AFTER-NEXT:    [[MUL:%.*]] = fmul fast double [[CONV2]], [[CONV1]]
; SINK-AFTER-NEXT:    [[SUB:%.*]] = fsub fast double [[CONV]], [[MUL]]
; SINK-AFTER-NEXT:    store double [[SUB]], double* [[B:%.*]], align 8
; SINK-AFTER-NEXT:    [[CMP25:%.*]] = icmp sgt i32 [[N:%.*]], 1
; SINK-AFTER-NEXT:    br i1 [[CMP25]], label [[FOR_PREHEADER:%.*]], label [[FOR_END:%.*]]
; SINK-AFTER:       for.preheader:
; SINK-AFTER-NEXT:    [[TMP1:%.*]] = add i32 [[N]], -2
; SINK-AFTER-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
; SINK-AFTER-NEXT:    [[TMP3:%.*]] = add nuw nsw i64 [[TMP2]], 1
; SINK-AFTER-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP3]], 4
; SINK-AFTER-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; SINK-AFTER:       vector.ph:
; SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP3]], 4
; SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP3]], [[N_MOD_VF]]
; SINK-AFTER-NEXT:    [[IND_END:%.*]] = add i64 1, [[N_VEC]]
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i16> poison, i16 [[TMP0]], i32 3
; SINK-AFTER-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x double> poison, double [[CONV1]], i32 0
; SINK-AFTER-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x double> [[BROADCAST_SPLATINSERT]], <4 x double> poison, <4 x i32> zeroinitializer
; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
; SINK-AFTER:       vector.body:
; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
; SINK-AFTER-NEXT:    [[TMP4:%.*]] = add i64 [[OFFSET_IDX]], 0
; SINK-AFTER-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP4]]
; SINK-AFTER-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, i16* [[TMP5]], i32 0
; SINK-AFTER-NEXT:    [[TMP7:%.*]] = bitcast i16* [[TMP6]] to <4 x i16>*
; SINK-AFTER-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, <4 x i16>* [[TMP7]], align 2
; SINK-AFTER-NEXT:    [[TMP8:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; SINK-AFTER-NEXT:    [[TMP9:%.*]] = sitofp <4 x i16> [[WIDE_LOAD]] to <4 x double>
; SINK-AFTER-NEXT:    [[TMP10:%.*]] = sitofp <4 x i16> [[TMP8]] to <4 x double>
; SINK-AFTER-NEXT:    [[TMP11:%.*]] = fmul fast <4 x double> [[TMP10]], [[BROADCAST_SPLAT]]
; SINK-AFTER-NEXT:    [[TMP12:%.*]] = fsub fast <4 x double> [[TMP9]], [[TMP11]]
; SINK-AFTER-NEXT:    [[TMP13:%.*]] = getelementptr inbounds double, double* [[B]], i64 [[TMP4]]
; SINK-AFTER-NEXT:    [[TMP14:%.*]] = getelementptr inbounds double, double* [[TMP13]], i32 0
; SINK-AFTER-NEXT:    [[TMP15:%.*]] = bitcast double* [[TMP14]] to <4 x double>*
; SINK-AFTER-NEXT:    store <4 x double> [[TMP12]], <4 x double>* [[TMP15]], align 8
; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; SINK-AFTER-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; SINK-AFTER-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
; SINK-AFTER:       middle.block:
; SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP3]], [[N_VEC]]
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 3
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 2
; SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
; SINK-AFTER:       scalar.ph:
; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[TMP0]], [[FOR_PREHEADER]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 1, [[FOR_PREHEADER]] ]
; SINK-AFTER-NEXT:    br label [[SCALAR_BODY:%.*]]
; SINK-AFTER:       scalar.body:
; SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP17:%.*]], [[SCALAR_BODY]] ]
; SINK-AFTER-NEXT:    [[ADVARS_IV:%.*]] = phi i64 [ [[ADVARS_IV_NEXT:%.*]], [[SCALAR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; SINK-AFTER-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[ADVARS_IV]]
; SINK-AFTER-NEXT:    [[TMP17]] = load i16, i16* [[ARRAYIDX5]], align 2
; SINK-AFTER-NEXT:    [[CONV6:%.*]] = sitofp i16 [[TMP17]] to double
; SINK-AFTER-NEXT:    [[CONV11:%.*]] = sitofp i16 [[SCALAR_RECUR]] to double
; SINK-AFTER-NEXT:    [[MUL12:%.*]] = fmul fast double [[CONV11]], [[CONV1]]
; SINK-AFTER-NEXT:    [[SUB13:%.*]] = fsub fast double [[CONV6]], [[MUL12]]
; SINK-AFTER-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[B]], i64 [[ADVARS_IV]]
; SINK-AFTER-NEXT:    store double [[SUB13]], double* [[ARRAYIDX15]], align 8
; SINK-AFTER-NEXT:    [[ADVARS_IV_NEXT]] = add nuw nsw i64 [[ADVARS_IV]], 1
; SINK-AFTER-NEXT:    [[LFTR_WIDEIV:%.*]] = trunc i64 [[ADVARS_IV_NEXT]] to i32
; SINK-AFTER-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[N]]
; SINK-AFTER-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[SCALAR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
; SINK-AFTER:       for.end.loopexit:
; SINK-AFTER-NEXT:    br label [[FOR_END]]
; SINK-AFTER:       for.end:
; SINK-AFTER-NEXT:    ret void
;
entry:
  %0 = load i16, i16* %a, align 2
  %conv = sitofp i16 %0 to double
  %conv1 = fpext float %f to double
  %conv2 = sitofp i16 %p to double
  %mul = fmul fast double %conv2, %conv1
  %sub = fsub fast double %conv, %mul
  store double %sub, double* %b, align 8
  %cmp25 = icmp sgt i32 %n, 1
  br i1 %cmp25, label %for.preheader, label %for.end

for.preheader:
  br label %scalar.body

scalar.body:
  %1 = phi i16 [ %0, %for.preheader ], [ %2, %scalar.body ]
  %advars.iv = phi i64 [ %advars.iv.next, %scalar.body ], [ 1, %for.preheader ]
  %arrayidx5 = getelementptr inbounds i16, i16* %a, i64 %advars.iv
  %2 = load i16, i16* %arrayidx5, align 2
  %conv6 = sitofp i16 %2 to double
  %conv11 = sitofp i16 %1 to double
  %mul12 = fmul fast double %conv11, %conv1
  %sub13 = fsub fast double %conv6, %mul12
  %arrayidx15 = getelementptr inbounds double, double* %b, i64 %advars.iv
  store double %sub13, double* %arrayidx15, align 8
  %advars.iv.next = add nuw nsw i64 %advars.iv, 1
  %lftr.wideiv = trunc i64 %advars.iv.next to i32
  %exitcond = icmp eq i32 %lftr.wideiv, %n
  br i1 %exitcond, label %for.end.loopexit, label %scalar.body

for.end.loopexit:
  br label %for.end

for.end:
  ret void
}

; void PR26734(short *a, int *b, int *c, int d, short *e) {
;   for (; d != 21; d++) {
;     *b &= *c;
;     *e = *a - 6;
;     *c = *e;
;   }
; }
;
;
define void @PR26734(i16* %a, i32* %b, i32* %c, i32 %d, i16* %e) {
; UNROLL-NO-IC-LABEL: @PR26734(
; UNROLL-NO-IC-NEXT:  entry:
; UNROLL-NO-IC-NEXT:    [[CMP4:%.*]] = icmp eq i32 [[D:%.*]], 21
; UNROLL-NO-IC-NEXT:    br i1 [[CMP4]], label [[ENTRY_FOR_END_CRIT_EDGE:%.*]], label [[FOR_BODY_LR_PH:%.*]]
; UNROLL-NO-IC:       entry.for.end_crit_edge:
; UNROLL-NO-IC-NEXT:    [[DOTPRE:%.*]] = load i32, i32* [[B:%.*]], align 4
; UNROLL-NO-IC-NEXT:    br label [[FOR_END:%.*]]
; UNROLL-NO-IC:       for.body.lr.ph:
; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A:%.*]], align 2
; UNROLL-NO-IC-NEXT:    [[SUB:%.*]] = add i16 [[TMP0]], -6
; UNROLL-NO-IC-NEXT:    [[CONV2:%.*]] = sext i16 [[SUB]] to i32
; UNROLL-NO-IC-NEXT:    [[C_PROMOTED:%.*]] = load i32, i32* [[C:%.*]], align 4
; UNROLL-NO-IC-NEXT:    [[B_PROMOTED:%.*]] = load i32, i32* [[B]], align 4
; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY:%.*]]
; UNROLL-NO-IC:       for.body:
; UNROLL-NO-IC-NEXT:    [[INC7:%.*]] = phi i32 [ [[D]], [[FOR_BODY_LR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[AND6:%.*]] = phi i32 [ [[B_PROMOTED]], [[FOR_BODY_LR_PH]] ], [ [[AND:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[CONV25:%.*]] = phi i32 [ [[C_PROMOTED]], [[FOR_BODY_LR_PH]] ], [ [[CONV2]], [[FOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[AND]] = and i32 [[AND6]], [[CONV25]]
; UNROLL-NO-IC-NEXT:    [[INC]] = add nsw i32 [[INC7]], 1
; UNROLL-NO-IC-NEXT:    [[CMP:%.*]] = icmp eq i32 [[INC]], 21
; UNROLL-NO-IC-NEXT:    br i1 [[CMP]], label [[FOR_COND_FOR_END_CRIT_EDGE:%.*]], label [[FOR_BODY]]
; UNROLL-NO-IC:       for.cond.for.end_crit_edge:
; UNROLL-NO-IC-NEXT:    [[AND_LCSSA:%.*]] = phi i32 [ [[AND]], [[FOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    store i32 [[CONV2]], i32* [[C]], align 4
; UNROLL-NO-IC-NEXT:    store i32 [[AND_LCSSA]], i32* [[B]], align 4
; UNROLL-NO-IC-NEXT:    store i16 [[SUB]], i16* [[E:%.*]], align 2
; UNROLL-NO-IC-NEXT:    br label [[FOR_END]]
; UNROLL-NO-IC:       for.end:
; UNROLL-NO-IC-NEXT:    ret void
;
; UNROLL-NO-VF-LABEL: @PR26734(
; UNROLL-NO-VF-NEXT:  entry:
; UNROLL-NO-VF-NEXT:    [[CMP4:%.*]] = icmp eq i32 [[D:%.*]], 21
; UNROLL-NO-VF-NEXT:    br i1 [[CMP4]], label [[ENTRY_FOR_END_CRIT_EDGE:%.*]], label [[FOR_BODY_LR_PH:%.*]]
; UNROLL-NO-VF:       entry.for.end_crit_edge:
; UNROLL-NO-VF-NEXT:    [[DOTPRE:%.*]] = load i32, i32* [[B:%.*]], align 4
; UNROLL-NO-VF-NEXT:    br label [[FOR_END:%.*]]
; UNROLL-NO-VF:       for.body.lr.ph:
; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A:%.*]], align 2
; UNROLL-NO-VF-NEXT:    [[SUB:%.*]] = add i16 [[TMP0]], -6
; UNROLL-NO-VF-NEXT:    [[CONV2:%.*]] = sext i16 [[SUB]] to i32
; UNROLL-NO-VF-NEXT:    [[C_PROMOTED:%.*]] = load i32, i32* [[C:%.*]], align 4
; UNROLL-NO-VF-NEXT:    [[B_PROMOTED:%.*]] = load i32, i32* [[B]], align 4
; UNROLL-NO-VF-NEXT:    br label [[FOR_BODY:%.*]]
; UNROLL-NO-VF:       for.body:
; UNROLL-NO-VF-NEXT:    [[INC7:%.*]] = phi i32 [ [[D]], [[FOR_BODY_LR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[AND6:%.*]] = phi i32 [ [[B_PROMOTED]], [[FOR_BODY_LR_PH]] ], [ [[AND:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[CONV25:%.*]] = phi i32 [ [[C_PROMOTED]], [[FOR_BODY_LR_PH]] ], [ [[CONV2]], [[FOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[AND]] = and i32 [[AND6]], [[CONV25]]
; UNROLL-NO-VF-NEXT:    [[INC]] = add nsw i32 [[INC7]], 1
; UNROLL-NO-VF-NEXT:    [[CMP:%.*]] = icmp eq i32 [[INC]], 21
; UNROLL-NO-VF-NEXT:    br i1 [[CMP]], label [[FOR_COND_FOR_END_CRIT_EDGE:%.*]], label [[FOR_BODY]]
; UNROLL-NO-VF:       for.cond.for.end_crit_edge:
; UNROLL-NO-VF-NEXT:    [[AND_LCSSA:%.*]] = phi i32 [ [[AND]], [[FOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    store i32 [[CONV2]], i32* [[C]], align 4
; UNROLL-NO-VF-NEXT:    store i32 [[AND_LCSSA]], i32* [[B]], align 4
; UNROLL-NO-VF-NEXT:    store i16 [[SUB]], i16* [[E:%.*]], align 2
; UNROLL-NO-VF-NEXT:    br label [[FOR_END]]
; UNROLL-NO-VF:       for.end:
; UNROLL-NO-VF-NEXT:    ret void
;
; SINK-AFTER-LABEL: @PR26734(
; SINK-AFTER-NEXT:  entry:
; SINK-AFTER-NEXT:    [[CMP4:%.*]] = icmp eq i32 [[D:%.*]], 21
; SINK-AFTER-NEXT:    br i1 [[CMP4]], label [[ENTRY_FOR_END_CRIT_EDGE:%.*]], label [[FOR_BODY_LR_PH:%.*]]
; SINK-AFTER:       entry.for.end_crit_edge:
; SINK-AFTER-NEXT:    [[DOTPRE:%.*]] = load i32, i32* [[B:%.*]], align 4
; SINK-AFTER-NEXT:    br label [[FOR_END:%.*]]
; SINK-AFTER:       for.body.lr.ph:
; SINK-AFTER-NEXT:    [[TMP0:%.*]] = load i16, i16* [[A:%.*]], align 2
; SINK-AFTER-NEXT:    [[SUB:%.*]] = add i16 [[TMP0]], -6
; SINK-AFTER-NEXT:    [[CONV2:%.*]] = sext i16 [[SUB]] to i32
; SINK-AFTER-NEXT:    [[C_PROMOTED:%.*]] = load i32, i32* [[C:%.*]], align 4
; SINK-AFTER-NEXT:    [[B_PROMOTED:%.*]] = load i32, i32* [[B]], align 4
; SINK-AFTER-NEXT:    br label [[FOR_BODY:%.*]]
; SINK-AFTER:       for.body:
; SINK-AFTER-NEXT:    [[INC7:%.*]] = phi i32 [ [[D]], [[FOR_BODY_LR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
; SINK-AFTER-NEXT:    [[AND6:%.*]] = phi i32 [ [[B_PROMOTED]], [[FOR_BODY_LR_PH]] ], [ [[AND:%.*]], [[FOR_BODY]] ]
; SINK-AFTER-NEXT:    [[CONV25:%.*]] = phi i32 [ [[C_PROMOTED]], [[FOR_BODY_LR_PH]] ], [ [[CONV2]], [[FOR_BODY]] ]
; SINK-AFTER-NEXT:    [[AND]] = and i32 [[AND6]], [[CONV25]]
; SINK-AFTER-NEXT:    [[INC]] = add nsw i32 [[INC7]], 1
; SINK-AFTER-NEXT:    [[CMP:%.*]] = icmp eq i32 [[INC]], 21
; SINK-AFTER-NEXT:    br i1 [[CMP]], label [[FOR_COND_FOR_END_CRIT_EDGE:%.*]], label [[FOR_BODY]]
; SINK-AFTER:       for.cond.for.end_crit_edge:
; SINK-AFTER-NEXT:    [[AND_LCSSA:%.*]] = phi i32 [ [[AND]], [[FOR_BODY]] ]
; SINK-AFTER-NEXT:    store i32 [[CONV2]], i32* [[C]], align 4
; SINK-AFTER-NEXT:    store i32 [[AND_LCSSA]], i32* [[B]], align 4
; SINK-AFTER-NEXT:    store i16 [[SUB]], i16* [[E:%.*]], align 2
; SINK-AFTER-NEXT:    br label [[FOR_END]]
; SINK-AFTER:       for.end:
; SINK-AFTER-NEXT:    ret void
;
entry:
  %cmp4 = icmp eq i32 %d, 21
  br i1 %cmp4, label %entry.for.end_crit_edge, label %for.body.lr.ph

entry.for.end_crit_edge:
  %.pre = load i32, i32* %b, align 4
  br label %for.end

for.body.lr.ph:
  %0 = load i16, i16* %a, align 2
  %sub = add i16 %0, -6
  %conv2 = sext i16 %sub to i32
  %c.promoted = load i32, i32* %c, align 4
  %b.promoted = load i32, i32* %b, align 4
  br label %for.body

for.body:
  %inc7 = phi i32 [ %d, %for.body.lr.ph ], [ %inc, %for.body ]
  %and6 = phi i32 [ %b.promoted, %for.body.lr.ph ], [ %and, %for.body ]
  %conv25 = phi i32 [ %c.promoted, %for.body.lr.ph ], [ %conv2, %for.body ]
  %and = and i32 %and6, %conv25
  %inc = add nsw i32 %inc7, 1
  %cmp = icmp eq i32 %inc, 21
  br i1 %cmp, label %for.cond.for.end_crit_edge, label %for.body

for.cond.for.end_crit_edge:
  %and.lcssa = phi i32 [ %and, %for.body ]
  store i32 %conv2, i32* %c, align 4
  store i32 %and.lcssa, i32* %b, align 4
  store i16 %sub, i16* %e, align 2
  br label %for.end

for.end:
  ret void
}

; int PR27246() {
;   unsigned int e, n;
;   for (int i = 1; i < 49; ++i) {
;     for (int k = i; k > 1; --k)
;       e = k;
;     n = e;
;   }
;   return n;
; }
;
;
define i32 @PR27246() {
; UNROLL-NO-IC-LABEL: @PR27246(
; UNROLL-NO-IC-NEXT:  entry:
; UNROLL-NO-IC-NEXT:    br label [[FOR_COND1_PREHEADER:%.*]]
; UNROLL-NO-IC:       for.cond1.preheader:
; UNROLL-NO-IC-NEXT:    [[I_016:%.*]] = phi i32 [ 1, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_COND_CLEANUP3:%.*]] ]
; UNROLL-NO-IC-NEXT:    [[E_015:%.*]] = phi i32 [ poison, [[ENTRY]] ], [ [[E_1_LCSSA:%.*]], [[FOR_COND_CLEANUP3]] ]
; UNROLL-NO-IC-NEXT:    br label [[FOR_COND1:%.*]]
; UNROLL-NO-IC:       for.cond.cleanup:
; UNROLL-NO-IC-NEXT:    [[E_1_LCSSA_LCSSA:%.*]] = phi i32 [ [[E_1_LCSSA]], [[FOR_COND_CLEANUP3]] ]
; UNROLL-NO-IC-NEXT:    ret i32 [[E_1_LCSSA_LCSSA]]
; UNROLL-NO-IC:       for.cond1:
; UNROLL-NO-IC-NEXT:    [[E_1:%.*]] = phi i32 [ [[K_0:%.*]], [[FOR_COND1]] ], [ [[E_015]], [[FOR_COND1_PREHEADER]] ]
; UNROLL-NO-IC-NEXT:    [[K_0]] = phi i32 [ [[DEC:%.*]], [[FOR_COND1]] ], [ [[I_016]], [[FOR_COND1_PREHEADER]] ]
; UNROLL-NO-IC-NEXT:    [[CMP2:%.*]] = icmp sgt i32 [[K_0]], 1
; UNROLL-NO-IC-NEXT:    [[DEC]] = add nsw i32 [[K_0]], -1
; UNROLL-NO-IC-NEXT:    br i1 [[CMP2]], label [[FOR_COND1]], label [[FOR_COND_CLEANUP3]]
; UNROLL-NO-IC:       for.cond.cleanup3:
; UNROLL-NO-IC-NEXT:    [[E_1_LCSSA]] = phi i32 [ [[E_1]], [[FOR_COND1]] ]
; UNROLL-NO-IC-NEXT:    [[INC]] = add nuw nsw i32 [[I_016]], 1
; UNROLL-NO-IC-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INC]], 49
; UNROLL-NO-IC-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP:%.*]], label [[FOR_COND1_PREHEADER]]
;
; UNROLL-NO-VF-LABEL: @PR27246(
; UNROLL-NO-VF-NEXT:  entry:
; UNROLL-NO-VF-NEXT:    br label [[FOR_COND1_PREHEADER:%.*]]
; UNROLL-NO-VF:       for.cond1.preheader:
; UNROLL-NO-VF-NEXT:    [[I_016:%.*]] = phi i32 [ 1, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_COND_CLEANUP3:%.*]] ]
; UNROLL-NO-VF-NEXT:    [[E_015:%.*]] = phi i32 [ poison, [[ENTRY]] ], [ [[E_1_LCSSA:%.*]], [[FOR_COND_CLEANUP3]] ]
; UNROLL-NO-VF-NEXT:    br label [[FOR_COND1:%.*]]
; UNROLL-NO-VF:       for.cond.cleanup:
; UNROLL-NO-VF-NEXT:    [[E_1_LCSSA_LCSSA:%.*]] = phi i32 [ [[E_1_LCSSA]], [[FOR_COND_CLEANUP3]] ]
; UNROLL-NO-VF-NEXT:    ret i32 [[E_1_LCSSA_LCSSA]]
; UNROLL-NO-VF:       for.cond1:
; UNROLL-NO-VF-NEXT:    [[E_1:%.*]] = phi i32 [ [[K_0:%.*]], [[FOR_COND1]] ], [ [[E_015]], [[FOR_COND1_PREHEADER]] ]
; UNROLL-NO-VF-NEXT:    [[K_0]] = phi i32 [ [[DEC:%.*]], [[FOR_COND1]] ], [ [[I_016]], [[FOR_COND1_PREHEADER]] ]
; UNROLL-NO-VF-NEXT:    [[CMP2:%.*]] = icmp sgt i32 [[K_0]], 1
; UNROLL-NO-VF-NEXT:    [[DEC]] = add nsw i32 [[K_0]], -1
; UNROLL-NO-VF-NEXT:    br i1 [[CMP2]], label [[FOR_COND1]], label [[FOR_COND_CLEANUP3]]
; UNROLL-NO-VF:       for.cond.cleanup3:
; UNROLL-NO-VF-NEXT:    [[E_1_LCSSA]] = phi i32 [ [[E_1]], [[FOR_COND1]] ]
; UNROLL-NO-VF-NEXT:    [[INC]] = add nuw nsw i32 [[I_016]], 1
; UNROLL-NO-VF-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INC]], 49
; UNROLL-NO-VF-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP:%.*]], label [[FOR_COND1_PREHEADER]]
;
; SINK-AFTER-LABEL: @PR27246(
; SINK-AFTER-NEXT:  entry:
; SINK-AFTER-NEXT:    br label [[FOR_COND1_PREHEADER:%.*]]
; SINK-AFTER:       for.cond1.preheader:
; SINK-AFTER-NEXT:    [[I_016:%.*]] = phi i32 [ 1, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_COND_CLEANUP3:%.*]] ]
; SINK-AFTER-NEXT:    [[E_015:%.*]] = phi i32 [ poison, [[ENTRY]] ], [ [[E_1_LCSSA:%.*]], [[FOR_COND_CLEANUP3]] ]
; SINK-AFTER-NEXT:    br label [[FOR_COND1:%.*]]
; SINK-AFTER:       for.cond.cleanup:
; SINK-AFTER-NEXT:    [[E_1_LCSSA_LCSSA:%.*]] = phi i32 [ [[E_1_LCSSA]], [[FOR_COND_CLEANUP3]] ]
; SINK-AFTER-NEXT:    ret i32 [[E_1_LCSSA_LCSSA]]
; SINK-AFTER:       for.cond1:
; SINK-AFTER-NEXT:    [[E_1:%.*]] = phi i32 [ [[K_0:%.*]], [[FOR_COND1]] ], [ [[E_015]], [[FOR_COND1_PREHEADER]] ]
; SINK-AFTER-NEXT:    [[K_0]] = phi i32 [ [[DEC:%.*]], [[FOR_COND1]] ], [ [[I_016]], [[FOR_COND1_PREHEADER]] ]
; SINK-AFTER-NEXT:    [[CMP2:%.*]] = icmp sgt i32 [[K_0]], 1
; SINK-AFTER-NEXT:    [[DEC]] = add nsw i32 [[K_0]], -1
; SINK-AFTER-NEXT:    br i1 [[CMP2]], label [[FOR_COND1]], label [[FOR_COND_CLEANUP3]]
; SINK-AFTER:       for.cond.cleanup3:
; SINK-AFTER-NEXT:    [[E_1_LCSSA]] = phi i32 [ [[E_1]], [[FOR_COND1]] ]
; SINK-AFTER-NEXT:    [[INC]] = add nuw nsw i32 [[I_016]], 1
; SINK-AFTER-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INC]], 49
; SINK-AFTER-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP:%.*]], label [[FOR_COND1_PREHEADER]]
;
entry:
  br label %for.cond1.preheader

for.cond1.preheader:
  %i.016 = phi i32 [ 1, %entry ], [ %inc, %for.cond.cleanup3 ]
  %e.015 = phi i32 [ poison, %entry ], [ %e.1.lcssa, %for.cond.cleanup3 ]
  br label %for.cond1

for.cond.cleanup:
  %e.1.lcssa.lcssa = phi i32 [ %e.1.lcssa, %for.cond.cleanup3 ]
  ret i32 %e.1.lcssa.lcssa

for.cond1:
  %e.1 = phi i32 [ %k.0, %for.cond1 ], [ %e.015, %for.cond1.preheader ]
  %k.0 = phi i32 [ %dec, %for.cond1 ], [ %i.016, %for.cond1.preheader ]
  %cmp2 = icmp sgt i32 %k.0, 1
  %dec = add nsw i32 %k.0, -1
  br i1 %cmp2, label %for.cond1, label %for.cond.cleanup3

for.cond.cleanup3:
  %e.1.lcssa = phi i32 [ %e.1, %for.cond1 ]
  %inc = add nuw nsw i32 %i.016, 1
  %exitcond = icmp eq i32 %inc, 49
  br i1 %exitcond, label %for.cond.cleanup, label %for.cond1.preheader
}

;
define void @PR30183(i32 %pre_load, i32* %a, i32* %b, i64 %n) {
; UNROLL-NO-IC-LABEL: @PR30183(
; UNROLL-NO-IC-NEXT:  entry:
; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i64 [[N:%.*]], -2
; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = lshr i64 [[TMP0]], 1
; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add nuw i64 [[TMP1]], 1
; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 8
; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC:       vector.ph:
; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 8
; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
; UNROLL-NO-IC-NEXT:    [[IND_END:%.*]] = mul i64 [[N_VEC]], 2
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i32> poison, i32 [[PRE_LOAD:%.*]], i32 3
; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC:       vector.body:
; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[TMP42:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 2
; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add i64 [[OFFSET_IDX]], 0
; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = add i64 [[OFFSET_IDX]], 2
; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = add i64 [[OFFSET_IDX]], 4
; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = add i64 [[OFFSET_IDX]], 6
; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = add i64 [[OFFSET_IDX]], 8
; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = add i64 [[OFFSET_IDX]], 10
; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = add i64 [[OFFSET_IDX]], 12
; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = add i64 [[OFFSET_IDX]], 14
; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = add nuw nsw i64 [[TMP3]], 2
; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = add nuw nsw i64 [[TMP4]], 2
; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = add nuw nsw i64 [[TMP5]], 2
; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = add nuw nsw i64 [[TMP6]], 2
; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = add nuw nsw i64 [[TMP7]], 2
; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = add nuw nsw i64 [[TMP8]], 2
; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = add nuw nsw i64 [[TMP9]], 2
; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = add nuw nsw i64 [[TMP10]], 2
; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP11]]
; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP12]]
; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP13]]
; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP14]]
; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP15]]
; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP16]]
; UNROLL-NO-IC-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP17]]
; UNROLL-NO-IC-NEXT:    [[TMP26:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP18]]
; UNROLL-NO-IC-NEXT:    [[TMP27:%.*]] = load i32, i32* [[TMP19]], align 4
; UNROLL-NO-IC-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP20]], align 4
; UNROLL-NO-IC-NEXT:    [[TMP29:%.*]] = load i32, i32* [[TMP21]], align 4
; UNROLL-NO-IC-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP22]], align 4
; UNROLL-NO-IC-NEXT:    [[TMP31:%.*]] = insertelement <4 x i32> poison, i32 [[TMP27]], i32 0
; UNROLL-NO-IC-NEXT:    [[TMP32:%.*]] = insertelement <4 x i32> [[TMP31]], i32 [[TMP28]], i32 1
; UNROLL-NO-IC-NEXT:    [[TMP33:%.*]] = insertelement <4 x i32> [[TMP32]], i32 [[TMP29]], i32 2
; UNROLL-NO-IC-NEXT:    [[TMP34:%.*]] = insertelement <4 x i32> [[TMP33]], i32 [[TMP30]], i32 3
; UNROLL-NO-IC-NEXT:    [[TMP35:%.*]] = load i32, i32* [[TMP23]], align 4
; UNROLL-NO-IC-NEXT:    [[TMP36:%.*]] = load i32, i32* [[TMP24]], align 4
; UNROLL-NO-IC-NEXT:    [[TMP37:%.*]] = load i32, i32* [[TMP25]], align 4
; UNROLL-NO-IC-NEXT:    [[TMP38:%.*]] = load i32, i32* [[TMP26]], align 4
; UNROLL-NO-IC-NEXT:    [[TMP39:%.*]] = insertelement <4 x i32> poison, i32 [[TMP35]], i32 0
; UNROLL-NO-IC-NEXT:    [[TMP40:%.*]] = insertelement <4 x i32> [[TMP39]], i32 [[TMP36]], i32 1
; UNROLL-NO-IC-NEXT:    [[TMP41:%.*]] = insertelement <4 x i32> [[TMP40]], i32 [[TMP37]], i32 2
; UNROLL-NO-IC-NEXT:    [[TMP42]] = insertelement <4 x i32> [[TMP41]], i32 [[TMP38]], i32 3
; UNROLL-NO-IC-NEXT:    [[TMP43:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP34]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; UNROLL-NO-IC-NEXT:    [[TMP44:%.*]] = shufflevector <4 x i32> [[TMP34]], <4 x i32> [[TMP42]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
; UNROLL-NO-IC-NEXT:    [[TMP45:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-IC-NEXT:    br i1 [[TMP45]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
; UNROLL-NO-IC:       middle.block:
; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP42]], i32 3
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP42]], i32 2
; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC:       scalar.ph:
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[PRE_LOAD]], [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; UNROLL-NO-IC-NEXT:    br label [[SCALAR_BODY:%.*]]
; UNROLL-NO-IC:       scalar.body:
; UNROLL-NO-IC-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[SCALAR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[VAR2:%.*]], [[SCALAR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 2
; UNROLL-NO-IC-NEXT:    [[VAR1:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[I_NEXT]]
; UNROLL-NO-IC-NEXT:    [[VAR2]] = load i32, i32* [[VAR1]], align 4
; UNROLL-NO-IC-NEXT:    [[COND:%.*]] = icmp eq i64 [[I_NEXT]], [[N]]
; UNROLL-NO-IC-NEXT:    br i1 [[COND]], label [[FOR_END]], label [[SCALAR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
; UNROLL-NO-IC:       for.end:
; UNROLL-NO-IC-NEXT:    ret void
;
; UNROLL-NO-VF-LABEL: @PR30183(
; UNROLL-NO-VF-NEXT:  entry:
; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = add i64 [[N:%.*]], -2
; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = lshr i64 [[TMP0]], 1
; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = add nuw i64 [[TMP1]], 1
; UNROLL-NO-VF-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 2
; UNROLL-NO-VF-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-VF:       vector.ph:
; UNROLL-NO-VF-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 2
; UNROLL-NO-VF-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
; UNROLL-NO-VF-NEXT:    [[IND_END:%.*]] = mul i64 [[N_VEC]], 2
; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-VF:       vector.body:
; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi i32 [ [[PRE_LOAD:%.*]], [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 2
; UNROLL-NO-VF-NEXT:    [[INDUCTION:%.*]] = add i64 [[OFFSET_IDX]], 0
; UNROLL-NO-VF-NEXT:    [[INDUCTION1:%.*]] = add i64 [[OFFSET_IDX]], 2
; UNROLL-NO-VF-NEXT:    [[TMP3:%.*]] = add nuw nsw i64 [[INDUCTION]], 2
; UNROLL-NO-VF-NEXT:    [[TMP4:%.*]] = add nuw nsw i64 [[INDUCTION1]], 2
; UNROLL-NO-VF-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP3]]
; UNROLL-NO-VF-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP4]]
; UNROLL-NO-VF-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP5]], align 4
; UNROLL-NO-VF-NEXT:    [[TMP8]] = load i32, i32* [[TMP6]], align 4
; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; UNROLL-NO-VF-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-VF-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
; UNROLL-NO-VF:       middle.block:
; UNROLL-NO-VF-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; UNROLL-NO-VF-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-VF:       scalar.ph:
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[PRE_LOAD]], [[ENTRY:%.*]] ], [ [[TMP8]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; UNROLL-NO-VF-NEXT:    br label [[SCALAR_BODY:%.*]]
; UNROLL-NO-VF:       scalar.body:
; UNROLL-NO-VF-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[SCALAR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[VAR2:%.*]], [[SCALAR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 2
; UNROLL-NO-VF-NEXT:    [[VAR1:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[I_NEXT]]
; UNROLL-NO-VF-NEXT:    [[VAR2]] = load i32, i32* [[VAR1]], align 4
; UNROLL-NO-VF-NEXT:    [[COND:%.*]] = icmp eq i64 [[I_NEXT]], [[N]]
; UNROLL-NO-VF-NEXT:    br i1 [[COND]], label [[FOR_END]], label [[SCALAR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
; UNROLL-NO-VF:       for.end:
; UNROLL-NO-VF-NEXT:    ret void
;
; SINK-AFTER-LABEL: @PR30183(
; SINK-AFTER-NEXT:  entry:
; SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i64 [[N:%.*]], -2
; SINK-AFTER-NEXT:    [[TMP1:%.*]] = lshr i64 [[TMP0]], 1
; SINK-AFTER-NEXT:    [[TMP2:%.*]] = add nuw i64 [[TMP1]], 1
; SINK-AFTER-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 4
; SINK-AFTER-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; SINK-AFTER:       vector.ph:
; SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 4
; SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
; SINK-AFTER-NEXT:    [[IND_END:%.*]] = mul i64 [[N_VEC]], 2
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i32> poison, i32 [[PRE_LOAD:%.*]], i32 3
; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
; SINK-AFTER:       vector.body:
; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[TMP22:%.*]], [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 2
; SINK-AFTER-NEXT:    [[TMP3:%.*]] = add i64 [[OFFSET_IDX]], 0
; SINK-AFTER-NEXT:    [[TMP4:%.*]] = add i64 [[OFFSET_IDX]], 2
; SINK-AFTER-NEXT:    [[TMP5:%.*]] = add i64 [[OFFSET_IDX]], 4
; SINK-AFTER-NEXT:    [[TMP6:%.*]] = add i64 [[OFFSET_IDX]], 6
; SINK-AFTER-NEXT:    [[TMP7:%.*]] = add nuw nsw i64 [[TMP3]], 2
; SINK-AFTER-NEXT:    [[TMP8:%.*]] = add nuw nsw i64 [[TMP4]], 2
; SINK-AFTER-NEXT:    [[TMP9:%.*]] = add nuw nsw i64 [[TMP5]], 2
; SINK-AFTER-NEXT:    [[TMP10:%.*]] = add nuw nsw i64 [[TMP6]], 2
; SINK-AFTER-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[TMP7]]
; SINK-AFTER-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP8]]
; SINK-AFTER-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP9]]
; SINK-AFTER-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[TMP10]]
; SINK-AFTER-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP11]], align 4
; SINK-AFTER-NEXT:    [[TMP16:%.*]] = load i32, i32* [[TMP12]], align 4
; SINK-AFTER-NEXT:    [[TMP17:%.*]] = load i32, i32* [[TMP13]], align 4
; SINK-AFTER-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP14]], align 4
; SINK-AFTER-NEXT:    [[TMP19:%.*]] = insertelement <4 x i32> poison, i32 [[TMP15]], i32 0
; SINK-AFTER-NEXT:    [[TMP20:%.*]] = insertelement <4 x i32> [[TMP19]], i32 [[TMP16]], i32 1
; SINK-AFTER-NEXT:    [[TMP21:%.*]] = insertelement <4 x i32> [[TMP20]], i32 [[TMP17]], i32 2
; SINK-AFTER-NEXT:    [[TMP22]] = insertelement <4 x i32> [[TMP21]], i32 [[TMP18]], i32 3
; SINK-AFTER-NEXT:    [[TMP23:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP22]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; SINK-AFTER-NEXT:    [[TMP24:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; SINK-AFTER-NEXT:    br i1 [[TMP24]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
; SINK-AFTER:       middle.block:
; SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP22]], i32 3
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP22]], i32 2
; SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; SINK-AFTER:       scalar.ph:
; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[PRE_LOAD]], [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; SINK-AFTER-NEXT:    br label [[SCALAR_BODY:%.*]]
; SINK-AFTER:       scalar.body:
; SINK-AFTER-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[SCALAR_BODY]] ]
; SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[VAR2:%.*]], [[SCALAR_BODY]] ]
; SINK-AFTER-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 2
; SINK-AFTER-NEXT:    [[VAR1:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[I_NEXT]]
; SINK-AFTER-NEXT:    [[VAR2]] = load i32, i32* [[VAR1]], align 4
; SINK-AFTER-NEXT:    [[COND:%.*]] = icmp eq i64 [[I_NEXT]], [[N]]
; SINK-AFTER-NEXT:    br i1 [[COND]], label [[FOR_END]], label [[SCALAR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
; SINK-AFTER:       for.end:
; SINK-AFTER-NEXT:    ret void
;
entry:
  br label %scalar.body

scalar.body:
  %i = phi i64 [ 0, %entry ], [ %i.next, %scalar.body ]
  %var0 = phi i32 [ %pre_load, %entry ], [ %var2, %scalar.body ]
  %i.next = add nuw nsw i64 %i, 2
  %var1 = getelementptr inbounds i32, i32* %a, i64 %i.next
  %var2 = load i32, i32* %var1
  %cond = icmp eq i64 %i.next,%n
  br i1 %cond, label %for.end, label %scalar.body

for.end:
  ret void
}

;
define void @constant_folded_previous_value() {
; UNROLL-NO-IC-LABEL: @constant_folded_previous_value(
; UNROLL-NO-IC-NEXT:  entry:
; UNROLL-NO-IC-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC:       vector.ph:
; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC:       vector.body:
; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i64> [ <i64 poison, i64 poison, i64 poison, i64 0>, [[VECTOR_PH]] ], [ <i64 1, i64 1, i64 1, i64 1>, [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = shufflevector <4 x i64> [[VECTOR_RECUR]], <4 x i64> <i64 1, i64 1, i64 1, i64 1>, <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = icmp eq i64 [[INDEX_NEXT]], undef
; UNROLL-NO-IC-NEXT:    br i1 [[TMP1]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
; UNROLL-NO-IC:       middle.block:
; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 undef, undef
; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC:       scalar.ph:
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ 1, [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ undef, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; UNROLL-NO-IC-NEXT:    br label [[SCALAR_BODY:%.*]]
; UNROLL-NO-IC:       scalar.body:
; UNROLL-NO-IC-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[SCALAR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi i64 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[VAR3:%.*]], [[SCALAR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[VAR3]] = add i64 0, 1
; UNROLL-NO-IC-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; UNROLL-NO-IC-NEXT:    [[COND:%.*]] = icmp eq i64 [[I_NEXT]], undef
; UNROLL-NO-IC-NEXT:    br i1 [[COND]], label [[FOR_END]], label [[SCALAR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
; UNROLL-NO-IC:       for.end:
; UNROLL-NO-IC-NEXT:    ret void
;
; UNROLL-NO-VF-LABEL: @constant_folded_previous_value(
; UNROLL-NO-VF-NEXT:  entry:
; UNROLL-NO-VF-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-VF:       vector.ph:
; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-VF:       vector.body:
; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[TMP1:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = add i64 0, 1
; UNROLL-NO-VF-NEXT:    [[TMP1]] = add i64 0, 1
; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[INDEX_NEXT]], undef
; UNROLL-NO-VF-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
; UNROLL-NO-VF:       middle.block:
; UNROLL-NO-VF-NEXT:    [[CMP_N:%.*]] = icmp eq i64 undef, undef
; UNROLL-NO-VF-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-VF:       scalar.ph:
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[TMP1]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ undef, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; UNROLL-NO-VF-NEXT:    br label [[SCALAR_BODY:%.*]]
; UNROLL-NO-VF:       scalar.body:
; UNROLL-NO-VF-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[SCALAR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR:%.*]] = phi i64 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[VAR3:%.*]], [[SCALAR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[VAR3]] = add i64 0, 1
; UNROLL-NO-VF-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; UNROLL-NO-VF-NEXT:    [[COND:%.*]] = icmp eq i64 [[I_NEXT]], undef
; UNROLL-NO-VF-NEXT:    br i1 [[COND]], label [[FOR_END]], label [[SCALAR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
; UNROLL-NO-VF:       for.end:
; UNROLL-NO-VF-NEXT:    ret void
;
; SINK-AFTER-LABEL: @constant_folded_previous_value(
; SINK-AFTER-NEXT:  entry:
; SINK-AFTER-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; SINK-AFTER:       vector.ph:
; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
; SINK-AFTER:       vector.body:
; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i64> [ <i64 poison, i64 poison, i64 poison, i64 0>, [[VECTOR_PH]] ], [ <i64 1, i64 1, i64 1, i64 1>, [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[TMP0:%.*]] = shufflevector <4 x i64> [[VECTOR_RECUR]], <4 x i64> <i64 1, i64 1, i64 1, i64 1>, <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; SINK-AFTER-NEXT:    [[TMP1:%.*]] = icmp eq i64 [[INDEX_NEXT]], undef
; SINK-AFTER-NEXT:    br i1 [[TMP1]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
; SINK-AFTER:       middle.block:
; SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 undef, undef
; SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; SINK-AFTER:       scalar.ph:
; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ 1, [[MIDDLE_BLOCK]] ]
; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ undef, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; SINK-AFTER-NEXT:    br label [[SCALAR_BODY:%.*]]
; SINK-AFTER:       scalar.body:
; SINK-AFTER-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[SCALAR_BODY]] ]
; SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i64 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[VAR3:%.*]], [[SCALAR_BODY]] ]
; SINK-AFTER-NEXT:    [[VAR3]] = add i64 0, 1
; SINK-AFTER-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; SINK-AFTER-NEXT:    [[COND:%.*]] = icmp eq i64 [[I_NEXT]], undef
; SINK-AFTER-NEXT:    br i1 [[COND]], label [[FOR_END]], label [[SCALAR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
; SINK-AFTER:       for.end:
; SINK-AFTER-NEXT:    ret void
;
entry:
  br label %scalar.body

scalar.body:
  %i = phi i64 [ 0, %entry ], [ %i.next, %scalar.body ]
  %var2 = phi i64 [ 0, %entry ], [ %var3, %scalar.body ]
  %var3 = add i64 0, 1
  %i.next = add nuw nsw i64 %i, 1
  %cond = icmp eq i64 %i.next, undef
  br i1 %cond, label %for.end, label %scalar.body

for.end:
  ret void
}

; We vectorize this first order recurrence, by generating two
; extracts for the phi `val.phi` - one at the last index and
; another at the second last index. We need these 2 extracts because
; the first order recurrence phi is used outside the loop, so we require the phi
; itself and not its update (addx).
; Check the case when unrolled but not vectorized.
define i32 @extract_second_last_iteration(i32* %cval, i32 %x)  {
; UNROLL-NO-IC-LABEL: @extract_second_last_iteration(
; UNROLL-NO-IC-NEXT:  entry:
; UNROLL-NO-IC-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC:       vector.ph:
; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[X:%.*]], i32 0
; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <4 x i32> poison, i32 [[X]], i32 0
; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLAT3:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT2]], <4 x i32> poison, <4 x i32> zeroinitializer
; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC:       vector.body:
; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP1:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[STEP_ADD:%.*]] = add <4 x i32> [[VEC_IND]], <i32 4, i32 4, i32 4, i32 4>
; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add <4 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
; UNROLL-NO-IC-NEXT:    [[TMP1]] = add <4 x i32> [[STEP_ADD]], [[BROADCAST_SPLAT3]]
; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP0]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = shufflevector <4 x i32> [[TMP0]], <4 x i32> [[TMP1]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[STEP_ADD]], <i32 4, i32 4, i32 4, i32 4>
; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = icmp eq i32 [[INDEX_NEXT]], 96
; UNROLL-NO-IC-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
; UNROLL-NO-IC:       middle.block:
; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i32 96, 96
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP1]], i32 3
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP1]], i32 2
; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC:       scalar.ph:
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 96, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY:%.*]]
; UNROLL-NO-IC:       for.body:
; UNROLL-NO-IC-NEXT:    [[INC_PHI:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[ADDX:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[INC]] = add i32 [[INC_PHI]], 1
; UNROLL-NO-IC-NEXT:    [[BC:%.*]] = zext i32 [[INC_PHI]] to i64
; UNROLL-NO-IC-NEXT:    [[ADDX]] = add i32 [[INC_PHI]], [[X]]
; UNROLL-NO-IC-NEXT:    [[CMP:%.*]] = icmp eq i32 [[INC_PHI]], 95
; UNROLL-NO-IC-NEXT:    br i1 [[CMP]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
; UNROLL-NO-IC:       for.end:
; UNROLL-NO-IC-NEXT:    [[VAL_PHI_LCSSA:%.*]] = phi i32 [ [[SCALAR_RECUR]], [[FOR_BODY]] ], [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT:    ret i32 [[VAL_PHI_LCSSA]]
;
; UNROLL-NO-VF-LABEL: @extract_second_last_iteration(
; UNROLL-NO-VF-NEXT:  entry:
; UNROLL-NO-VF-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-VF:       vector.ph:
; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-VF:       vector.body:
; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP1:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[INDUCTION:%.*]] = add i32 [[INDEX]], 0
; UNROLL-NO-VF-NEXT:    [[INDUCTION1:%.*]] = add i32 [[INDEX]], 1
; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = add i32 [[INDUCTION]], [[X:%.*]]
; UNROLL-NO-VF-NEXT:    [[TMP1]] = add i32 [[INDUCTION1]], [[X]]
; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], 96
; UNROLL-NO-VF-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
; UNROLL-NO-VF:       middle.block:
; UNROLL-NO-VF-NEXT:    [[CMP_N:%.*]] = icmp eq i32 96, 96
; UNROLL-NO-VF-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-VF:       scalar.ph:
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[TMP1]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 96, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; UNROLL-NO-VF-NEXT:    br label [[FOR_BODY:%.*]]
; UNROLL-NO-VF:       for.body:
; UNROLL-NO-VF-NEXT:    [[INC_PHI:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[ADDX:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[INC]] = add i32 [[INC_PHI]], 1
; UNROLL-NO-VF-NEXT:    [[BC:%.*]] = zext i32 [[INC_PHI]] to i64
; UNROLL-NO-VF-NEXT:    [[ADDX]] = add i32 [[INC_PHI]], [[X]]
; UNROLL-NO-VF-NEXT:    [[CMP:%.*]] = icmp eq i32 [[INC_PHI]], 95
; UNROLL-NO-VF-NEXT:    br i1 [[CMP]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
; UNROLL-NO-VF:       for.end:
; UNROLL-NO-VF-NEXT:    [[VAL_PHI_LCSSA:%.*]] = phi i32 [ [[SCALAR_RECUR]], [[FOR_BODY]] ], [ [[TMP0]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-VF-NEXT:    ret i32 [[VAL_PHI_LCSSA]]
;
; SINK-AFTER-LABEL: @extract_second_last_iteration(
; SINK-AFTER-NEXT:  entry:
; SINK-AFTER-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; SINK-AFTER:       vector.ph:
; SINK-AFTER-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[X:%.*]], i32 0
; SINK-AFTER-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
; SINK-AFTER:       vector.body:
; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP0:%.*]], [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[TMP0]] = add <4 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
; SINK-AFTER-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP0]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; SINK-AFTER-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 4, i32 4, i32 4, i32 4>
; SINK-AFTER-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], 96
; SINK-AFTER-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
; SINK-AFTER:       middle.block:
; SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i32 96, 96
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP0]], i32 3
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP0]], i32 2
; SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; SINK-AFTER:       scalar.ph:
; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 96, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; SINK-AFTER-NEXT:    br label [[FOR_BODY:%.*]]
; SINK-AFTER:       for.body:
; SINK-AFTER-NEXT:    [[INC_PHI:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
; SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[ADDX:%.*]], [[FOR_BODY]] ]
; SINK-AFTER-NEXT:    [[INC]] = add i32 [[INC_PHI]], 1
; SINK-AFTER-NEXT:    [[BC:%.*]] = zext i32 [[INC_PHI]] to i64
; SINK-AFTER-NEXT:    [[ADDX]] = add i32 [[INC_PHI]], [[X]]
; SINK-AFTER-NEXT:    [[CMP:%.*]] = icmp eq i32 [[INC_PHI]], 95
; SINK-AFTER-NEXT:    br i1 [[CMP]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
; SINK-AFTER:       for.end:
; SINK-AFTER-NEXT:    [[VAL_PHI_LCSSA:%.*]] = phi i32 [ [[SCALAR_RECUR]], [[FOR_BODY]] ], [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], [[MIDDLE_BLOCK]] ]
; SINK-AFTER-NEXT:    ret i32 [[VAL_PHI_LCSSA]]
;
entry:
  br label %for.body

for.body:
  %inc.phi = phi i32 [ 0, %entry ], [ %inc, %for.body ]
  %val.phi = phi i32 [ 0, %entry ], [ %addx, %for.body ]
  %inc = add i32 %inc.phi, 1
  %bc = zext i32 %inc.phi to i64
  %addx = add i32 %inc.phi, %x
  %cmp = icmp eq i32 %inc.phi, 95
  br i1 %cmp, label %for.end, label %for.body

for.end:
  ret i32 %val.phi
}

; We vectorize this first order recurrence, with a set of insertelements for
; each unrolled part. Make sure these insertelements are generated in-order,
; because the shuffle of the first order recurrence will be added after the
; insertelement of the last part UF - 1, assuming the latter appears after the
; insertelements of all other parts.
;
; int PR33613(double *b, double j, int d) {
;   int a = 0;
;   for(int i = 0; i < 10240; i++, b+=25) {
;     double f = b[d]; // Scalarize to form insertelements
;     if (j * f)
;       a++;
;     j = f;
;   }
;   return a;
; }
;
;
define i32 @PR33613(double* %b, double %j, i32 %d) {
; UNROLL-NO-IC-LABEL: @PR33613(
; UNROLL-NO-IC-NEXT:  entry:
; UNROLL-NO-IC-NEXT:    [[IDXPROM:%.*]] = sext i32 [[D:%.*]] to i64
; UNROLL-NO-IC-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC:       vector.ph:
; UNROLL-NO-IC-NEXT:    [[IND_END:%.*]] = getelementptr double, double* [[B:%.*]], i64 256000
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x double> poison, double [[J:%.*]], i32 3
; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC:       vector.body:
; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP48:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[VEC_PHI9:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP49:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x double> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[TMP39:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = mul i64 [[TMP0]], 25
; UNROLL-NO-IC-NEXT:    [[NEXT_GEP:%.*]] = getelementptr double, double* [[B]], i64 [[TMP1]]
; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add i64 [[INDEX]], 1
; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = mul i64 [[TMP2]], 25
; UNROLL-NO-IC-NEXT:    [[NEXT_GEP2:%.*]] = getelementptr double, double* [[B]], i64 [[TMP3]]
; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 2
; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 25
; UNROLL-NO-IC-NEXT:    [[NEXT_GEP3:%.*]] = getelementptr double, double* [[B]], i64 [[TMP5]]
; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = add i64 [[INDEX]], 3
; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = mul i64 [[TMP6]], 25
; UNROLL-NO-IC-NEXT:    [[NEXT_GEP4:%.*]] = getelementptr double, double* [[B]], i64 [[TMP7]]
; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = add i64 [[INDEX]], 4
; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = mul i64 [[TMP8]], 25
; UNROLL-NO-IC-NEXT:    [[NEXT_GEP5:%.*]] = getelementptr double, double* [[B]], i64 [[TMP9]]
; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = add i64 [[INDEX]], 5
; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 25
; UNROLL-NO-IC-NEXT:    [[NEXT_GEP6:%.*]] = getelementptr double, double* [[B]], i64 [[TMP11]]
; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = add i64 [[INDEX]], 6
; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 25
; UNROLL-NO-IC-NEXT:    [[NEXT_GEP7:%.*]] = getelementptr double, double* [[B]], i64 [[TMP13]]
; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = add i64 [[INDEX]], 7
; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = mul i64 [[TMP14]], 25
; UNROLL-NO-IC-NEXT:    [[NEXT_GEP8:%.*]] = getelementptr double, double* [[B]], i64 [[TMP15]]
; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP]], i64 [[IDXPROM]]
; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP2]], i64 [[IDXPROM]]
; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP3]], i64 [[IDXPROM]]
; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP4]], i64 [[IDXPROM]]
; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP5]], i64 [[IDXPROM]]
; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP6]], i64 [[IDXPROM]]
; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP7]], i64 [[IDXPROM]]
; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP8]], i64 [[IDXPROM]]
; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = load double, double* [[TMP16]], align 8
; UNROLL-NO-IC-NEXT:    [[TMP25:%.*]] = load double, double* [[TMP17]], align 8
; UNROLL-NO-IC-NEXT:    [[TMP26:%.*]] = load double, double* [[TMP18]], align 8
; UNROLL-NO-IC-NEXT:    [[TMP27:%.*]] = load double, double* [[TMP19]], align 8
; UNROLL-NO-IC-NEXT:    [[TMP28:%.*]] = insertelement <4 x double> poison, double [[TMP24]], i32 0
; UNROLL-NO-IC-NEXT:    [[TMP29:%.*]] = insertelement <4 x double> [[TMP28]], double [[TMP25]], i32 1
; UNROLL-NO-IC-NEXT:    [[TMP30:%.*]] = insertelement <4 x double> [[TMP29]], double [[TMP26]], i32 2
; UNROLL-NO-IC-NEXT:    [[TMP31:%.*]] = insertelement <4 x double> [[TMP30]], double [[TMP27]], i32 3
; UNROLL-NO-IC-NEXT:    [[TMP32:%.*]] = load double, double* [[TMP20]], align 8
; UNROLL-NO-IC-NEXT:    [[TMP33:%.*]] = load double, double* [[TMP21]], align 8
; UNROLL-NO-IC-NEXT:    [[TMP34:%.*]] = load double, double* [[TMP22]], align 8
; UNROLL-NO-IC-NEXT:    [[TMP35:%.*]] = load double, double* [[TMP23]], align 8
; UNROLL-NO-IC-NEXT:    [[TMP36:%.*]] = insertelement <4 x double> poison, double [[TMP32]], i32 0
; UNROLL-NO-IC-NEXT:    [[TMP37:%.*]] = insertelement <4 x double> [[TMP36]], double [[TMP33]], i32 1
; UNROLL-NO-IC-NEXT:    [[TMP38:%.*]] = insertelement <4 x double> [[TMP37]], double [[TMP34]], i32 2
; UNROLL-NO-IC-NEXT:    [[TMP39]] = insertelement <4 x double> [[TMP38]], double [[TMP35]], i32 3
; UNROLL-NO-IC-NEXT:    [[TMP40:%.*]] = shufflevector <4 x double> [[VECTOR_RECUR]], <4 x double> [[TMP31]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; UNROLL-NO-IC-NEXT:    [[TMP41:%.*]] = shufflevector <4 x double> [[TMP31]], <4 x double> [[TMP39]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; UNROLL-NO-IC-NEXT:    [[TMP42:%.*]] = fmul <4 x double> [[TMP40]], [[TMP31]]
; UNROLL-NO-IC-NEXT:    [[TMP43:%.*]] = fmul <4 x double> [[TMP41]], [[TMP39]]
; UNROLL-NO-IC-NEXT:    [[TMP44:%.*]] = fcmp une <4 x double> [[TMP42]], zeroinitializer
; UNROLL-NO-IC-NEXT:    [[TMP45:%.*]] = fcmp une <4 x double> [[TMP43]], zeroinitializer
; UNROLL-NO-IC-NEXT:    [[TMP46:%.*]] = zext <4 x i1> [[TMP44]] to <4 x i32>
; UNROLL-NO-IC-NEXT:    [[TMP47:%.*]] = zext <4 x i1> [[TMP45]] to <4 x i32>
; UNROLL-NO-IC-NEXT:    [[TMP48]] = add <4 x i32> [[VEC_PHI]], [[TMP46]]
; UNROLL-NO-IC-NEXT:    [[TMP49]] = add <4 x i32> [[VEC_PHI9]], [[TMP47]]
; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
; UNROLL-NO-IC-NEXT:    [[TMP50:%.*]] = icmp eq i64 [[INDEX_NEXT]], 10240
; UNROLL-NO-IC-NEXT:    br i1 [[TMP50]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
; UNROLL-NO-IC:       middle.block:
; UNROLL-NO-IC-NEXT:    [[BIN_RDX:%.*]] = add <4 x i32> [[TMP49]], [[TMP48]]
; UNROLL-NO-IC-NEXT:    [[TMP51:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX]])
; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 10240, 10240
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x double> [[TMP39]], i32 3
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x double> [[TMP39]], i32 2
; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC:       scalar.ph:
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi double [ [[J]], [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi double* [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[B]], [[ENTRY]] ]
; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 10240, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; UNROLL-NO-IC-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[TMP51]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY:%.*]]
; UNROLL-NO-IC:       for.cond.cleanup:
; UNROLL-NO-IC-NEXT:    [[A_1_LCSSA:%.*]] = phi i32 [ [[A_1:%.*]], [[FOR_BODY]] ], [ [[TMP51]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT:    ret i32 [[A_1_LCSSA]]
; UNROLL-NO-IC:       for.body:
; UNROLL-NO-IC-NEXT:    [[B_ADDR_012:%.*]] = phi double* [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[ADD_PTR:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[I_011:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[INC1:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[A_010:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[A_1]], [[FOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi double [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP52:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds double, double* [[B_ADDR_012]], i64 [[IDXPROM]]
; UNROLL-NO-IC-NEXT:    [[TMP52]] = load double, double* [[ARRAYIDX]], align 8
; UNROLL-NO-IC-NEXT:    [[MUL:%.*]] = fmul double [[SCALAR_RECUR]], [[TMP52]]
; UNROLL-NO-IC-NEXT:    [[TOBOOL:%.*]] = fcmp une double [[MUL]], 0.000000e+00
; UNROLL-NO-IC-NEXT:    [[INC:%.*]] = zext i1 [[TOBOOL]] to i32
; UNROLL-NO-IC-NEXT:    [[A_1]] = add nsw i32 [[A_010]], [[INC]]
; UNROLL-NO-IC-NEXT:    [[INC1]] = add nuw nsw i32 [[I_011]], 1
; UNROLL-NO-IC-NEXT:    [[ADD_PTR]] = getelementptr inbounds double, double* [[B_ADDR_012]], i64 25
; UNROLL-NO-IC-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INC1]], 10240
; UNROLL-NO-IC-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
;
; UNROLL-NO-VF-LABEL: @PR33613(
; UNROLL-NO-VF-NEXT:  entry:
; UNROLL-NO-VF-NEXT:    [[IDXPROM:%.*]] = sext i32 [[D:%.*]] to i64
; UNROLL-NO-VF-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-VF:       vector.ph:
; UNROLL-NO-VF-NEXT:    [[IND_END:%.*]] = getelementptr double, double* [[B:%.*]], i64 256000
; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-VF:       vector.body:
; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP14:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[VEC_PHI3:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi double [ [[J:%.*]], [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = mul i64 [[TMP0]], 25
; UNROLL-NO-VF-NEXT:    [[NEXT_GEP:%.*]] = getelementptr double, double* [[B]], i64 [[TMP1]]
; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = add i64 [[INDEX]], 1
; UNROLL-NO-VF-NEXT:    [[TMP3:%.*]] = mul i64 [[TMP2]], 25
; UNROLL-NO-VF-NEXT:    [[NEXT_GEP2:%.*]] = getelementptr double, double* [[B]], i64 [[TMP3]]
; UNROLL-NO-VF-NEXT:    [[TMP4:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP]], i64 [[IDXPROM]]
; UNROLL-NO-VF-NEXT:    [[TMP5:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP2]], i64 [[IDXPROM]]
; UNROLL-NO-VF-NEXT:    [[TMP6:%.*]] = load double, double* [[TMP4]], align 8
; UNROLL-NO-VF-NEXT:    [[TMP7]] = load double, double* [[TMP5]], align 8
; UNROLL-NO-VF-NEXT:    [[TMP8:%.*]] = fmul double [[VECTOR_RECUR]], [[TMP6]]
; UNROLL-NO-VF-NEXT:    [[TMP9:%.*]] = fmul double [[TMP6]], [[TMP7]]
; UNROLL-NO-VF-NEXT:    [[TMP10:%.*]] = fcmp une double [[TMP8]], 0.000000e+00
; UNROLL-NO-VF-NEXT:    [[TMP11:%.*]] = fcmp une double [[TMP9]], 0.000000e+00
; UNROLL-NO-VF-NEXT:    [[TMP12:%.*]] = zext i1 [[TMP10]] to i32
; UNROLL-NO-VF-NEXT:    [[TMP13:%.*]] = zext i1 [[TMP11]] to i32
; UNROLL-NO-VF-NEXT:    [[TMP14]] = add i32 [[VEC_PHI]], [[TMP12]]
; UNROLL-NO-VF-NEXT:    [[TMP15]] = add i32 [[VEC_PHI3]], [[TMP13]]
; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; UNROLL-NO-VF-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], 10240
; UNROLL-NO-VF-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
; UNROLL-NO-VF:       middle.block:
; UNROLL-NO-VF-NEXT:    [[BIN_RDX:%.*]] = add i32 [[TMP15]], [[TMP14]]
; UNROLL-NO-VF-NEXT:    [[CMP_N:%.*]] = icmp eq i64 10240, 10240
; UNROLL-NO-VF-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-VF:       scalar.ph:
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi double [ [[J]], [[ENTRY:%.*]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi double* [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[B]], [[ENTRY]] ]
; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 10240, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; UNROLL-NO-VF-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[BIN_RDX]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-VF-NEXT:    br label [[FOR_BODY:%.*]]
; UNROLL-NO-VF:       for.cond.cleanup:
; UNROLL-NO-VF-NEXT:    [[A_1_LCSSA:%.*]] = phi i32 [ [[A_1:%.*]], [[FOR_BODY]] ], [ [[BIN_RDX]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-VF-NEXT:    ret i32 [[A_1_LCSSA]]
; UNROLL-NO-VF:       for.body:
; UNROLL-NO-VF-NEXT:    [[B_ADDR_012:%.*]] = phi double* [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[ADD_PTR:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[I_011:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[INC1:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[A_010:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[A_1]], [[FOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR:%.*]] = phi double [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP17:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds double, double* [[B_ADDR_012]], i64 [[IDXPROM]]
; UNROLL-NO-VF-NEXT:    [[TMP17]] = load double, double* [[ARRAYIDX]], align 8
; UNROLL-NO-VF-NEXT:    [[MUL:%.*]] = fmul double [[SCALAR_RECUR]], [[TMP17]]
; UNROLL-NO-VF-NEXT:    [[TOBOOL:%.*]] = fcmp une double [[MUL]], 0.000000e+00
; UNROLL-NO-VF-NEXT:    [[INC:%.*]] = zext i1 [[TOBOOL]] to i32
; UNROLL-NO-VF-NEXT:    [[A_1]] = add nsw i32 [[A_010]], [[INC]]
; UNROLL-NO-VF-NEXT:    [[INC1]] = add nuw nsw i32 [[I_011]], 1
; UNROLL-NO-VF-NEXT:    [[ADD_PTR]] = getelementptr inbounds double, double* [[B_ADDR_012]], i64 25
; UNROLL-NO-VF-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INC1]], 10240
; UNROLL-NO-VF-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
;
; SINK-AFTER-LABEL: @PR33613(
; SINK-AFTER-NEXT:  entry:
; SINK-AFTER-NEXT:    [[IDXPROM:%.*]] = sext i32 [[D:%.*]] to i64
; SINK-AFTER-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; SINK-AFTER:       vector.ph:
; SINK-AFTER-NEXT:    [[IND_END:%.*]] = getelementptr double, double* [[B:%.*]], i64 256000
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x double> poison, double [[J:%.*]], i32 3
; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
; SINK-AFTER:       vector.body:
; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP24:%.*]], [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x double> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[TMP19:%.*]], [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
; SINK-AFTER-NEXT:    [[TMP1:%.*]] = mul i64 [[TMP0]], 25
; SINK-AFTER-NEXT:    [[NEXT_GEP:%.*]] = getelementptr double, double* [[B]], i64 [[TMP1]]
; SINK-AFTER-NEXT:    [[TMP2:%.*]] = add i64 [[INDEX]], 1
; SINK-AFTER-NEXT:    [[TMP3:%.*]] = mul i64 [[TMP2]], 25
; SINK-AFTER-NEXT:    [[NEXT_GEP2:%.*]] = getelementptr double, double* [[B]], i64 [[TMP3]]
; SINK-AFTER-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 2
; SINK-AFTER-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 25
; SINK-AFTER-NEXT:    [[NEXT_GEP3:%.*]] = getelementptr double, double* [[B]], i64 [[TMP5]]
; SINK-AFTER-NEXT:    [[TMP6:%.*]] = add i64 [[INDEX]], 3
; SINK-AFTER-NEXT:    [[TMP7:%.*]] = mul i64 [[TMP6]], 25
; SINK-AFTER-NEXT:    [[NEXT_GEP4:%.*]] = getelementptr double, double* [[B]], i64 [[TMP7]]
; SINK-AFTER-NEXT:    [[TMP8:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP]], i64 [[IDXPROM]]
; SINK-AFTER-NEXT:    [[TMP9:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP2]], i64 [[IDXPROM]]
; SINK-AFTER-NEXT:    [[TMP10:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP3]], i64 [[IDXPROM]]
; SINK-AFTER-NEXT:    [[TMP11:%.*]] = getelementptr inbounds double, double* [[NEXT_GEP4]], i64 [[IDXPROM]]
; SINK-AFTER-NEXT:    [[TMP12:%.*]] = load double, double* [[TMP8]], align 8
; SINK-AFTER-NEXT:    [[TMP13:%.*]] = load double, double* [[TMP9]], align 8
; SINK-AFTER-NEXT:    [[TMP14:%.*]] = load double, double* [[TMP10]], align 8
; SINK-AFTER-NEXT:    [[TMP15:%.*]] = load double, double* [[TMP11]], align 8
; SINK-AFTER-NEXT:    [[TMP16:%.*]] = insertelement <4 x double> poison, double [[TMP12]], i32 0
; SINK-AFTER-NEXT:    [[TMP17:%.*]] = insertelement <4 x double> [[TMP16]], double [[TMP13]], i32 1
; SINK-AFTER-NEXT:    [[TMP18:%.*]] = insertelement <4 x double> [[TMP17]], double [[TMP14]], i32 2
; SINK-AFTER-NEXT:    [[TMP19]] = insertelement <4 x double> [[TMP18]], double [[TMP15]], i32 3
; SINK-AFTER-NEXT:    [[TMP20:%.*]] = shufflevector <4 x double> [[VECTOR_RECUR]], <4 x double> [[TMP19]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; SINK-AFTER-NEXT:    [[TMP21:%.*]] = fmul <4 x double> [[TMP20]], [[TMP19]]
; SINK-AFTER-NEXT:    [[TMP22:%.*]] = fcmp une <4 x double> [[TMP21]], zeroinitializer
; SINK-AFTER-NEXT:    [[TMP23:%.*]] = zext <4 x i1> [[TMP22]] to <4 x i32>
; SINK-AFTER-NEXT:    [[TMP24]] = add <4 x i32> [[VEC_PHI]], [[TMP23]]
; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; SINK-AFTER-NEXT:    [[TMP25:%.*]] = icmp eq i64 [[INDEX_NEXT]], 10240
; SINK-AFTER-NEXT:    br i1 [[TMP25]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
; SINK-AFTER:       middle.block:
; SINK-AFTER-NEXT:    [[TMP26:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP24]])
; SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 10240, 10240
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x double> [[TMP19]], i32 3
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x double> [[TMP19]], i32 2
; SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
; SINK-AFTER:       scalar.ph:
; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi double [ [[J]], [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi double* [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[B]], [[ENTRY]] ]
; SINK-AFTER-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 10240, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; SINK-AFTER-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[TMP26]], [[MIDDLE_BLOCK]] ]
; SINK-AFTER-NEXT:    br label [[FOR_BODY:%.*]]
; SINK-AFTER:       for.cond.cleanup:
; SINK-AFTER-NEXT:    [[A_1_LCSSA:%.*]] = phi i32 [ [[A_1:%.*]], [[FOR_BODY]] ], [ [[TMP26]], [[MIDDLE_BLOCK]] ]
; SINK-AFTER-NEXT:    ret i32 [[A_1_LCSSA]]
; SINK-AFTER:       for.body:
; SINK-AFTER-NEXT:    [[B_ADDR_012:%.*]] = phi double* [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[ADD_PTR:%.*]], [[FOR_BODY]] ]
; SINK-AFTER-NEXT:    [[I_011:%.*]] = phi i32 [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[INC1:%.*]], [[FOR_BODY]] ]
; SINK-AFTER-NEXT:    [[A_010:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[A_1]], [[FOR_BODY]] ]
; SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi double [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP27:%.*]], [[FOR_BODY]] ]
; SINK-AFTER-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds double, double* [[B_ADDR_012]], i64 [[IDXPROM]]
; SINK-AFTER-NEXT:    [[TMP27]] = load double, double* [[ARRAYIDX]], align 8
; SINK-AFTER-NEXT:    [[MUL:%.*]] = fmul double [[SCALAR_RECUR]], [[TMP27]]
; SINK-AFTER-NEXT:    [[TOBOOL:%.*]] = fcmp une double [[MUL]], 0.000000e+00
; SINK-AFTER-NEXT:    [[INC:%.*]] = zext i1 [[TOBOOL]] to i32
; SINK-AFTER-NEXT:    [[A_1]] = add nsw i32 [[A_010]], [[INC]]
; SINK-AFTER-NEXT:    [[INC1]] = add nuw nsw i32 [[I_011]], 1
; SINK-AFTER-NEXT:    [[ADD_PTR]] = getelementptr inbounds double, double* [[B_ADDR_012]], i64 25
; SINK-AFTER-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INC1]], 10240
; SINK-AFTER-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
;
entry:
  %idxprom = sext i32 %d to i64
  br label %for.body

for.cond.cleanup:
  %a.1.lcssa = phi i32 [ %a.1, %for.body ]
  ret i32 %a.1.lcssa

for.body:
  %b.addr.012 = phi double* [ %b, %entry ], [ %add.ptr, %for.body ]
  %i.011 = phi i32 [ 0, %entry ], [ %inc1, %for.body ]
  %a.010 = phi i32 [ 0, %entry ], [ %a.1, %for.body ]
  %j.addr.09 = phi double [ %j, %entry ], [ %0, %for.body ]
  %arrayidx = getelementptr inbounds double, double* %b.addr.012, i64 %idxprom
  %0 = load double, double* %arrayidx, align 8
  %mul = fmul double %j.addr.09, %0
  %tobool = fcmp une double %mul, 0.000000e+00
  %inc = zext i1 %tobool to i32
  %a.1 = add nsw i32 %a.010, %inc
  %inc1 = add nuw nsw i32 %i.011, 1
  %add.ptr = getelementptr inbounds double, double* %b.addr.012, i64 25
  %exitcond = icmp eq i32 %inc1, 10240
  br i1 %exitcond, label %for.cond.cleanup, label %for.body
}

; void sink_after(short *a, int n, int *b) {
;   for(int i = 0; i < n; i++)
;     b[i] = (a[i] * a[i + 1]);
; }
;
; Check that the sext sank after the load in the vector loop.
;
define void @sink_after(i16* noalias %a, i32* noalias %b, i64 %n) {
; UNROLL-NO-IC-LABEL: @sink_after(
; UNROLL-NO-IC-NEXT:  entry:
; UNROLL-NO-IC-NEXT:    [[DOTPRE:%.*]] = load i16, i16* [[A:%.*]], align 2
; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 8
; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC:       vector.ph:
; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 8
; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i16> poison, i16 [[DOTPRE]], i32 3
; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC:       vector.body:
; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD1:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = add i64 [[INDEX]], 4
; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP0]], 1
; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add nuw nsw i64 [[TMP1]], 1
; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP2]]
; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP3]]
; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, i16* [[TMP4]], i32 0
; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = bitcast i16* [[TMP6]] to <4 x i16>*
; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, <4 x i16>* [[TMP7]], align 2
; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i16, i16* [[TMP4]], i32 4
; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = bitcast i16* [[TMP8]] to <4 x i16>*
; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD1]] = load <4 x i16>, <4 x i16>* [[TMP9]], align 2
; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = shufflevector <4 x i16> [[WIDE_LOAD]], <4 x i16> [[WIDE_LOAD1]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = sext <4 x i16> [[TMP10]] to <4 x i32>
; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = sext <4 x i16> [[TMP11]] to <4 x i32>
; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = sext <4 x i16> [[WIDE_LOAD]] to <4 x i32>
; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = sext <4 x i16> [[WIDE_LOAD1]] to <4 x i32>
; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = mul nsw <4 x i32> [[TMP14]], [[TMP12]]
; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = mul nsw <4 x i32> [[TMP15]], [[TMP13]]
; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], i64 [[TMP0]]
; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[TMP1]]
; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, i32* [[TMP18]], i32 0
; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = bitcast i32* [[TMP20]] to <4 x i32>*
; UNROLL-NO-IC-NEXT:    store <4 x i32> [[TMP16]], <4 x i32>* [[TMP21]], align 4
; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i32, i32* [[TMP18]], i32 4
; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = bitcast i32* [[TMP22]] to <4 x i32>*
; UNROLL-NO-IC-NEXT:    store <4 x i32> [[TMP17]], <4 x i32>* [[TMP23]], align 4
; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-IC-NEXT:    br i1 [[TMP24]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
; UNROLL-NO-IC:       middle.block:
; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[WIDE_LOAD1]], i32 3
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[WIDE_LOAD1]], i32 2
; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC:       scalar.ph:
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[DOTPRE]], [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY:%.*]]
; UNROLL-NO-IC:       for.body:
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP25:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[CONV:%.*]] = sext i16 [[SCALAR_RECUR]] to i32
; UNROLL-NO-IC-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; UNROLL-NO-IC-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[INDVARS_IV_NEXT]]
; UNROLL-NO-IC-NEXT:    [[TMP25]] = load i16, i16* [[ARRAYIDX2]], align 2
; UNROLL-NO-IC-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP25]] to i32
; UNROLL-NO-IC-NEXT:    [[MUL:%.*]] = mul nsw i32 [[CONV3]], [[CONV]]
; UNROLL-NO-IC-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
; UNROLL-NO-IC-NEXT:    store i32 [[MUL]], i32* [[ARRAYIDX5]], align 4
; UNROLL-NO-IC-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
; UNROLL-NO-IC-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
; UNROLL-NO-IC:       for.end:
; UNROLL-NO-IC-NEXT:    ret void
;
; UNROLL-NO-VF-LABEL: @sink_after(
; UNROLL-NO-VF-NEXT:  entry:
; UNROLL-NO-VF-NEXT:    [[DOTPRE:%.*]] = load i16, i16* [[A:%.*]], align 2
; UNROLL-NO-VF-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 2
; UNROLL-NO-VF-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-VF:       vector.ph:
; UNROLL-NO-VF-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 2
; UNROLL-NO-VF-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-VF:       vector.body:
; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi i16 [ [[DOTPRE]], [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[INDUCTION:%.*]] = add i64 [[INDEX]], 0
; UNROLL-NO-VF-NEXT:    [[INDUCTION1:%.*]] = add i64 [[INDEX]], 1
; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = add nuw nsw i64 [[INDUCTION]], 1
; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = add nuw nsw i64 [[INDUCTION1]], 1
; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP0]]
; UNROLL-NO-VF-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP1]]
; UNROLL-NO-VF-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP2]], align 2
; UNROLL-NO-VF-NEXT:    [[TMP5]] = load i16, i16* [[TMP3]], align 2
; UNROLL-NO-VF-NEXT:    [[TMP6:%.*]] = sext i16 [[VECTOR_RECUR]] to i32
; UNROLL-NO-VF-NEXT:    [[TMP7:%.*]] = sext i16 [[TMP4]] to i32
; UNROLL-NO-VF-NEXT:    [[TMP8:%.*]] = sext i16 [[TMP4]] to i32
; UNROLL-NO-VF-NEXT:    [[TMP9:%.*]] = sext i16 [[TMP5]] to i32
; UNROLL-NO-VF-NEXT:    [[TMP10:%.*]] = mul nsw i32 [[TMP8]], [[TMP6]]
; UNROLL-NO-VF-NEXT:    [[TMP11:%.*]] = mul nsw i32 [[TMP9]], [[TMP7]]
; UNROLL-NO-VF-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], i64 [[INDUCTION]]
; UNROLL-NO-VF-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDUCTION1]]
; UNROLL-NO-VF-NEXT:    store i32 [[TMP10]], i32* [[TMP12]], align 4
; UNROLL-NO-VF-NEXT:    store i32 [[TMP11]], i32* [[TMP13]], align 4
; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; UNROLL-NO-VF-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-VF-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
; UNROLL-NO-VF:       middle.block:
; UNROLL-NO-VF-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; UNROLL-NO-VF-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-VF:       scalar.ph:
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[DOTPRE]], [[ENTRY:%.*]] ], [ [[TMP5]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; UNROLL-NO-VF-NEXT:    br label [[FOR_BODY:%.*]]
; UNROLL-NO-VF:       for.body:
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP15:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[CONV:%.*]] = sext i16 [[SCALAR_RECUR]] to i32
; UNROLL-NO-VF-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; UNROLL-NO-VF-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[INDVARS_IV_NEXT]]
; UNROLL-NO-VF-NEXT:    [[TMP15]] = load i16, i16* [[ARRAYIDX2]], align 2
; UNROLL-NO-VF-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP15]] to i32
; UNROLL-NO-VF-NEXT:    [[MUL:%.*]] = mul nsw i32 [[CONV3]], [[CONV]]
; UNROLL-NO-VF-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
; UNROLL-NO-VF-NEXT:    store i32 [[MUL]], i32* [[ARRAYIDX5]], align 4
; UNROLL-NO-VF-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
; UNROLL-NO-VF-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
; UNROLL-NO-VF:       for.end:
; UNROLL-NO-VF-NEXT:    ret void
;
; SINK-AFTER-LABEL: @sink_after(
; SINK-AFTER-NEXT:  entry:
; SINK-AFTER-NEXT:    [[DOTPRE:%.*]] = load i16, i16* [[A:%.*]], align 2
; SINK-AFTER-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
; SINK-AFTER-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; SINK-AFTER:       vector.ph:
; SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
; SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i16> poison, i16 [[DOTPRE]], i32 3
; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
; SINK-AFTER:       vector.body:
; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
; SINK-AFTER-NEXT:    [[TMP1:%.*]] = add nuw nsw i64 [[TMP0]], 1
; SINK-AFTER-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP1]]
; SINK-AFTER-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i16, i16* [[TMP2]], i32 0
; SINK-AFTER-NEXT:    [[TMP4:%.*]] = bitcast i16* [[TMP3]] to <4 x i16>*
; SINK-AFTER-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, <4 x i16>* [[TMP4]], align 2
; SINK-AFTER-NEXT:    [[TMP5:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; SINK-AFTER-NEXT:    [[TMP6:%.*]] = sext <4 x i16> [[TMP5]] to <4 x i32>
; SINK-AFTER-NEXT:    [[TMP7:%.*]] = sext <4 x i16> [[WIDE_LOAD]] to <4 x i32>
; SINK-AFTER-NEXT:    [[TMP8:%.*]] = mul nsw <4 x i32> [[TMP7]], [[TMP6]]
; SINK-AFTER-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], i64 [[TMP0]]
; SINK-AFTER-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, i32* [[TMP9]], i32 0
; SINK-AFTER-NEXT:    [[TMP11:%.*]] = bitcast i32* [[TMP10]] to <4 x i32>*
; SINK-AFTER-NEXT:    store <4 x i32> [[TMP8]], <4 x i32>* [[TMP11]], align 4
; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; SINK-AFTER-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; SINK-AFTER-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
; SINK-AFTER:       middle.block:
; SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 3
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 2
; SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; SINK-AFTER:       scalar.ph:
; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[DOTPRE]], [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; SINK-AFTER-NEXT:    br label [[FOR_BODY:%.*]]
; SINK-AFTER:       for.body:
; SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP13:%.*]], [[FOR_BODY]] ]
; SINK-AFTER-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
; SINK-AFTER-NEXT:    [[CONV:%.*]] = sext i16 [[SCALAR_RECUR]] to i32
; SINK-AFTER-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; SINK-AFTER-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[INDVARS_IV_NEXT]]
; SINK-AFTER-NEXT:    [[TMP13]] = load i16, i16* [[ARRAYIDX2]], align 2
; SINK-AFTER-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP13]] to i32
; SINK-AFTER-NEXT:    [[MUL:%.*]] = mul nsw i32 [[CONV3]], [[CONV]]
; SINK-AFTER-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
; SINK-AFTER-NEXT:    store i32 [[MUL]], i32* [[ARRAYIDX5]], align 4
; SINK-AFTER-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
; SINK-AFTER-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
; SINK-AFTER:       for.end:
; SINK-AFTER-NEXT:    ret void
;
entry:
  %.pre = load i16, i16* %a
  br label %for.body

for.body:
  %0 = phi i16 [ %.pre, %entry ], [ %1, %for.body ]
  %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
  %conv = sext i16 %0 to i32
  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
  %arrayidx2 = getelementptr inbounds i16, i16* %a, i64 %indvars.iv.next
  %1 = load i16, i16* %arrayidx2
  %conv3 = sext i16 %1 to i32
  %mul = mul nsw i32 %conv3, %conv
  %arrayidx5 = getelementptr inbounds i32, i32* %b, i64 %indvars.iv
  store i32 %mul, i32* %arrayidx5
  %exitcond = icmp eq i64 %indvars.iv.next, %n
  br i1 %exitcond, label %for.end, label %for.body

for.end:
  ret void
}

; PR34711: given three consecutive instructions such that the first will be
; widened, the second is a cast that will be widened and needs to sink after the
; third, and the third is a first-order-recurring load that will be replicated
; instead of widened. Although the cast and the first instruction will both be
; widened, and are originally adjacent to each other, make sure the replicated
; load ends up appearing between them.
;
; void PR34711(short[2] *a, int *b, int *c, int n) {
;   for(int i = 0; i < n; i++) {
;     c[i] = 7;
;     b[i] = (a[i][0] * a[i][1]);
;   }
; }
;
; Check that the sext sank after the load in the vector loop.
;
define void @PR34711([2 x i16]* noalias %a, i32* noalias %b, i32* noalias %c, i64 %n) {
; UNROLL-NO-IC-LABEL: @PR34711(
; UNROLL-NO-IC-NEXT:  entry:
; UNROLL-NO-IC-NEXT:    [[PRE_INDEX:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A:%.*]], i64 0, i64 0
; UNROLL-NO-IC-NEXT:    [[DOTPRE:%.*]] = load i16, i16* [[PRE_INDEX]], align 2
; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 8
; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC:       vector.ph:
; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 8
; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i16> poison, i16 [[DOTPRE]], i32 3
; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC:       vector.body:
; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[TMP37:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = add i64 [[INDEX]], 1
; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add i64 [[INDEX]], 2
; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add i64 [[INDEX]], 3
; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 4
; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = add i64 [[INDEX]], 5
; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = add i64 [[INDEX]], 6
; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = add i64 [[INDEX]], 7
; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, i32* [[C:%.*]], i64 [[TMP0]]
; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[C]], i64 [[TMP4]]
; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP0]], i64 1
; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP1]], i64 1
; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP2]], i64 1
; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP3]], i64 1
; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP4]], i64 1
; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP5]], i64 1
; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP6]], i64 1
; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP7]], i64 1
; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, i32* [[TMP8]], i32 0
; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = bitcast i32* [[TMP18]] to <4 x i32>*
; UNROLL-NO-IC-NEXT:    store <4 x i32> <i32 7, i32 7, i32 7, i32 7>, <4 x i32>* [[TMP19]], align 4
; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, i32* [[TMP8]], i32 4
; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = bitcast i32* [[TMP20]] to <4 x i32>*
; UNROLL-NO-IC-NEXT:    store <4 x i32> <i32 7, i32 7, i32 7, i32 7>, <4 x i32>* [[TMP21]], align 4
; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = load i16, i16* [[TMP10]], align 2
; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = load i16, i16* [[TMP11]], align 2
; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = load i16, i16* [[TMP12]], align 2
; UNROLL-NO-IC-NEXT:    [[TMP25:%.*]] = load i16, i16* [[TMP13]], align 2
; UNROLL-NO-IC-NEXT:    [[TMP26:%.*]] = insertelement <4 x i16> poison, i16 [[TMP22]], i32 0
; UNROLL-NO-IC-NEXT:    [[TMP27:%.*]] = insertelement <4 x i16> [[TMP26]], i16 [[TMP23]], i32 1
; UNROLL-NO-IC-NEXT:    [[TMP28:%.*]] = insertelement <4 x i16> [[TMP27]], i16 [[TMP24]], i32 2
; UNROLL-NO-IC-NEXT:    [[TMP29:%.*]] = insertelement <4 x i16> [[TMP28]], i16 [[TMP25]], i32 3
; UNROLL-NO-IC-NEXT:    [[TMP30:%.*]] = load i16, i16* [[TMP14]], align 2
; UNROLL-NO-IC-NEXT:    [[TMP31:%.*]] = load i16, i16* [[TMP15]], align 2
; UNROLL-NO-IC-NEXT:    [[TMP32:%.*]] = load i16, i16* [[TMP16]], align 2
; UNROLL-NO-IC-NEXT:    [[TMP33:%.*]] = load i16, i16* [[TMP17]], align 2
; UNROLL-NO-IC-NEXT:    [[TMP34:%.*]] = insertelement <4 x i16> poison, i16 [[TMP30]], i32 0
; UNROLL-NO-IC-NEXT:    [[TMP35:%.*]] = insertelement <4 x i16> [[TMP34]], i16 [[TMP31]], i32 1
; UNROLL-NO-IC-NEXT:    [[TMP36:%.*]] = insertelement <4 x i16> [[TMP35]], i16 [[TMP32]], i32 2
; UNROLL-NO-IC-NEXT:    [[TMP37]] = insertelement <4 x i16> [[TMP36]], i16 [[TMP33]], i32 3
; UNROLL-NO-IC-NEXT:    [[TMP38:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[TMP29]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; UNROLL-NO-IC-NEXT:    [[TMP39:%.*]] = shufflevector <4 x i16> [[TMP29]], <4 x i16> [[TMP37]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; UNROLL-NO-IC-NEXT:    [[TMP40:%.*]] = sext <4 x i16> [[TMP38]] to <4 x i32>
; UNROLL-NO-IC-NEXT:    [[TMP41:%.*]] = sext <4 x i16> [[TMP39]] to <4 x i32>
; UNROLL-NO-IC-NEXT:    [[TMP42:%.*]] = sext <4 x i16> [[TMP29]] to <4 x i32>
; UNROLL-NO-IC-NEXT:    [[TMP43:%.*]] = sext <4 x i16> [[TMP37]] to <4 x i32>
; UNROLL-NO-IC-NEXT:    [[TMP44:%.*]] = mul nsw <4 x i32> [[TMP42]], [[TMP40]]
; UNROLL-NO-IC-NEXT:    [[TMP45:%.*]] = mul nsw <4 x i32> [[TMP43]], [[TMP41]]
; UNROLL-NO-IC-NEXT:    [[TMP46:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], i64 [[TMP0]]
; UNROLL-NO-IC-NEXT:    [[TMP47:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[TMP4]]
; UNROLL-NO-IC-NEXT:    [[TMP48:%.*]] = getelementptr inbounds i32, i32* [[TMP46]], i32 0
; UNROLL-NO-IC-NEXT:    [[TMP49:%.*]] = bitcast i32* [[TMP48]] to <4 x i32>*
; UNROLL-NO-IC-NEXT:    store <4 x i32> [[TMP44]], <4 x i32>* [[TMP49]], align 4
; UNROLL-NO-IC-NEXT:    [[TMP50:%.*]] = getelementptr inbounds i32, i32* [[TMP46]], i32 4
; UNROLL-NO-IC-NEXT:    [[TMP51:%.*]] = bitcast i32* [[TMP50]] to <4 x i32>*
; UNROLL-NO-IC-NEXT:    store <4 x i32> [[TMP45]], <4 x i32>* [[TMP51]], align 4
; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
; UNROLL-NO-IC-NEXT:    [[TMP52:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-IC-NEXT:    br i1 [[TMP52]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
; UNROLL-NO-IC:       middle.block:
; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[TMP37]], i32 3
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[TMP37]], i32 2
; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC:       scalar.ph:
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[DOTPRE]], [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY:%.*]]
; UNROLL-NO-IC:       for.body:
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP53:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[ARRAYCIDX:%.*]] = getelementptr inbounds i32, i32* [[C]], i64 [[INDVARS_IV]]
; UNROLL-NO-IC-NEXT:    [[CUR_INDEX:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[INDVARS_IV]], i64 1
; UNROLL-NO-IC-NEXT:    store i32 7, i32* [[ARRAYCIDX]], align 4
; UNROLL-NO-IC-NEXT:    [[CONV:%.*]] = sext i16 [[SCALAR_RECUR]] to i32
; UNROLL-NO-IC-NEXT:    [[TMP53]] = load i16, i16* [[CUR_INDEX]], align 2
; UNROLL-NO-IC-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP53]] to i32
; UNROLL-NO-IC-NEXT:    [[MUL:%.*]] = mul nsw i32 [[CONV3]], [[CONV]]
; UNROLL-NO-IC-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
; UNROLL-NO-IC-NEXT:    store i32 [[MUL]], i32* [[ARRAYIDX5]], align 4
; UNROLL-NO-IC-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; UNROLL-NO-IC-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
; UNROLL-NO-IC-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
; UNROLL-NO-IC:       for.end:
; UNROLL-NO-IC-NEXT:    ret void
;
; UNROLL-NO-VF-LABEL: @PR34711(
; UNROLL-NO-VF-NEXT:  entry:
; UNROLL-NO-VF-NEXT:    [[PRE_INDEX:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A:%.*]], i64 0, i64 0
; UNROLL-NO-VF-NEXT:    [[DOTPRE:%.*]] = load i16, i16* [[PRE_INDEX]], align 2
; UNROLL-NO-VF-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 2
; UNROLL-NO-VF-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-VF:       vector.ph:
; UNROLL-NO-VF-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 2
; UNROLL-NO-VF-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-VF:       vector.body:
; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi i16 [ [[DOTPRE]], [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[INDUCTION:%.*]] = add i64 [[INDEX]], 0
; UNROLL-NO-VF-NEXT:    [[INDUCTION1:%.*]] = add i64 [[INDEX]], 1
; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i32, i32* [[C:%.*]], i64 [[INDUCTION]]
; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, i32* [[C]], i64 [[INDUCTION1]]
; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[INDUCTION]], i64 1
; UNROLL-NO-VF-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[INDUCTION1]], i64 1
; UNROLL-NO-VF-NEXT:    store i32 7, i32* [[TMP0]], align 4
; UNROLL-NO-VF-NEXT:    store i32 7, i32* [[TMP1]], align 4
; UNROLL-NO-VF-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP2]], align 2
; UNROLL-NO-VF-NEXT:    [[TMP5]] = load i16, i16* [[TMP3]], align 2
; UNROLL-NO-VF-NEXT:    [[TMP6:%.*]] = sext i16 [[VECTOR_RECUR]] to i32
; UNROLL-NO-VF-NEXT:    [[TMP7:%.*]] = sext i16 [[TMP4]] to i32
; UNROLL-NO-VF-NEXT:    [[TMP8:%.*]] = sext i16 [[TMP4]] to i32
; UNROLL-NO-VF-NEXT:    [[TMP9:%.*]] = sext i16 [[TMP5]] to i32
; UNROLL-NO-VF-NEXT:    [[TMP10:%.*]] = mul nsw i32 [[TMP8]], [[TMP6]]
; UNROLL-NO-VF-NEXT:    [[TMP11:%.*]] = mul nsw i32 [[TMP9]], [[TMP7]]
; UNROLL-NO-VF-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], i64 [[INDUCTION]]
; UNROLL-NO-VF-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDUCTION1]]
; UNROLL-NO-VF-NEXT:    store i32 [[TMP10]], i32* [[TMP12]], align 4
; UNROLL-NO-VF-NEXT:    store i32 [[TMP11]], i32* [[TMP13]], align 4
; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; UNROLL-NO-VF-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-VF-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
; UNROLL-NO-VF:       middle.block:
; UNROLL-NO-VF-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; UNROLL-NO-VF-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-VF:       scalar.ph:
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[DOTPRE]], [[ENTRY:%.*]] ], [ [[TMP5]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; UNROLL-NO-VF-NEXT:    br label [[FOR_BODY:%.*]]
; UNROLL-NO-VF:       for.body:
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP15:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[ARRAYCIDX:%.*]] = getelementptr inbounds i32, i32* [[C]], i64 [[INDVARS_IV]]
; UNROLL-NO-VF-NEXT:    [[CUR_INDEX:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[INDVARS_IV]], i64 1
; UNROLL-NO-VF-NEXT:    store i32 7, i32* [[ARRAYCIDX]], align 4
; UNROLL-NO-VF-NEXT:    [[CONV:%.*]] = sext i16 [[SCALAR_RECUR]] to i32
; UNROLL-NO-VF-NEXT:    [[TMP15]] = load i16, i16* [[CUR_INDEX]], align 2
; UNROLL-NO-VF-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP15]] to i32
; UNROLL-NO-VF-NEXT:    [[MUL:%.*]] = mul nsw i32 [[CONV3]], [[CONV]]
; UNROLL-NO-VF-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
; UNROLL-NO-VF-NEXT:    store i32 [[MUL]], i32* [[ARRAYIDX5]], align 4
; UNROLL-NO-VF-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; UNROLL-NO-VF-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
; UNROLL-NO-VF-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
; UNROLL-NO-VF:       for.end:
; UNROLL-NO-VF-NEXT:    ret void
;
; SINK-AFTER-LABEL: @PR34711(
; SINK-AFTER-NEXT:  entry:
; SINK-AFTER-NEXT:    [[PRE_INDEX:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A:%.*]], i64 0, i64 0
; SINK-AFTER-NEXT:    [[DOTPRE:%.*]] = load i16, i16* [[PRE_INDEX]], align 2
; SINK-AFTER-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
; SINK-AFTER-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; SINK-AFTER:       vector.ph:
; SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
; SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i16> poison, i16 [[DOTPRE]], i32 3
; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
; SINK-AFTER:       vector.body:
; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[TMP18:%.*]], [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
; SINK-AFTER-NEXT:    [[TMP1:%.*]] = add i64 [[INDEX]], 1
; SINK-AFTER-NEXT:    [[TMP2:%.*]] = add i64 [[INDEX]], 2
; SINK-AFTER-NEXT:    [[TMP3:%.*]] = add i64 [[INDEX]], 3
; SINK-AFTER-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[C:%.*]], i64 [[TMP0]]
; SINK-AFTER-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP0]], i64 1
; SINK-AFTER-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP1]], i64 1
; SINK-AFTER-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP2]], i64 1
; SINK-AFTER-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[TMP3]], i64 1
; SINK-AFTER-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[TMP4]], i32 0
; SINK-AFTER-NEXT:    [[TMP10:%.*]] = bitcast i32* [[TMP9]] to <4 x i32>*
; SINK-AFTER-NEXT:    store <4 x i32> <i32 7, i32 7, i32 7, i32 7>, <4 x i32>* [[TMP10]], align 4
; SINK-AFTER-NEXT:    [[TMP11:%.*]] = load i16, i16* [[TMP5]], align 2
; SINK-AFTER-NEXT:    [[TMP12:%.*]] = load i16, i16* [[TMP6]], align 2
; SINK-AFTER-NEXT:    [[TMP13:%.*]] = load i16, i16* [[TMP7]], align 2
; SINK-AFTER-NEXT:    [[TMP14:%.*]] = load i16, i16* [[TMP8]], align 2
; SINK-AFTER-NEXT:    [[TMP15:%.*]] = insertelement <4 x i16> poison, i16 [[TMP11]], i32 0
; SINK-AFTER-NEXT:    [[TMP16:%.*]] = insertelement <4 x i16> [[TMP15]], i16 [[TMP12]], i32 1
; SINK-AFTER-NEXT:    [[TMP17:%.*]] = insertelement <4 x i16> [[TMP16]], i16 [[TMP13]], i32 2
; SINK-AFTER-NEXT:    [[TMP18]] = insertelement <4 x i16> [[TMP17]], i16 [[TMP14]], i32 3
; SINK-AFTER-NEXT:    [[TMP19:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[TMP18]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; SINK-AFTER-NEXT:    [[TMP20:%.*]] = sext <4 x i16> [[TMP19]] to <4 x i32>
; SINK-AFTER-NEXT:    [[TMP21:%.*]] = sext <4 x i16> [[TMP18]] to <4 x i32>
; SINK-AFTER-NEXT:    [[TMP22:%.*]] = mul nsw <4 x i32> [[TMP21]], [[TMP20]]
; SINK-AFTER-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], i64 [[TMP0]]
; SINK-AFTER-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i32, i32* [[TMP23]], i32 0
; SINK-AFTER-NEXT:    [[TMP25:%.*]] = bitcast i32* [[TMP24]] to <4 x i32>*
; SINK-AFTER-NEXT:    store <4 x i32> [[TMP22]], <4 x i32>* [[TMP25]], align 4
; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; SINK-AFTER-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; SINK-AFTER-NEXT:    br i1 [[TMP26]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
; SINK-AFTER:       middle.block:
; SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[TMP18]], i32 3
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[TMP18]], i32 2
; SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; SINK-AFTER:       scalar.ph:
; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[DOTPRE]], [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; SINK-AFTER-NEXT:    br label [[FOR_BODY:%.*]]
; SINK-AFTER:       for.body:
; SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP27:%.*]], [[FOR_BODY]] ]
; SINK-AFTER-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
; SINK-AFTER-NEXT:    [[ARRAYCIDX:%.*]] = getelementptr inbounds i32, i32* [[C]], i64 [[INDVARS_IV]]
; SINK-AFTER-NEXT:    [[CUR_INDEX:%.*]] = getelementptr inbounds [2 x i16], [2 x i16]* [[A]], i64 [[INDVARS_IV]], i64 1
; SINK-AFTER-NEXT:    store i32 7, i32* [[ARRAYCIDX]], align 4
; SINK-AFTER-NEXT:    [[CONV:%.*]] = sext i16 [[SCALAR_RECUR]] to i32
; SINK-AFTER-NEXT:    [[TMP27]] = load i16, i16* [[CUR_INDEX]], align 2
; SINK-AFTER-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP27]] to i32
; SINK-AFTER-NEXT:    [[MUL:%.*]] = mul nsw i32 [[CONV3]], [[CONV]]
; SINK-AFTER-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
; SINK-AFTER-NEXT:    store i32 [[MUL]], i32* [[ARRAYIDX5]], align 4
; SINK-AFTER-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; SINK-AFTER-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
; SINK-AFTER-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
; SINK-AFTER:       for.end:
; SINK-AFTER-NEXT:    ret void
;

entry:
  %pre.index = getelementptr inbounds [2 x i16], [2 x i16]* %a, i64 0, i64 0
  %.pre = load i16, i16* %pre.index
  br label %for.body

for.body:
  %0 = phi i16 [ %.pre, %entry ], [ %1, %for.body ]
  %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
  %arraycidx = getelementptr inbounds i32, i32* %c, i64 %indvars.iv
  %cur.index = getelementptr inbounds [2 x i16], [2 x i16]* %a, i64 %indvars.iv, i64 1
  store i32 7, i32* %arraycidx   ; 1st instruction, to be widened.
  %conv = sext i16 %0 to i32     ; 2nd, cast to sink after third.
  %1 = load i16, i16* %cur.index ; 3rd, first-order-recurring load not widened.
  %conv3 = sext i16 %1 to i32
  %mul = mul nsw i32 %conv3, %conv
  %arrayidx5 = getelementptr inbounds i32, i32* %b, i64 %indvars.iv
  store i32 %mul, i32* %arrayidx5
  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
  %exitcond = icmp eq i64 %indvars.iv.next, %n
  br i1 %exitcond, label %for.end, label %for.body

for.end:
  ret void
}

; void no_sink_after(short *a, int n, int *b) {
;   for(int i = 0; i < n; i++)
;     b[i] = ((a[i] + 2) * a[i + 1]);
; }
;

;
define void @sink_after_with_multiple_users(i16* noalias %a, i32* noalias %b, i64 %n) {
; UNROLL-NO-IC-LABEL: @sink_after_with_multiple_users(
; UNROLL-NO-IC-NEXT:  entry:
; UNROLL-NO-IC-NEXT:    [[DOTPRE:%.*]] = load i16, i16* [[A:%.*]], align 2
; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 8
; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC:       vector.ph:
; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 8
; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i16> poison, i16 [[DOTPRE]], i32 3
; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC:       vector.body:
; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD1:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = add i64 [[INDEX]], 4
; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP0]], 1
; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add nuw nsw i64 [[TMP1]], 1
; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP2]]
; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP3]]
; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, i16* [[TMP4]], i32 0
; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = bitcast i16* [[TMP6]] to <4 x i16>*
; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, <4 x i16>* [[TMP7]], align 2
; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i16, i16* [[TMP4]], i32 4
; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = bitcast i16* [[TMP8]] to <4 x i16>*
; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD1]] = load <4 x i16>, <4 x i16>* [[TMP9]], align 2
; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = shufflevector <4 x i16> [[WIDE_LOAD]], <4 x i16> [[WIDE_LOAD1]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = sext <4 x i16> [[TMP10]] to <4 x i32>
; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = sext <4 x i16> [[TMP11]] to <4 x i32>
; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = add nsw <4 x i32> [[TMP12]], <i32 2, i32 2, i32 2, i32 2>
; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = add nsw <4 x i32> [[TMP13]], <i32 2, i32 2, i32 2, i32 2>
; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = sext <4 x i16> [[WIDE_LOAD]] to <4 x i32>
; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = sext <4 x i16> [[WIDE_LOAD1]] to <4 x i32>
; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = mul nsw <4 x i32> [[TMP14]], [[TMP16]]
; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = mul nsw <4 x i32> [[TMP15]], [[TMP17]]
; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], i64 [[TMP0]]
; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[TMP1]]
; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i32, i32* [[TMP20]], i32 0
; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = bitcast i32* [[TMP22]] to <4 x i32>*
; UNROLL-NO-IC-NEXT:    store <4 x i32> [[TMP18]], <4 x i32>* [[TMP23]], align 4
; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i32, i32* [[TMP20]], i32 4
; UNROLL-NO-IC-NEXT:    [[TMP25:%.*]] = bitcast i32* [[TMP24]] to <4 x i32>*
; UNROLL-NO-IC-NEXT:    store <4 x i32> [[TMP19]], <4 x i32>* [[TMP25]], align 4
; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
; UNROLL-NO-IC-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-IC-NEXT:    br i1 [[TMP26]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
; UNROLL-NO-IC:       middle.block:
; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[WIDE_LOAD1]], i32 3
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[WIDE_LOAD1]], i32 2
; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC:       scalar.ph:
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[DOTPRE]], [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY:%.*]]
; UNROLL-NO-IC:       for.body:
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP27:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[CONV:%.*]] = sext i16 [[SCALAR_RECUR]] to i32
; UNROLL-NO-IC-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 2
; UNROLL-NO-IC-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; UNROLL-NO-IC-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[INDVARS_IV_NEXT]]
; UNROLL-NO-IC-NEXT:    [[TMP27]] = load i16, i16* [[ARRAYIDX2]], align 2
; UNROLL-NO-IC-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP27]] to i32
; UNROLL-NO-IC-NEXT:    [[MUL:%.*]] = mul nsw i32 [[ADD]], [[CONV3]]
; UNROLL-NO-IC-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
; UNROLL-NO-IC-NEXT:    store i32 [[MUL]], i32* [[ARRAYIDX5]], align 4
; UNROLL-NO-IC-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
; UNROLL-NO-IC-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
; UNROLL-NO-IC:       for.end:
; UNROLL-NO-IC-NEXT:    ret void
;
; UNROLL-NO-VF-LABEL: @sink_after_with_multiple_users(
; UNROLL-NO-VF-NEXT:  entry:
; UNROLL-NO-VF-NEXT:    [[DOTPRE:%.*]] = load i16, i16* [[A:%.*]], align 2
; UNROLL-NO-VF-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 2
; UNROLL-NO-VF-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-VF:       vector.ph:
; UNROLL-NO-VF-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 2
; UNROLL-NO-VF-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-VF:       vector.body:
; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi i16 [ [[DOTPRE]], [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[INDUCTION:%.*]] = add i64 [[INDEX]], 0
; UNROLL-NO-VF-NEXT:    [[INDUCTION1:%.*]] = add i64 [[INDEX]], 1
; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = add nuw nsw i64 [[INDUCTION]], 1
; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = add nuw nsw i64 [[INDUCTION1]], 1
; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP0]]
; UNROLL-NO-VF-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP1]]
; UNROLL-NO-VF-NEXT:    [[TMP4:%.*]] = load i16, i16* [[TMP2]], align 2
; UNROLL-NO-VF-NEXT:    [[TMP5]] = load i16, i16* [[TMP3]], align 2
; UNROLL-NO-VF-NEXT:    [[TMP6:%.*]] = sext i16 [[VECTOR_RECUR]] to i32
; UNROLL-NO-VF-NEXT:    [[TMP7:%.*]] = sext i16 [[TMP4]] to i32
; UNROLL-NO-VF-NEXT:    [[TMP8:%.*]] = add nsw i32 [[TMP6]], 2
; UNROLL-NO-VF-NEXT:    [[TMP9:%.*]] = add nsw i32 [[TMP7]], 2
; UNROLL-NO-VF-NEXT:    [[TMP10:%.*]] = sext i16 [[TMP4]] to i32
; UNROLL-NO-VF-NEXT:    [[TMP11:%.*]] = sext i16 [[TMP5]] to i32
; UNROLL-NO-VF-NEXT:    [[TMP12:%.*]] = mul nsw i32 [[TMP8]], [[TMP10]]
; UNROLL-NO-VF-NEXT:    [[TMP13:%.*]] = mul nsw i32 [[TMP9]], [[TMP11]]
; UNROLL-NO-VF-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], i64 [[INDUCTION]]
; UNROLL-NO-VF-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDUCTION1]]
; UNROLL-NO-VF-NEXT:    store i32 [[TMP12]], i32* [[TMP14]], align 4
; UNROLL-NO-VF-NEXT:    store i32 [[TMP13]], i32* [[TMP15]], align 4
; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; UNROLL-NO-VF-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-VF-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
; UNROLL-NO-VF:       middle.block:
; UNROLL-NO-VF-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; UNROLL-NO-VF-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-VF:       scalar.ph:
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[DOTPRE]], [[ENTRY:%.*]] ], [ [[TMP5]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; UNROLL-NO-VF-NEXT:    br label [[FOR_BODY:%.*]]
; UNROLL-NO-VF:       for.body:
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP17:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[CONV:%.*]] = sext i16 [[SCALAR_RECUR]] to i32
; UNROLL-NO-VF-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 2
; UNROLL-NO-VF-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; UNROLL-NO-VF-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[INDVARS_IV_NEXT]]
; UNROLL-NO-VF-NEXT:    [[TMP17]] = load i16, i16* [[ARRAYIDX2]], align 2
; UNROLL-NO-VF-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP17]] to i32
; UNROLL-NO-VF-NEXT:    [[MUL:%.*]] = mul nsw i32 [[ADD]], [[CONV3]]
; UNROLL-NO-VF-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
; UNROLL-NO-VF-NEXT:    store i32 [[MUL]], i32* [[ARRAYIDX5]], align 4
; UNROLL-NO-VF-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
; UNROLL-NO-VF-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
; UNROLL-NO-VF:       for.end:
; UNROLL-NO-VF-NEXT:    ret void
;
; SINK-AFTER-LABEL: @sink_after_with_multiple_users(
; SINK-AFTER-NEXT:  entry:
; SINK-AFTER-NEXT:    [[DOTPRE:%.*]] = load i16, i16* [[A:%.*]], align 2
; SINK-AFTER-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
; SINK-AFTER-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; SINK-AFTER:       vector.ph:
; SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
; SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x i16> poison, i16 [[DOTPRE]], i32 3
; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
; SINK-AFTER:       vector.body:
; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
; SINK-AFTER-NEXT:    [[TMP1:%.*]] = add nuw nsw i64 [[TMP0]], 1
; SINK-AFTER-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[TMP1]]
; SINK-AFTER-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i16, i16* [[TMP2]], i32 0
; SINK-AFTER-NEXT:    [[TMP4:%.*]] = bitcast i16* [[TMP3]] to <4 x i16>*
; SINK-AFTER-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, <4 x i16>* [[TMP4]], align 2
; SINK-AFTER-NEXT:    [[TMP5:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; SINK-AFTER-NEXT:    [[TMP6:%.*]] = sext <4 x i16> [[TMP5]] to <4 x i32>
; SINK-AFTER-NEXT:    [[TMP7:%.*]] = add nsw <4 x i32> [[TMP6]], <i32 2, i32 2, i32 2, i32 2>
; SINK-AFTER-NEXT:    [[TMP8:%.*]] = sext <4 x i16> [[WIDE_LOAD]] to <4 x i32>
; SINK-AFTER-NEXT:    [[TMP9:%.*]] = mul nsw <4 x i32> [[TMP7]], [[TMP8]]
; SINK-AFTER-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], i64 [[TMP0]]
; SINK-AFTER-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, i32* [[TMP10]], i32 0
; SINK-AFTER-NEXT:    [[TMP12:%.*]] = bitcast i32* [[TMP11]] to <4 x i32>*
; SINK-AFTER-NEXT:    store <4 x i32> [[TMP9]], <4 x i32>* [[TMP12]], align 4
; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; SINK-AFTER-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; SINK-AFTER-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
; SINK-AFTER:       middle.block:
; SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 3
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 2
; SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; SINK-AFTER:       scalar.ph:
; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[DOTPRE]], [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; SINK-AFTER-NEXT:    br label [[FOR_BODY:%.*]]
; SINK-AFTER:       for.body:
; SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[TMP14:%.*]], [[FOR_BODY]] ]
; SINK-AFTER-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
; SINK-AFTER-NEXT:    [[CONV:%.*]] = sext i16 [[SCALAR_RECUR]] to i32
; SINK-AFTER-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 2
; SINK-AFTER-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; SINK-AFTER-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[A]], i64 [[INDVARS_IV_NEXT]]
; SINK-AFTER-NEXT:    [[TMP14]] = load i16, i16* [[ARRAYIDX2]], align 2
; SINK-AFTER-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP14]] to i32
; SINK-AFTER-NEXT:    [[MUL:%.*]] = mul nsw i32 [[ADD]], [[CONV3]]
; SINK-AFTER-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[INDVARS_IV]]
; SINK-AFTER-NEXT:    store i32 [[MUL]], i32* [[ARRAYIDX5]], align 4
; SINK-AFTER-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
; SINK-AFTER-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
; SINK-AFTER:       for.end:
; SINK-AFTER-NEXT:    ret void
;
entry:
  %.pre = load i16, i16* %a
  br label %for.body

for.body:
  %0 = phi i16 [ %.pre, %entry ], [ %1, %for.body ]
  %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
  %conv = sext i16 %0 to i32
  %add = add nsw i32 %conv, 2
  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
  %arrayidx2 = getelementptr inbounds i16, i16* %a, i64 %indvars.iv.next
  %1 = load i16, i16* %arrayidx2
  %conv3 = sext i16 %1 to i32
  %mul = mul nsw i32 %add, %conv3
  %arrayidx5 = getelementptr inbounds i32, i32* %b, i64 %indvars.iv
  store i32 %mul, i32* %arrayidx5
  %exitcond = icmp eq i64 %indvars.iv.next, %n
  br i1 %exitcond, label %for.end, label %for.body

for.end:
  ret void
}

; Do not sink branches: While branches are if-converted and do not require
; sinking, instructions with side effects (e.g. loads) conditioned by those
; branches will become users of the condition bit after vectorization and would
; need to be sunk if the loop is vectorized.
define void @do_not_sink_branch(i32 %x, i32* %in, i32* %out, i32 %tc) local_unnamed_addr #0 {
; UNROLL-NO-IC-LABEL: @do_not_sink_branch(
; UNROLL-NO-IC-NEXT:  entry:
; UNROLL-NO-IC-NEXT:    [[CMP530:%.*]] = icmp slt i32 0, [[TC:%.*]]
; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY4:%.*]]
; UNROLL-NO-IC:       for.body4:
; UNROLL-NO-IC-NEXT:    [[INDVARS_IV:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[COND_END:%.*]] ]
; UNROLL-NO-IC-NEXT:    [[CMP534:%.*]] = phi i1 [ [[CMP530]], [[ENTRY]] ], [ [[CMP5:%.*]], [[COND_END]] ]
; UNROLL-NO-IC-NEXT:    br i1 [[CMP534]], label [[COND_TRUE:%.*]], label [[COND_END]]
; UNROLL-NO-IC:       cond.true:
; UNROLL-NO-IC-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i32, i32* [[IN:%.*]], i32 [[INDVARS_IV]]
; UNROLL-NO-IC-NEXT:    [[IN_VAL:%.*]] = load i32, i32* [[ARRAYIDX7]], align 4
; UNROLL-NO-IC-NEXT:    br label [[COND_END]]
; UNROLL-NO-IC:       cond.end:
; UNROLL-NO-IC-NEXT:    [[COND:%.*]] = phi i32 [ [[IN_VAL]], [[COND_TRUE]] ], [ 0, [[FOR_BODY4]] ]
; UNROLL-NO-IC-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i32, i32* [[OUT:%.*]], i32 [[INDVARS_IV]]
; UNROLL-NO-IC-NEXT:    store i32 [[COND]], i32* [[ARRAYIDX8]], align 4
; UNROLL-NO-IC-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
; UNROLL-NO-IC-NEXT:    [[CMP5]] = icmp slt i32 [[INDVARS_IV_NEXT]], [[TC]]
; UNROLL-NO-IC-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[X:%.*]]
; UNROLL-NO-IC-NEXT:    br i1 [[EXITCOND]], label [[FOR_END12_LOOPEXIT:%.*]], label [[FOR_BODY4]]
; UNROLL-NO-IC:       for.end12.loopexit:
; UNROLL-NO-IC-NEXT:    ret void
;
; UNROLL-NO-VF-LABEL: @do_not_sink_branch(
; UNROLL-NO-VF-NEXT:  entry:
; UNROLL-NO-VF-NEXT:    [[CMP530:%.*]] = icmp slt i32 0, [[TC:%.*]]
; UNROLL-NO-VF-NEXT:    br label [[FOR_BODY4:%.*]]
; UNROLL-NO-VF:       for.body4:
; UNROLL-NO-VF-NEXT:    [[INDVARS_IV:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[COND_END:%.*]] ]
; UNROLL-NO-VF-NEXT:    [[CMP534:%.*]] = phi i1 [ [[CMP530]], [[ENTRY]] ], [ [[CMP5:%.*]], [[COND_END]] ]
; UNROLL-NO-VF-NEXT:    br i1 [[CMP534]], label [[COND_TRUE:%.*]], label [[COND_END]]
; UNROLL-NO-VF:       cond.true:
; UNROLL-NO-VF-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i32, i32* [[IN:%.*]], i32 [[INDVARS_IV]]
; UNROLL-NO-VF-NEXT:    [[IN_VAL:%.*]] = load i32, i32* [[ARRAYIDX7]], align 4
; UNROLL-NO-VF-NEXT:    br label [[COND_END]]
; UNROLL-NO-VF:       cond.end:
; UNROLL-NO-VF-NEXT:    [[COND:%.*]] = phi i32 [ [[IN_VAL]], [[COND_TRUE]] ], [ 0, [[FOR_BODY4]] ]
; UNROLL-NO-VF-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i32, i32* [[OUT:%.*]], i32 [[INDVARS_IV]]
; UNROLL-NO-VF-NEXT:    store i32 [[COND]], i32* [[ARRAYIDX8]], align 4
; UNROLL-NO-VF-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
; UNROLL-NO-VF-NEXT:    [[CMP5]] = icmp slt i32 [[INDVARS_IV_NEXT]], [[TC]]
; UNROLL-NO-VF-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[X:%.*]]
; UNROLL-NO-VF-NEXT:    br i1 [[EXITCOND]], label [[FOR_END12_LOOPEXIT:%.*]], label [[FOR_BODY4]]
; UNROLL-NO-VF:       for.end12.loopexit:
; UNROLL-NO-VF-NEXT:    ret void
;
; SINK-AFTER-LABEL: @do_not_sink_branch(
; SINK-AFTER-NEXT:  entry:
; SINK-AFTER-NEXT:    [[CMP530:%.*]] = icmp slt i32 0, [[TC:%.*]]
; SINK-AFTER-NEXT:    br label [[FOR_BODY4:%.*]]
; SINK-AFTER:       for.body4:
; SINK-AFTER-NEXT:    [[INDVARS_IV:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[COND_END:%.*]] ]
; SINK-AFTER-NEXT:    [[CMP534:%.*]] = phi i1 [ [[CMP530]], [[ENTRY]] ], [ [[CMP5:%.*]], [[COND_END]] ]
; SINK-AFTER-NEXT:    br i1 [[CMP534]], label [[COND_TRUE:%.*]], label [[COND_END]]
; SINK-AFTER:       cond.true:
; SINK-AFTER-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i32, i32* [[IN:%.*]], i32 [[INDVARS_IV]]
; SINK-AFTER-NEXT:    [[IN_VAL:%.*]] = load i32, i32* [[ARRAYIDX7]], align 4
; SINK-AFTER-NEXT:    br label [[COND_END]]
; SINK-AFTER:       cond.end:
; SINK-AFTER-NEXT:    [[COND:%.*]] = phi i32 [ [[IN_VAL]], [[COND_TRUE]] ], [ 0, [[FOR_BODY4]] ]
; SINK-AFTER-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i32, i32* [[OUT:%.*]], i32 [[INDVARS_IV]]
; SINK-AFTER-NEXT:    store i32 [[COND]], i32* [[ARRAYIDX8]], align 4
; SINK-AFTER-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 1
; SINK-AFTER-NEXT:    [[CMP5]] = icmp slt i32 [[INDVARS_IV_NEXT]], [[TC]]
; SINK-AFTER-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], [[X:%.*]]
; SINK-AFTER-NEXT:    br i1 [[EXITCOND]], label [[FOR_END12_LOOPEXIT:%.*]], label [[FOR_BODY4]]
; SINK-AFTER:       for.end12.loopexit:
; SINK-AFTER-NEXT:    ret void
;
entry:
  %cmp530 = icmp slt i32 0, %tc
  br label %for.body4

for.body4:                                        ; preds = %cond.end, %entry
  %indvars.iv = phi i32 [ 0, %entry ], [ %indvars.iv.next, %cond.end ]
  %cmp534 = phi i1 [ %cmp530, %entry ], [ %cmp5, %cond.end ]
  br i1 %cmp534, label %cond.true, label %cond.end

cond.true:                                        ; preds = %for.body4
  %arrayidx7 = getelementptr inbounds i32, i32* %in, i32 %indvars.iv
  %in.val = load i32, i32* %arrayidx7, align 4
  br label %cond.end

cond.end:                                         ; preds = %for.body4, %cond.true
  %cond = phi i32 [ %in.val, %cond.true ], [ 0, %for.body4 ]
  %arrayidx8 = getelementptr inbounds i32, i32* %out, i32 %indvars.iv
  store i32 %cond, i32* %arrayidx8, align 4
  %indvars.iv.next = add nuw nsw i32 %indvars.iv, 1
  %cmp5 = icmp slt i32 %indvars.iv.next, %tc
  %exitcond = icmp eq i32 %indvars.iv.next, %x
  br i1 %exitcond, label %for.end12.loopexit, label %for.body4

for.end12.loopexit:                               ; preds = %cond.end
  ret void
}

; Dead instructions, like the exit condition are not part of the actual VPlan
; and do not need to be sunk. PR44634.
define void @sink_dead_inst() {
; UNROLL-NO-IC-LABEL: @sink_dead_inst(
; UNROLL-NO-IC-NEXT:  entry:
; UNROLL-NO-IC-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC:       vector.ph:
; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC:       vector.body:
; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[VEC_IND:%.*]] = phi <4 x i16> [ <i16 -27, i16 -26, i16 -25, i16 -24>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ <i16 poison, i16 poison, i16 poison, i16 0>, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR2:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 -27>, [[VECTOR_PH]] ], [ [[TMP3:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[STEP_ADD:%.*]] = add <4 x i16> [[VEC_IND]], <i16 4, i16 4, i16 4, i16 4>
; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add <4 x i16> [[VEC_IND]], <i16 1, i16 1, i16 1, i16 1>
; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = add <4 x i16> [[STEP_ADD]], <i16 1, i16 1, i16 1, i16 1>
; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = zext <4 x i16> [[TMP0]] to <4 x i32>
; UNROLL-NO-IC-NEXT:    [[TMP3]] = zext <4 x i16> [[TMP1]] to <4 x i32>
; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR2]], <4 x i32> [[TMP2]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = shufflevector <4 x i32> [[TMP2]], <4 x i32> [[TMP3]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = add <4 x i16> [[TMP0]], <i16 5, i16 5, i16 5, i16 5>
; UNROLL-NO-IC-NEXT:    [[TMP7]] = add <4 x i16> [[TMP1]], <i16 5, i16 5, i16 5, i16 5>
; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[TMP6]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = shufflevector <4 x i16> [[TMP6]], <4 x i16> [[TMP7]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <4 x i16> [[STEP_ADD]], <i16 4, i16 4, i16 4, i16 4>
; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = icmp eq i32 [[INDEX_NEXT]], 40
; UNROLL-NO-IC-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
; UNROLL-NO-IC:       middle.block:
; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i32 43, 40
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[TMP7]], i32 3
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[TMP7]], i32 2
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT3:%.*]] = extractelement <4 x i32> [[TMP3]], i32 3
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI4:%.*]] = extractelement <4 x i32> [[TMP3]], i32 2
; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC:       scalar.ph:
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT5:%.*]] = phi i32 [ -27, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT3]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ 0, [[ENTRY]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 13, [[MIDDLE_BLOCK]] ], [ -27, [[ENTRY]] ]
; UNROLL-NO-IC-NEXT:    br label [[FOR_COND:%.*]]
; UNROLL-NO-IC:       for.cond:
; UNROLL-NO-IC-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_COND]] ]
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[REC_1_PREV:%.*]], [[FOR_COND]] ]
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR6:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT5]], [[SCALAR_PH]] ], [ [[REC_2_PREV:%.*]], [[FOR_COND]] ]
; UNROLL-NO-IC-NEXT:    [[USE_REC_1:%.*]] = sub i16 [[SCALAR_RECUR]], 10
; UNROLL-NO-IC-NEXT:    [[CMP:%.*]] = icmp eq i32 [[SCALAR_RECUR6]], 15
; UNROLL-NO-IC-NEXT:    [[IV_NEXT]] = add i16 [[IV]], 1
; UNROLL-NO-IC-NEXT:    [[REC_2_PREV]] = zext i16 [[IV_NEXT]] to i32
; UNROLL-NO-IC-NEXT:    [[REC_1_PREV]] = add i16 [[IV_NEXT]], 5
; UNROLL-NO-IC-NEXT:    br i1 [[CMP]], label [[FOR_END]], label [[FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
; UNROLL-NO-IC:       for.end:
; UNROLL-NO-IC-NEXT:    ret void
;
; UNROLL-NO-VF-LABEL: @sink_dead_inst(
; UNROLL-NO-VF-NEXT:  entry:
; UNROLL-NO-VF-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-VF:       vector.ph:
; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-VF:       vector.body:
; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi i16 [ 0, [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR1:%.*]] = phi i32 [ -27, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = trunc i32 [[INDEX]] to i16
; UNROLL-NO-VF-NEXT:    [[OFFSET_IDX:%.*]] = add i16 -27, [[TMP0]]
; UNROLL-NO-VF-NEXT:    [[INDUCTION:%.*]] = add i16 [[OFFSET_IDX]], 0
; UNROLL-NO-VF-NEXT:    [[INDUCTION2:%.*]] = add i16 [[OFFSET_IDX]], 1
; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = add i16 [[INDUCTION]], 1
; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = add i16 [[INDUCTION2]], 1
; UNROLL-NO-VF-NEXT:    [[TMP3:%.*]] = zext i16 [[TMP1]] to i32
; UNROLL-NO-VF-NEXT:    [[TMP4]] = zext i16 [[TMP2]] to i32
; UNROLL-NO-VF-NEXT:    [[TMP5:%.*]] = add i16 [[TMP1]], 5
; UNROLL-NO-VF-NEXT:    [[TMP6]] = add i16 [[TMP2]], 5
; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; UNROLL-NO-VF-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], 42
; UNROLL-NO-VF-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
; UNROLL-NO-VF:       middle.block:
; UNROLL-NO-VF-NEXT:    [[CMP_N:%.*]] = icmp eq i32 43, 42
; UNROLL-NO-VF-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-VF:       scalar.ph:
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT3:%.*]] = phi i32 [ -27, [[ENTRY:%.*]] ], [ [[TMP4]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ 0, [[ENTRY]] ], [ [[TMP6]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 15, [[MIDDLE_BLOCK]] ], [ -27, [[ENTRY]] ]
; UNROLL-NO-VF-NEXT:    br label [[FOR_COND:%.*]]
; UNROLL-NO-VF:       for.cond:
; UNROLL-NO-VF-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_COND]] ]
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[REC_1_PREV:%.*]], [[FOR_COND]] ]
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR4:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT3]], [[SCALAR_PH]] ], [ [[REC_2_PREV:%.*]], [[FOR_COND]] ]
; UNROLL-NO-VF-NEXT:    [[USE_REC_1:%.*]] = sub i16 [[SCALAR_RECUR]], 10
; UNROLL-NO-VF-NEXT:    [[CMP:%.*]] = icmp eq i32 [[SCALAR_RECUR4]], 15
; UNROLL-NO-VF-NEXT:    [[IV_NEXT]] = add i16 [[IV]], 1
; UNROLL-NO-VF-NEXT:    [[REC_2_PREV]] = zext i16 [[IV_NEXT]] to i32
; UNROLL-NO-VF-NEXT:    [[REC_1_PREV]] = add i16 [[IV_NEXT]], 5
; UNROLL-NO-VF-NEXT:    br i1 [[CMP]], label [[FOR_END]], label [[FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]]
; UNROLL-NO-VF:       for.end:
; UNROLL-NO-VF-NEXT:    ret void
;
; SINK-AFTER-LABEL: @sink_dead_inst(
; SINK-AFTER-NEXT:  entry:
; SINK-AFTER-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; SINK-AFTER:       vector.ph:
; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
; SINK-AFTER:       vector.body:
; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[VEC_IND:%.*]] = phi <4 x i16> [ <i16 -27, i16 -26, i16 -25, i16 -24>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ <i16 poison, i16 poison, i16 poison, i16 0>, [[VECTOR_PH]] ], [ [[TMP3:%.*]], [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[VECTOR_RECUR1:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 -27>, [[VECTOR_PH]] ], [ [[TMP1:%.*]], [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[TMP0:%.*]] = add <4 x i16> [[VEC_IND]], <i16 1, i16 1, i16 1, i16 1>
; SINK-AFTER-NEXT:    [[TMP1]] = zext <4 x i16> [[TMP0]] to <4 x i32>
; SINK-AFTER-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR1]], <4 x i32> [[TMP1]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; SINK-AFTER-NEXT:    [[TMP3]] = add <4 x i16> [[TMP0]], <i16 5, i16 5, i16 5, i16 5>
; SINK-AFTER-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[TMP3]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; SINK-AFTER-NEXT:    [[VEC_IND_NEXT]] = add <4 x i16> [[VEC_IND]], <i16 4, i16 4, i16 4, i16 4>
; SINK-AFTER-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], 40
; SINK-AFTER-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
; SINK-AFTER:       middle.block:
; SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i32 43, 40
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[TMP3]], i32 3
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[TMP3]], i32 2
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT2:%.*]] = extractelement <4 x i32> [[TMP1]], i32 3
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI3:%.*]] = extractelement <4 x i32> [[TMP1]], i32 2
; SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; SINK-AFTER:       scalar.ph:
; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT4:%.*]] = phi i32 [ -27, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT2]], [[MIDDLE_BLOCK]] ]
; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ 0, [[ENTRY]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 13, [[MIDDLE_BLOCK]] ], [ -27, [[ENTRY]] ]
; SINK-AFTER-NEXT:    br label [[FOR_COND:%.*]]
; SINK-AFTER:       for.cond:
; SINK-AFTER-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_COND]] ]
; SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[REC_1_PREV:%.*]], [[FOR_COND]] ]
; SINK-AFTER-NEXT:    [[SCALAR_RECUR5:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT4]], [[SCALAR_PH]] ], [ [[REC_2_PREV:%.*]], [[FOR_COND]] ]
; SINK-AFTER-NEXT:    [[USE_REC_1:%.*]] = sub i16 [[SCALAR_RECUR]], 10
; SINK-AFTER-NEXT:    [[CMP:%.*]] = icmp eq i32 [[SCALAR_RECUR5]], 15
; SINK-AFTER-NEXT:    [[IV_NEXT]] = add i16 [[IV]], 1
; SINK-AFTER-NEXT:    [[REC_2_PREV]] = zext i16 [[IV_NEXT]] to i32
; SINK-AFTER-NEXT:    [[REC_1_PREV]] = add i16 [[IV_NEXT]], 5
; SINK-AFTER-NEXT:    br i1 [[CMP]], label [[FOR_END]], label [[FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
; SINK-AFTER:       for.end:
; SINK-AFTER-NEXT:    ret void
;
entry:
  br label %for.cond

for.cond:
  %iv = phi i16 [ -27, %entry ], [ %iv.next, %for.cond ]
  %rec.1 = phi i16 [ 0, %entry ], [ %rec.1.prev, %for.cond ]
  %rec.2 = phi i32 [ -27, %entry ], [ %rec.2.prev, %for.cond ]
  %use.rec.1 = sub i16 %rec.1, 10
  %cmp = icmp eq i32 %rec.2, 15
  %iv.next = add i16 %iv, 1
  %rec.2.prev = zext i16 %iv.next to i32
  %rec.1.prev = add i16 %iv.next, 5
  br i1 %cmp, label %for.end, label %for.cond

for.end:
  ret void
}

define i32 @sink_into_replication_region(i32 %y) {
; UNROLL-NO-IC-LABEL: @sink_into_replication_region(
; UNROLL-NO-IC-NEXT:  bb:
; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i32 [[Y:%.*]], 1
; UNROLL-NO-IC-NEXT:    [[SMIN:%.*]] = call i32 @llvm.smin.i32(i32 [[Y]], i32 1)
; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = sub i32 [[TMP0]], [[SMIN]]
; UNROLL-NO-IC-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC:       vector.ph:
; UNROLL-NO-IC-NEXT:    [[N_RND_UP:%.*]] = add i32 [[TMP1]], 7
; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[N_RND_UP]], 8
; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i32 [[N_RND_UP]], [[N_MOD_VF]]
; UNROLL-NO-IC-NEXT:    [[IND_END:%.*]] = sub i32 [[Y]], [[N_VEC]]
; UNROLL-NO-IC-NEXT:    [[TRIP_COUNT_MINUS_1:%.*]] = sub i32 [[TMP1]], 1
; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[TRIP_COUNT_MINUS_1]], i32 0
; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC:       vector.body:
; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UDIV_CONTINUE18:%.*]] ]
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP43:%.*]], [[PRED_UDIV_CONTINUE18]] ]
; UNROLL-NO-IC-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP46:%.*]], [[PRED_UDIV_CONTINUE18]] ]
; UNROLL-NO-IC-NEXT:    [[VEC_PHI1:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP47:%.*]], [[PRED_UDIV_CONTINUE18]] ]
; UNROLL-NO-IC-NEXT:    [[OFFSET_IDX:%.*]] = sub i32 [[Y]], [[INDEX]]
; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <4 x i32> poison, i32 [[INDEX]], i32 0
; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLAT3:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT2]], <4 x i32> poison, <4 x i32> zeroinitializer
; UNROLL-NO-IC-NEXT:    [[VEC_IV:%.*]] = add <4 x i32> [[BROADCAST_SPLAT3]], <i32 0, i32 1, i32 2, i32 3>
; UNROLL-NO-IC-NEXT:    [[VEC_IV4:%.*]] = add <4 x i32> [[BROADCAST_SPLAT3]], <i32 4, i32 5, i32 6, i32 7>
; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = icmp ule <4 x i32> [[VEC_IV]], [[BROADCAST_SPLAT]]
; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = icmp ule <4 x i32> [[VEC_IV4]], [[BROADCAST_SPLAT]]
; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = extractelement <4 x i1> [[TMP2]], i32 0
; UNROLL-NO-IC-NEXT:    br i1 [[TMP4]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
; UNROLL-NO-IC:       pred.udiv.if:
; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = add i32 [[OFFSET_IDX]], 0
; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = udiv i32 219220132, [[TMP5]]
; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = insertelement <4 x i32> poison, i32 [[TMP6]], i32 0
; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE]]
; UNROLL-NO-IC:       pred.udiv.continue:
; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = phi <4 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP7]], [[PRED_UDIV_IF]] ]
; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = extractelement <4 x i1> [[TMP2]], i32 1
; UNROLL-NO-IC-NEXT:    br i1 [[TMP9]], label [[PRED_UDIV_IF5:%.*]], label [[PRED_UDIV_CONTINUE6:%.*]]
; UNROLL-NO-IC:       pred.udiv.if5:
; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = add i32 [[OFFSET_IDX]], -1
; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = udiv i32 219220132, [[TMP10]]
; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = insertelement <4 x i32> [[TMP8]], i32 [[TMP11]], i32 1
; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE6]]
; UNROLL-NO-IC:       pred.udiv.continue6:
; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = phi <4 x i32> [ [[TMP8]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP12]], [[PRED_UDIV_IF5]] ]
; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = extractelement <4 x i1> [[TMP2]], i32 2
; UNROLL-NO-IC-NEXT:    br i1 [[TMP14]], label [[PRED_UDIV_IF7:%.*]], label [[PRED_UDIV_CONTINUE8:%.*]]
; UNROLL-NO-IC:       pred.udiv.if7:
; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = add i32 [[OFFSET_IDX]], -2
; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = udiv i32 219220132, [[TMP15]]
; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = insertelement <4 x i32> [[TMP13]], i32 [[TMP16]], i32 2
; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE8]]
; UNROLL-NO-IC:       pred.udiv.continue8:
; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = phi <4 x i32> [ [[TMP13]], [[PRED_UDIV_CONTINUE6]] ], [ [[TMP17]], [[PRED_UDIV_IF7]] ]
; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = extractelement <4 x i1> [[TMP2]], i32 3
; UNROLL-NO-IC-NEXT:    br i1 [[TMP19]], label [[PRED_UDIV_IF9:%.*]], label [[PRED_UDIV_CONTINUE10:%.*]]
; UNROLL-NO-IC:       pred.udiv.if9:
; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = add i32 [[OFFSET_IDX]], -3
; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = udiv i32 219220132, [[TMP20]]
; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = insertelement <4 x i32> [[TMP18]], i32 [[TMP21]], i32 3
; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE10]]
; UNROLL-NO-IC:       pred.udiv.continue10:
; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = phi <4 x i32> [ [[TMP18]], [[PRED_UDIV_CONTINUE8]] ], [ [[TMP22]], [[PRED_UDIV_IF9]] ]
; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = extractelement <4 x i1> [[TMP3]], i32 0
; UNROLL-NO-IC-NEXT:    br i1 [[TMP24]], label [[PRED_UDIV_IF11:%.*]], label [[PRED_UDIV_CONTINUE12:%.*]]
; UNROLL-NO-IC:       pred.udiv.if11:
; UNROLL-NO-IC-NEXT:    [[TMP25:%.*]] = add i32 [[OFFSET_IDX]], -4
; UNROLL-NO-IC-NEXT:    [[TMP26:%.*]] = udiv i32 219220132, [[TMP25]]
; UNROLL-NO-IC-NEXT:    [[TMP27:%.*]] = insertelement <4 x i32> poison, i32 [[TMP26]], i32 0
; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE12]]
; UNROLL-NO-IC:       pred.udiv.continue12:
; UNROLL-NO-IC-NEXT:    [[TMP28:%.*]] = phi <4 x i32> [ poison, [[PRED_UDIV_CONTINUE10]] ], [ [[TMP27]], [[PRED_UDIV_IF11]] ]
; UNROLL-NO-IC-NEXT:    [[TMP29:%.*]] = extractelement <4 x i1> [[TMP3]], i32 1
; UNROLL-NO-IC-NEXT:    br i1 [[TMP29]], label [[PRED_UDIV_IF13:%.*]], label [[PRED_UDIV_CONTINUE14:%.*]]
; UNROLL-NO-IC:       pred.udiv.if13:
; UNROLL-NO-IC-NEXT:    [[TMP30:%.*]] = add i32 [[OFFSET_IDX]], -5
; UNROLL-NO-IC-NEXT:    [[TMP31:%.*]] = udiv i32 219220132, [[TMP30]]
; UNROLL-NO-IC-NEXT:    [[TMP32:%.*]] = insertelement <4 x i32> [[TMP28]], i32 [[TMP31]], i32 1
; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE14]]
; UNROLL-NO-IC:       pred.udiv.continue14:
; UNROLL-NO-IC-NEXT:    [[TMP33:%.*]] = phi <4 x i32> [ [[TMP28]], [[PRED_UDIV_CONTINUE12]] ], [ [[TMP32]], [[PRED_UDIV_IF13]] ]
; UNROLL-NO-IC-NEXT:    [[TMP34:%.*]] = extractelement <4 x i1> [[TMP3]], i32 2
; UNROLL-NO-IC-NEXT:    br i1 [[TMP34]], label [[PRED_UDIV_IF15:%.*]], label [[PRED_UDIV_CONTINUE16:%.*]]
; UNROLL-NO-IC:       pred.udiv.if15:
; UNROLL-NO-IC-NEXT:    [[TMP35:%.*]] = add i32 [[OFFSET_IDX]], -6
; UNROLL-NO-IC-NEXT:    [[TMP36:%.*]] = udiv i32 219220132, [[TMP35]]
; UNROLL-NO-IC-NEXT:    [[TMP37:%.*]] = insertelement <4 x i32> [[TMP33]], i32 [[TMP36]], i32 2
; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE16]]
; UNROLL-NO-IC:       pred.udiv.continue16:
; UNROLL-NO-IC-NEXT:    [[TMP38:%.*]] = phi <4 x i32> [ [[TMP33]], [[PRED_UDIV_CONTINUE14]] ], [ [[TMP37]], [[PRED_UDIV_IF15]] ]
; UNROLL-NO-IC-NEXT:    [[TMP39:%.*]] = extractelement <4 x i1> [[TMP3]], i32 3
; UNROLL-NO-IC-NEXT:    br i1 [[TMP39]], label [[PRED_UDIV_IF17:%.*]], label [[PRED_UDIV_CONTINUE18]]
; UNROLL-NO-IC:       pred.udiv.if17:
; UNROLL-NO-IC-NEXT:    [[TMP40:%.*]] = add i32 [[OFFSET_IDX]], -7
; UNROLL-NO-IC-NEXT:    [[TMP41:%.*]] = udiv i32 219220132, [[TMP40]]
; UNROLL-NO-IC-NEXT:    [[TMP42:%.*]] = insertelement <4 x i32> [[TMP38]], i32 [[TMP41]], i32 3
; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE18]]
; UNROLL-NO-IC:       pred.udiv.continue18:
; UNROLL-NO-IC-NEXT:    [[TMP43]] = phi <4 x i32> [ [[TMP38]], [[PRED_UDIV_CONTINUE16]] ], [ [[TMP42]], [[PRED_UDIV_IF17]] ]
; UNROLL-NO-IC-NEXT:    [[TMP44:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP23]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; UNROLL-NO-IC-NEXT:    [[TMP45:%.*]] = shufflevector <4 x i32> [[TMP23]], <4 x i32> [[TMP43]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; UNROLL-NO-IC-NEXT:    [[TMP46]] = add <4 x i32> [[VEC_PHI]], [[TMP44]]
; UNROLL-NO-IC-NEXT:    [[TMP47]] = add <4 x i32> [[VEC_PHI1]], [[TMP45]]
; UNROLL-NO-IC-NEXT:    [[TMP48:%.*]] = select <4 x i1> [[TMP2]], <4 x i32> [[TMP46]], <4 x i32> [[VEC_PHI]]
; UNROLL-NO-IC-NEXT:    [[TMP49:%.*]] = select <4 x i1> [[TMP3]], <4 x i32> [[TMP47]], <4 x i32> [[VEC_PHI1]]
; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], 8
; UNROLL-NO-IC-NEXT:    [[TMP50:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-IC-NEXT:    br i1 [[TMP50]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !prof [[PROF24:![0-9]+]], !llvm.loop [[LOOP25:![0-9]+]]
; UNROLL-NO-IC:       middle.block:
; UNROLL-NO-IC-NEXT:    [[BIN_RDX:%.*]] = add <4 x i32> [[TMP49]], [[TMP48]]
; UNROLL-NO-IC-NEXT:    [[TMP51:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX]])
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP43]], i32 3
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP43]], i32 2
; UNROLL-NO-IC-NEXT:    br i1 true, label [[BB1:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC:       scalar.ph:
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[BB:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[Y]], [[BB]] ]
; UNROLL-NO-IC-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[BB]] ], [ [[TMP51]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT:    br label [[BB2:%.*]]
; UNROLL-NO-IC:       bb1:
; UNROLL-NO-IC-NEXT:    [[VAR:%.*]] = phi i32 [ [[VAR6:%.*]], [[BB2]] ], [ [[TMP51]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT:    ret i32 [[VAR]]
; UNROLL-NO-IC:       bb2:
; UNROLL-NO-IC-NEXT:    [[VAR3:%.*]] = phi i32 [ [[VAR8:%.*]], [[BB2]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[VAR7:%.*]], [[BB2]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
; UNROLL-NO-IC-NEXT:    [[VAR5:%.*]] = phi i32 [ [[VAR6]], [[BB2]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
; UNROLL-NO-IC-NEXT:    [[VAR6]] = add i32 [[VAR5]], [[SCALAR_RECUR]]
; UNROLL-NO-IC-NEXT:    [[VAR7]] = udiv i32 219220132, [[VAR3]]
; UNROLL-NO-IC-NEXT:    [[VAR8]] = add nsw i32 [[VAR3]], -1
; UNROLL-NO-IC-NEXT:    [[VAR9:%.*]] = icmp slt i32 [[VAR3]], 2
; UNROLL-NO-IC-NEXT:    br i1 [[VAR9]], label [[BB1]], label [[BB2]], !prof [[PROF26:![0-9]+]], !llvm.loop [[LOOP27:![0-9]+]]
;
; UNROLL-NO-VF-LABEL: @sink_into_replication_region(
; UNROLL-NO-VF-NEXT:  bb:
; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = add i32 [[Y:%.*]], 1
; UNROLL-NO-VF-NEXT:    [[SMIN:%.*]] = call i32 @llvm.smin.i32(i32 [[Y]], i32 1)
; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = sub i32 [[TMP0]], [[SMIN]]
; UNROLL-NO-VF-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-VF:       vector.ph:
; UNROLL-NO-VF-NEXT:    [[N_RND_UP:%.*]] = add i32 [[TMP1]], 1
; UNROLL-NO-VF-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[N_RND_UP]], 2
; UNROLL-NO-VF-NEXT:    [[N_VEC:%.*]] = sub i32 [[N_RND_UP]], [[N_MOD_VF]]
; UNROLL-NO-VF-NEXT:    [[IND_END:%.*]] = sub i32 [[Y]], [[N_VEC]]
; UNROLL-NO-VF-NEXT:    [[TRIP_COUNT_MINUS_1:%.*]] = sub i32 [[TMP1]], 1
; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-VF:       vector.body:
; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UDIV_CONTINUE5:%.*]] ]
; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[PRED_UDIV_CONTINUE5]] ]
; UNROLL-NO-VF-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[PRED_UDIV_CONTINUE5]] ]
; UNROLL-NO-VF-NEXT:    [[VEC_PHI1:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[PRED_UDIV_CONTINUE5]] ]
; UNROLL-NO-VF-NEXT:    [[OFFSET_IDX:%.*]] = sub i32 [[Y]], [[INDEX]]
; UNROLL-NO-VF-NEXT:    [[VEC_IV:%.*]] = add i32 [[INDEX]], 0
; UNROLL-NO-VF-NEXT:    [[VEC_IV3:%.*]] = add i32 [[INDEX]], 1
; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = icmp ule i32 [[VEC_IV]], [[TRIP_COUNT_MINUS_1]]
; UNROLL-NO-VF-NEXT:    [[TMP3:%.*]] = icmp ule i32 [[VEC_IV3]], [[TRIP_COUNT_MINUS_1]]
; UNROLL-NO-VF-NEXT:    br i1 [[TMP2]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
; UNROLL-NO-VF:       pred.udiv.if:
; UNROLL-NO-VF-NEXT:    [[INDUCTION:%.*]] = add i32 [[OFFSET_IDX]], 0
; UNROLL-NO-VF-NEXT:    [[TMP4:%.*]] = udiv i32 219220132, [[INDUCTION]]
; UNROLL-NO-VF-NEXT:    br label [[PRED_UDIV_CONTINUE]]
; UNROLL-NO-VF:       pred.udiv.continue:
; UNROLL-NO-VF-NEXT:    [[TMP5:%.*]] = phi i32 [ poison, [[VECTOR_BODY]] ], [ [[TMP4]], [[PRED_UDIV_IF]] ]
; UNROLL-NO-VF-NEXT:    br i1 [[TMP3]], label [[PRED_UDIV_IF4:%.*]], label [[PRED_UDIV_CONTINUE5]]
; UNROLL-NO-VF:       pred.udiv.if4:
; UNROLL-NO-VF-NEXT:    [[INDUCTION2:%.*]] = add i32 [[OFFSET_IDX]], -1
; UNROLL-NO-VF-NEXT:    [[TMP6:%.*]] = udiv i32 219220132, [[INDUCTION2]]
; UNROLL-NO-VF-NEXT:    br label [[PRED_UDIV_CONTINUE5]]
; UNROLL-NO-VF:       pred.udiv.continue5:
; UNROLL-NO-VF-NEXT:    [[TMP7]] = phi i32 [ poison, [[PRED_UDIV_CONTINUE]] ], [ [[TMP6]], [[PRED_UDIV_IF4]] ]
; UNROLL-NO-VF-NEXT:    [[TMP8]] = add i32 [[VEC_PHI]], [[VECTOR_RECUR]]
; UNROLL-NO-VF-NEXT:    [[TMP9]] = add i32 [[VEC_PHI1]], [[TMP5]]
; UNROLL-NO-VF-NEXT:    [[TMP10:%.*]] = select i1 [[TMP2]], i32 [[TMP8]], i32 [[VEC_PHI]]
; UNROLL-NO-VF-NEXT:    [[TMP11:%.*]] = select i1 [[TMP3]], i32 [[TMP9]], i32 [[VEC_PHI1]]
; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], 2
; UNROLL-NO-VF-NEXT:    [[TMP12:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-VF-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !prof [[PROF23:![0-9]+]], !llvm.loop [[LOOP24:![0-9]+]]
; UNROLL-NO-VF:       middle.block:
; UNROLL-NO-VF-NEXT:    [[BIN_RDX:%.*]] = add i32 [[TMP11]], [[TMP10]]
; UNROLL-NO-VF-NEXT:    br i1 true, label [[BB1:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-VF:       scalar.ph:
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[BB:%.*]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[Y]], [[BB]] ]
; UNROLL-NO-VF-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[BB]] ], [ [[BIN_RDX]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-VF-NEXT:    br label [[BB2:%.*]]
; UNROLL-NO-VF:       bb1:
; UNROLL-NO-VF-NEXT:    [[VAR:%.*]] = phi i32 [ [[VAR6:%.*]], [[BB2]] ], [ [[BIN_RDX]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-VF-NEXT:    ret i32 [[VAR]]
; UNROLL-NO-VF:       bb2:
; UNROLL-NO-VF-NEXT:    [[VAR3:%.*]] = phi i32 [ [[VAR8:%.*]], [[BB2]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[VAR7:%.*]], [[BB2]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
; UNROLL-NO-VF-NEXT:    [[VAR5:%.*]] = phi i32 [ [[VAR6]], [[BB2]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
; UNROLL-NO-VF-NEXT:    [[VAR6]] = add i32 [[VAR5]], [[SCALAR_RECUR]]
; UNROLL-NO-VF-NEXT:    [[VAR7]] = udiv i32 219220132, [[VAR3]]
; UNROLL-NO-VF-NEXT:    [[VAR8]] = add nsw i32 [[VAR3]], -1
; UNROLL-NO-VF-NEXT:    [[VAR9:%.*]] = icmp slt i32 [[VAR3]], 2
; UNROLL-NO-VF-NEXT:    br i1 [[VAR9]], label [[BB1]], label [[BB2]], !prof [[PROF25:![0-9]+]], !llvm.loop [[LOOP26:![0-9]+]]
;
; SINK-AFTER-LABEL: @sink_into_replication_region(
; SINK-AFTER-NEXT:  bb:
; SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i32 [[Y:%.*]], 1
; SINK-AFTER-NEXT:    [[SMIN:%.*]] = call i32 @llvm.smin.i32(i32 [[Y]], i32 1)
; SINK-AFTER-NEXT:    [[TMP1:%.*]] = sub i32 [[TMP0]], [[SMIN]]
; SINK-AFTER-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; SINK-AFTER:       vector.ph:
; SINK-AFTER-NEXT:    [[N_RND_UP:%.*]] = add i32 [[TMP1]], 3
; SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[N_RND_UP]], 4
; SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i32 [[N_RND_UP]], [[N_MOD_VF]]
; SINK-AFTER-NEXT:    [[IND_END:%.*]] = sub i32 [[Y]], [[N_VEC]]
; SINK-AFTER-NEXT:    [[TRIP_COUNT_MINUS_1:%.*]] = sub i32 [[TMP1]], 1
; SINK-AFTER-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[TRIP_COUNT_MINUS_1]], i32 0
; SINK-AFTER-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
; SINK-AFTER:       vector.body:
; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UDIV_CONTINUE8:%.*]] ]
; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP22:%.*]], [[PRED_UDIV_CONTINUE8]] ]
; SINK-AFTER-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP24:%.*]], [[PRED_UDIV_CONTINUE8]] ]
; SINK-AFTER-NEXT:    [[OFFSET_IDX:%.*]] = sub i32 [[Y]], [[INDEX]]
; SINK-AFTER-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <4 x i32> poison, i32 [[INDEX]], i32 0
; SINK-AFTER-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT1]], <4 x i32> poison, <4 x i32> zeroinitializer
; SINK-AFTER-NEXT:    [[VEC_IV:%.*]] = add <4 x i32> [[BROADCAST_SPLAT2]], <i32 0, i32 1, i32 2, i32 3>
; SINK-AFTER-NEXT:    [[TMP2:%.*]] = icmp ule <4 x i32> [[VEC_IV]], [[BROADCAST_SPLAT]]
; SINK-AFTER-NEXT:    [[TMP3:%.*]] = extractelement <4 x i1> [[TMP2]], i32 0
; SINK-AFTER-NEXT:    br i1 [[TMP3]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
; SINK-AFTER:       pred.udiv.if:
; SINK-AFTER-NEXT:    [[TMP4:%.*]] = add i32 [[OFFSET_IDX]], 0
; SINK-AFTER-NEXT:    [[TMP5:%.*]] = udiv i32 219220132, [[TMP4]]
; SINK-AFTER-NEXT:    [[TMP6:%.*]] = insertelement <4 x i32> poison, i32 [[TMP5]], i32 0
; SINK-AFTER-NEXT:    br label [[PRED_UDIV_CONTINUE]]
; SINK-AFTER:       pred.udiv.continue:
; SINK-AFTER-NEXT:    [[TMP7:%.*]] = phi <4 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP6]], [[PRED_UDIV_IF]] ]
; SINK-AFTER-NEXT:    [[TMP8:%.*]] = extractelement <4 x i1> [[TMP2]], i32 1
; SINK-AFTER-NEXT:    br i1 [[TMP8]], label [[PRED_UDIV_IF3:%.*]], label [[PRED_UDIV_CONTINUE4:%.*]]
; SINK-AFTER:       pred.udiv.if3:
; SINK-AFTER-NEXT:    [[TMP9:%.*]] = add i32 [[OFFSET_IDX]], -1
; SINK-AFTER-NEXT:    [[TMP10:%.*]] = udiv i32 219220132, [[TMP9]]
; SINK-AFTER-NEXT:    [[TMP11:%.*]] = insertelement <4 x i32> [[TMP7]], i32 [[TMP10]], i32 1
; SINK-AFTER-NEXT:    br label [[PRED_UDIV_CONTINUE4]]
; SINK-AFTER:       pred.udiv.continue4:
; SINK-AFTER-NEXT:    [[TMP12:%.*]] = phi <4 x i32> [ [[TMP7]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP11]], [[PRED_UDIV_IF3]] ]
; SINK-AFTER-NEXT:    [[TMP13:%.*]] = extractelement <4 x i1> [[TMP2]], i32 2
; SINK-AFTER-NEXT:    br i1 [[TMP13]], label [[PRED_UDIV_IF5:%.*]], label [[PRED_UDIV_CONTINUE6:%.*]]
; SINK-AFTER:       pred.udiv.if5:
; SINK-AFTER-NEXT:    [[TMP14:%.*]] = add i32 [[OFFSET_IDX]], -2
; SINK-AFTER-NEXT:    [[TMP15:%.*]] = udiv i32 219220132, [[TMP14]]
; SINK-AFTER-NEXT:    [[TMP16:%.*]] = insertelement <4 x i32> [[TMP12]], i32 [[TMP15]], i32 2
; SINK-AFTER-NEXT:    br label [[PRED_UDIV_CONTINUE6]]
; SINK-AFTER:       pred.udiv.continue6:
; SINK-AFTER-NEXT:    [[TMP17:%.*]] = phi <4 x i32> [ [[TMP12]], [[PRED_UDIV_CONTINUE4]] ], [ [[TMP16]], [[PRED_UDIV_IF5]] ]
; SINK-AFTER-NEXT:    [[TMP18:%.*]] = extractelement <4 x i1> [[TMP2]], i32 3
; SINK-AFTER-NEXT:    br i1 [[TMP18]], label [[PRED_UDIV_IF7:%.*]], label [[PRED_UDIV_CONTINUE8]]
; SINK-AFTER:       pred.udiv.if7:
; SINK-AFTER-NEXT:    [[TMP19:%.*]] = add i32 [[OFFSET_IDX]], -3
; SINK-AFTER-NEXT:    [[TMP20:%.*]] = udiv i32 219220132, [[TMP19]]
; SINK-AFTER-NEXT:    [[TMP21:%.*]] = insertelement <4 x i32> [[TMP17]], i32 [[TMP20]], i32 3
; SINK-AFTER-NEXT:    br label [[PRED_UDIV_CONTINUE8]]
; SINK-AFTER:       pred.udiv.continue8:
; SINK-AFTER-NEXT:    [[TMP22]] = phi <4 x i32> [ [[TMP17]], [[PRED_UDIV_CONTINUE6]] ], [ [[TMP21]], [[PRED_UDIV_IF7]] ]
; SINK-AFTER-NEXT:    [[TMP23:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP22]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; SINK-AFTER-NEXT:    [[TMP24]] = add <4 x i32> [[VEC_PHI]], [[TMP23]]
; SINK-AFTER-NEXT:    [[TMP25:%.*]] = select <4 x i1> [[TMP2]], <4 x i32> [[TMP24]], <4 x i32> [[VEC_PHI]]
; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], 4
; SINK-AFTER-NEXT:    [[TMP26:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; SINK-AFTER-NEXT:    br i1 [[TMP26]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !prof [[PROF24:![0-9]+]], !llvm.loop [[LOOP25:![0-9]+]]
; SINK-AFTER:       middle.block:
; SINK-AFTER-NEXT:    [[TMP27:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP25]])
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP22]], i32 3
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP22]], i32 2
; SINK-AFTER-NEXT:    br i1 true, label [[BB1:%.*]], label [[SCALAR_PH]]
; SINK-AFTER:       scalar.ph:
; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[BB:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[Y]], [[BB]] ]
; SINK-AFTER-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[BB]] ], [ [[TMP27]], [[MIDDLE_BLOCK]] ]
; SINK-AFTER-NEXT:    br label [[BB2:%.*]]
; SINK-AFTER:       bb1:
; SINK-AFTER-NEXT:    [[VAR:%.*]] = phi i32 [ [[VAR6:%.*]], [[BB2]] ], [ [[TMP27]], [[MIDDLE_BLOCK]] ]
; SINK-AFTER-NEXT:    ret i32 [[VAR]]
; SINK-AFTER:       bb2:
; SINK-AFTER-NEXT:    [[VAR3:%.*]] = phi i32 [ [[VAR8:%.*]], [[BB2]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[VAR7:%.*]], [[BB2]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
; SINK-AFTER-NEXT:    [[VAR5:%.*]] = phi i32 [ [[VAR6]], [[BB2]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
; SINK-AFTER-NEXT:    [[VAR6]] = add i32 [[VAR5]], [[SCALAR_RECUR]]
; SINK-AFTER-NEXT:    [[VAR7]] = udiv i32 219220132, [[VAR3]]
; SINK-AFTER-NEXT:    [[VAR8]] = add nsw i32 [[VAR3]], -1
; SINK-AFTER-NEXT:    [[VAR9:%.*]] = icmp slt i32 [[VAR3]], 2
; SINK-AFTER-NEXT:    br i1 [[VAR9]], label [[BB1]], label [[BB2]], !prof [[PROF26:![0-9]+]], !llvm.loop [[LOOP27:![0-9]+]]
;
bb:
  br label %bb2

  bb1:                                              ; preds = %bb2
  %var = phi i32 [ %var6, %bb2 ]
  ret i32 %var

  bb2:                                              ; preds = %bb2, %bb
  %var3 = phi i32 [ %var8, %bb2 ], [ %y, %bb ]
  %var4 = phi i32 [ %var7, %bb2 ], [ 0, %bb ]
  %var5 = phi i32 [ %var6, %bb2 ], [ 0, %bb ]
  %var6 = add i32 %var5, %var4
  %var7 = udiv i32 219220132, %var3
  %var8 = add nsw i32 %var3, -1
  %var9 = icmp slt i32 %var3, 2
  br i1 %var9, label %bb1, label %bb2, !prof !2
}

define i32 @sink_into_replication_region_multiple(i32 *%x, i32 %y) {
;
; CHECK-LABEL: @sink_into_replication_region_multiple(
; UNROLL-NO-IC-LABEL: @sink_into_replication_region_multiple(
; UNROLL-NO-IC-NEXT:  bb:
; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i32 [[Y:%.*]], 1
; UNROLL-NO-IC-NEXT:    [[SMIN:%.*]] = call i32 @llvm.smin.i32(i32 [[Y]], i32 1)
; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = sub i32 [[TMP0]], [[SMIN]]
; UNROLL-NO-IC-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC:       vector.ph:
; UNROLL-NO-IC-NEXT:    [[N_RND_UP:%.*]] = add i32 [[TMP1]], 7
; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[N_RND_UP]], 8
; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i32 [[N_RND_UP]], [[N_MOD_VF]]
; UNROLL-NO-IC-NEXT:    [[IND_END:%.*]] = sub i32 [[Y]], [[N_VEC]]
; UNROLL-NO-IC-NEXT:    [[TRIP_COUNT_MINUS_1:%.*]] = sub i32 [[TMP1]], 1
; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[TRIP_COUNT_MINUS_1]], i32 0
; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC:       vector.body:
; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE31:%.*]] ]
; UNROLL-NO-IC-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[PRED_STORE_CONTINUE31]] ]
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP43:%.*]], [[PRED_STORE_CONTINUE31]] ]
; UNROLL-NO-IC-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP46:%.*]], [[PRED_STORE_CONTINUE31]] ]
; UNROLL-NO-IC-NEXT:    [[VEC_PHI3:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP47:%.*]], [[PRED_STORE_CONTINUE31]] ]
; UNROLL-NO-IC-NEXT:    [[STEP_ADD:%.*]] = add <4 x i32> [[VEC_IND]], <i32 4, i32 4, i32 4, i32 4>
; UNROLL-NO-IC-NEXT:    [[OFFSET_IDX:%.*]] = sub i32 [[Y]], [[INDEX]]
; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add i32 [[OFFSET_IDX]], 0
; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add i32 [[OFFSET_IDX]], -1
; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = add i32 [[OFFSET_IDX]], -2
; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = add i32 [[OFFSET_IDX]], -3
; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = add i32 [[OFFSET_IDX]], -4
; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = add i32 [[OFFSET_IDX]], -5
; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = add i32 [[OFFSET_IDX]], -6
; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = add i32 [[OFFSET_IDX]], -7
; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = icmp ule <4 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = icmp ule <4 x i32> [[STEP_ADD]], [[BROADCAST_SPLAT]]
; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = extractelement <4 x i1> [[TMP10]], i32 0
; UNROLL-NO-IC-NEXT:    br i1 [[TMP12]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
; UNROLL-NO-IC:       pred.udiv.if:
; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = udiv i32 219220132, [[TMP2]]
; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = insertelement <4 x i32> poison, i32 [[TMP13]], i32 0
; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE]]
; UNROLL-NO-IC:       pred.udiv.continue:
; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = phi <4 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP14]], [[PRED_UDIV_IF]] ]
; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = extractelement <4 x i1> [[TMP10]], i32 1
; UNROLL-NO-IC-NEXT:    br i1 [[TMP16]], label [[PRED_UDIV_IF4:%.*]], label [[PRED_UDIV_CONTINUE5:%.*]]
; UNROLL-NO-IC:       pred.udiv.if4:
; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = udiv i32 219220132, [[TMP3]]
; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = insertelement <4 x i32> [[TMP15]], i32 [[TMP17]], i32 1
; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE5]]
; UNROLL-NO-IC:       pred.udiv.continue5:
; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = phi <4 x i32> [ [[TMP15]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP18]], [[PRED_UDIV_IF4]] ]
; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = extractelement <4 x i1> [[TMP10]], i32 2
; UNROLL-NO-IC-NEXT:    br i1 [[TMP20]], label [[PRED_UDIV_IF6:%.*]], label [[PRED_UDIV_CONTINUE7:%.*]]
; UNROLL-NO-IC:       pred.udiv.if6:
; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = udiv i32 219220132, [[TMP4]]
; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = insertelement <4 x i32> [[TMP19]], i32 [[TMP21]], i32 2
; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE7]]
; UNROLL-NO-IC:       pred.udiv.continue7:
; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = phi <4 x i32> [ [[TMP19]], [[PRED_UDIV_CONTINUE5]] ], [ [[TMP22]], [[PRED_UDIV_IF6]] ]
; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = extractelement <4 x i1> [[TMP10]], i32 3
; UNROLL-NO-IC-NEXT:    br i1 [[TMP24]], label [[PRED_UDIV_IF8:%.*]], label [[PRED_UDIV_CONTINUE9:%.*]]
; UNROLL-NO-IC:       pred.udiv.if8:
; UNROLL-NO-IC-NEXT:    [[TMP25:%.*]] = udiv i32 219220132, [[TMP5]]
; UNROLL-NO-IC-NEXT:    [[TMP26:%.*]] = insertelement <4 x i32> [[TMP23]], i32 [[TMP25]], i32 3
; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE9]]
; UNROLL-NO-IC:       pred.udiv.continue9:
; UNROLL-NO-IC-NEXT:    [[TMP27:%.*]] = phi <4 x i32> [ [[TMP23]], [[PRED_UDIV_CONTINUE7]] ], [ [[TMP26]], [[PRED_UDIV_IF8]] ]
; UNROLL-NO-IC-NEXT:    [[TMP28:%.*]] = extractelement <4 x i1> [[TMP11]], i32 0
; UNROLL-NO-IC-NEXT:    br i1 [[TMP28]], label [[PRED_UDIV_IF10:%.*]], label [[PRED_UDIV_CONTINUE11:%.*]]
; UNROLL-NO-IC:       pred.udiv.if10:
; UNROLL-NO-IC-NEXT:    [[TMP29:%.*]] = udiv i32 219220132, [[TMP6]]
; UNROLL-NO-IC-NEXT:    [[TMP30:%.*]] = insertelement <4 x i32> poison, i32 [[TMP29]], i32 0
; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE11]]
; UNROLL-NO-IC:       pred.udiv.continue11:
; UNROLL-NO-IC-NEXT:    [[TMP31:%.*]] = phi <4 x i32> [ poison, [[PRED_UDIV_CONTINUE9]] ], [ [[TMP30]], [[PRED_UDIV_IF10]] ]
; UNROLL-NO-IC-NEXT:    [[TMP32:%.*]] = extractelement <4 x i1> [[TMP11]], i32 1
; UNROLL-NO-IC-NEXT:    br i1 [[TMP32]], label [[PRED_UDIV_IF12:%.*]], label [[PRED_UDIV_CONTINUE13:%.*]]
; UNROLL-NO-IC:       pred.udiv.if12:
; UNROLL-NO-IC-NEXT:    [[TMP33:%.*]] = udiv i32 219220132, [[TMP7]]
; UNROLL-NO-IC-NEXT:    [[TMP34:%.*]] = insertelement <4 x i32> [[TMP31]], i32 [[TMP33]], i32 1
; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE13]]
; UNROLL-NO-IC:       pred.udiv.continue13:
; UNROLL-NO-IC-NEXT:    [[TMP35:%.*]] = phi <4 x i32> [ [[TMP31]], [[PRED_UDIV_CONTINUE11]] ], [ [[TMP34]], [[PRED_UDIV_IF12]] ]
; UNROLL-NO-IC-NEXT:    [[TMP36:%.*]] = extractelement <4 x i1> [[TMP11]], i32 2
; UNROLL-NO-IC-NEXT:    br i1 [[TMP36]], label [[PRED_UDIV_IF14:%.*]], label [[PRED_UDIV_CONTINUE15:%.*]]
; UNROLL-NO-IC:       pred.udiv.if14:
; UNROLL-NO-IC-NEXT:    [[TMP37:%.*]] = udiv i32 219220132, [[TMP8]]
; UNROLL-NO-IC-NEXT:    [[TMP38:%.*]] = insertelement <4 x i32> [[TMP35]], i32 [[TMP37]], i32 2
; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE15]]
; UNROLL-NO-IC:       pred.udiv.continue15:
; UNROLL-NO-IC-NEXT:    [[TMP39:%.*]] = phi <4 x i32> [ [[TMP35]], [[PRED_UDIV_CONTINUE13]] ], [ [[TMP38]], [[PRED_UDIV_IF14]] ]
; UNROLL-NO-IC-NEXT:    [[TMP40:%.*]] = extractelement <4 x i1> [[TMP11]], i32 3
; UNROLL-NO-IC-NEXT:    br i1 [[TMP40]], label [[PRED_UDIV_IF16:%.*]], label [[PRED_UDIV_CONTINUE17:%.*]]
; UNROLL-NO-IC:       pred.udiv.if16:
; UNROLL-NO-IC-NEXT:    [[TMP41:%.*]] = udiv i32 219220132, [[TMP9]]
; UNROLL-NO-IC-NEXT:    [[TMP42:%.*]] = insertelement <4 x i32> [[TMP39]], i32 [[TMP41]], i32 3
; UNROLL-NO-IC-NEXT:    br label [[PRED_UDIV_CONTINUE17]]
; UNROLL-NO-IC:       pred.udiv.continue17:
; UNROLL-NO-IC-NEXT:    [[TMP43]] = phi <4 x i32> [ [[TMP39]], [[PRED_UDIV_CONTINUE15]] ], [ [[TMP42]], [[PRED_UDIV_IF16]] ]
; UNROLL-NO-IC-NEXT:    [[TMP44:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP27]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; UNROLL-NO-IC-NEXT:    [[TMP45:%.*]] = shufflevector <4 x i32> [[TMP27]], <4 x i32> [[TMP43]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; UNROLL-NO-IC-NEXT:    [[TMP46]] = add <4 x i32> [[VEC_PHI]], [[TMP44]]
; UNROLL-NO-IC-NEXT:    [[TMP47]] = add <4 x i32> [[VEC_PHI3]], [[TMP45]]
; UNROLL-NO-IC-NEXT:    [[TMP48:%.*]] = extractelement <4 x i1> [[TMP10]], i32 0
; UNROLL-NO-IC-NEXT:    br i1 [[TMP48]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
; UNROLL-NO-IC:       pred.store.if:
; UNROLL-NO-IC-NEXT:    [[TMP49:%.*]] = add i32 [[INDEX]], 0
; UNROLL-NO-IC-NEXT:    [[TMP50:%.*]] = getelementptr inbounds i32, i32* [[X:%.*]], i32 [[TMP49]]
; UNROLL-NO-IC-NEXT:    store i32 [[TMP2]], i32* [[TMP50]], align 4
; UNROLL-NO-IC-NEXT:    br label [[PRED_STORE_CONTINUE]]
; UNROLL-NO-IC:       pred.store.continue:
; UNROLL-NO-IC-NEXT:    [[TMP51:%.*]] = extractelement <4 x i1> [[TMP10]], i32 1
; UNROLL-NO-IC-NEXT:    br i1 [[TMP51]], label [[PRED_STORE_IF18:%.*]], label [[PRED_STORE_CONTINUE19:%.*]]
; UNROLL-NO-IC:       pred.store.if18:
; UNROLL-NO-IC-NEXT:    [[TMP52:%.*]] = add i32 [[INDEX]], 1
; UNROLL-NO-IC-NEXT:    [[TMP53:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[TMP52]]
; UNROLL-NO-IC-NEXT:    store i32 [[TMP3]], i32* [[TMP53]], align 4
; UNROLL-NO-IC-NEXT:    br label [[PRED_STORE_CONTINUE19]]
; UNROLL-NO-IC:       pred.store.continue19:
; UNROLL-NO-IC-NEXT:    [[TMP54:%.*]] = extractelement <4 x i1> [[TMP10]], i32 2
; UNROLL-NO-IC-NEXT:    br i1 [[TMP54]], label [[PRED_STORE_IF20:%.*]], label [[PRED_STORE_CONTINUE21:%.*]]
; UNROLL-NO-IC:       pred.store.if20:
; UNROLL-NO-IC-NEXT:    [[TMP55:%.*]] = add i32 [[INDEX]], 2
; UNROLL-NO-IC-NEXT:    [[TMP56:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[TMP55]]
; UNROLL-NO-IC-NEXT:    store i32 [[TMP4]], i32* [[TMP56]], align 4
; UNROLL-NO-IC-NEXT:    br label [[PRED_STORE_CONTINUE21]]
; UNROLL-NO-IC:       pred.store.continue21:
; UNROLL-NO-IC-NEXT:    [[TMP57:%.*]] = extractelement <4 x i1> [[TMP10]], i32 3
; UNROLL-NO-IC-NEXT:    br i1 [[TMP57]], label [[PRED_STORE_IF22:%.*]], label [[PRED_STORE_CONTINUE23:%.*]]
; UNROLL-NO-IC:       pred.store.if22:
; UNROLL-NO-IC-NEXT:    [[TMP58:%.*]] = add i32 [[INDEX]], 3
; UNROLL-NO-IC-NEXT:    [[TMP59:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[TMP58]]
; UNROLL-NO-IC-NEXT:    store i32 [[TMP5]], i32* [[TMP59]], align 4
; UNROLL-NO-IC-NEXT:    br label [[PRED_STORE_CONTINUE23]]
; UNROLL-NO-IC:       pred.store.continue23:
; UNROLL-NO-IC-NEXT:    [[TMP60:%.*]] = extractelement <4 x i1> [[TMP11]], i32 0
; UNROLL-NO-IC-NEXT:    br i1 [[TMP60]], label [[PRED_STORE_IF24:%.*]], label [[PRED_STORE_CONTINUE25:%.*]]
; UNROLL-NO-IC:       pred.store.if24:
; UNROLL-NO-IC-NEXT:    [[TMP61:%.*]] = add i32 [[INDEX]], 4
; UNROLL-NO-IC-NEXT:    [[TMP62:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[TMP61]]
; UNROLL-NO-IC-NEXT:    store i32 [[TMP6]], i32* [[TMP62]], align 4
; UNROLL-NO-IC-NEXT:    br label [[PRED_STORE_CONTINUE25]]
; UNROLL-NO-IC:       pred.store.continue25:
; UNROLL-NO-IC-NEXT:    [[TMP63:%.*]] = extractelement <4 x i1> [[TMP11]], i32 1
; UNROLL-NO-IC-NEXT:    br i1 [[TMP63]], label [[PRED_STORE_IF26:%.*]], label [[PRED_STORE_CONTINUE27:%.*]]
; UNROLL-NO-IC:       pred.store.if26:
; UNROLL-NO-IC-NEXT:    [[TMP64:%.*]] = add i32 [[INDEX]], 5
; UNROLL-NO-IC-NEXT:    [[TMP65:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[TMP64]]
; UNROLL-NO-IC-NEXT:    store i32 [[TMP7]], i32* [[TMP65]], align 4
; UNROLL-NO-IC-NEXT:    br label [[PRED_STORE_CONTINUE27]]
; UNROLL-NO-IC:       pred.store.continue27:
; UNROLL-NO-IC-NEXT:    [[TMP66:%.*]] = extractelement <4 x i1> [[TMP11]], i32 2
; UNROLL-NO-IC-NEXT:    br i1 [[TMP66]], label [[PRED_STORE_IF28:%.*]], label [[PRED_STORE_CONTINUE29:%.*]]
; UNROLL-NO-IC:       pred.store.if28:
; UNROLL-NO-IC-NEXT:    [[TMP67:%.*]] = add i32 [[INDEX]], 6
; UNROLL-NO-IC-NEXT:    [[TMP68:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[TMP67]]
; UNROLL-NO-IC-NEXT:    store i32 [[TMP8]], i32* [[TMP68]], align 4
; UNROLL-NO-IC-NEXT:    br label [[PRED_STORE_CONTINUE29]]
; UNROLL-NO-IC:       pred.store.continue29:
; UNROLL-NO-IC-NEXT:    [[TMP69:%.*]] = extractelement <4 x i1> [[TMP11]], i32 3
; UNROLL-NO-IC-NEXT:    br i1 [[TMP69]], label [[PRED_STORE_IF30:%.*]], label [[PRED_STORE_CONTINUE31]]
; UNROLL-NO-IC:       pred.store.if30:
; UNROLL-NO-IC-NEXT:    [[TMP70:%.*]] = add i32 [[INDEX]], 7
; UNROLL-NO-IC-NEXT:    [[TMP71:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[TMP70]]
; UNROLL-NO-IC-NEXT:    store i32 [[TMP9]], i32* [[TMP71]], align 4
; UNROLL-NO-IC-NEXT:    br label [[PRED_STORE_CONTINUE31]]
; UNROLL-NO-IC:       pred.store.continue31:
; UNROLL-NO-IC-NEXT:    [[TMP72:%.*]] = select <4 x i1> [[TMP10]], <4 x i32> [[TMP46]], <4 x i32> [[VEC_PHI]]
; UNROLL-NO-IC-NEXT:    [[TMP73:%.*]] = select <4 x i1> [[TMP11]], <4 x i32> [[TMP47]], <4 x i32> [[VEC_PHI3]]
; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], 8
; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[STEP_ADD]], <i32 4, i32 4, i32 4, i32 4>
; UNROLL-NO-IC-NEXT:    [[TMP74:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-IC-NEXT:    br i1 [[TMP74]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !prof [[PROF24]], !llvm.loop [[LOOP28:![0-9]+]]
; UNROLL-NO-IC:       middle.block:
; UNROLL-NO-IC-NEXT:    [[BIN_RDX:%.*]] = add <4 x i32> [[TMP73]], [[TMP72]]
; UNROLL-NO-IC-NEXT:    [[TMP75:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX]])
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP43]], i32 3
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP43]], i32 2
; UNROLL-NO-IC-NEXT:    br i1 true, label [[BB1:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC:       scalar.ph:
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[BB:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[Y]], [[BB]] ]
; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[BB]] ]
; UNROLL-NO-IC-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[BB]] ], [ [[TMP75]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT:    br label [[BB2:%.*]]
; UNROLL-NO-IC:       bb1:
; UNROLL-NO-IC-NEXT:    [[VAR:%.*]] = phi i32 [ [[VAR6:%.*]], [[BB2]] ], [ [[TMP75]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT:    ret i32 [[VAR]]
; UNROLL-NO-IC:       bb2:
; UNROLL-NO-IC-NEXT:    [[VAR3:%.*]] = phi i32 [ [[VAR8:%.*]], [[BB2]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; UNROLL-NO-IC-NEXT:    [[IV:%.*]] = phi i32 [ [[IV_NEXT:%.*]], [[BB2]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[VAR7:%.*]], [[BB2]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
; UNROLL-NO-IC-NEXT:    [[VAR5:%.*]] = phi i32 [ [[VAR6]], [[BB2]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
; UNROLL-NO-IC-NEXT:    [[G:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[IV]]
; UNROLL-NO-IC-NEXT:    [[VAR6]] = add i32 [[VAR5]], [[SCALAR_RECUR]]
; UNROLL-NO-IC-NEXT:    [[VAR7]] = udiv i32 219220132, [[VAR3]]
; UNROLL-NO-IC-NEXT:    store i32 [[VAR3]], i32* [[G]], align 4
; UNROLL-NO-IC-NEXT:    [[VAR8]] = add nsw i32 [[VAR3]], -1
; UNROLL-NO-IC-NEXT:    [[IV_NEXT]] = add nsw i32 [[IV]], 1
; UNROLL-NO-IC-NEXT:    [[VAR9:%.*]] = icmp slt i32 [[VAR3]], 2
; UNROLL-NO-IC-NEXT:    br i1 [[VAR9]], label [[BB1]], label [[BB2]], !prof [[PROF26]], !llvm.loop [[LOOP29:![0-9]+]]
;
; UNROLL-NO-VF-LABEL: @sink_into_replication_region_multiple(
; UNROLL-NO-VF-NEXT:  bb:
; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = add i32 [[Y:%.*]], 1
; UNROLL-NO-VF-NEXT:    [[SMIN:%.*]] = call i32 @llvm.smin.i32(i32 [[Y]], i32 1)
; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = sub i32 [[TMP0]], [[SMIN]]
; UNROLL-NO-VF-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-VF:       vector.ph:
; UNROLL-NO-VF-NEXT:    [[N_RND_UP:%.*]] = add i32 [[TMP1]], 1
; UNROLL-NO-VF-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[N_RND_UP]], 2
; UNROLL-NO-VF-NEXT:    [[N_VEC:%.*]] = sub i32 [[N_RND_UP]], [[N_MOD_VF]]
; UNROLL-NO-VF-NEXT:    [[IND_END:%.*]] = sub i32 [[Y]], [[N_VEC]]
; UNROLL-NO-VF-NEXT:    [[TRIP_COUNT_MINUS_1:%.*]] = sub i32 [[TMP1]], 1
; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-VF:       vector.body:
; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE10:%.*]] ]
; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[PRED_STORE_CONTINUE10]] ]
; UNROLL-NO-VF-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[PRED_STORE_CONTINUE10]] ]
; UNROLL-NO-VF-NEXT:    [[VEC_PHI2:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[PRED_STORE_CONTINUE10]] ]
; UNROLL-NO-VF-NEXT:    [[OFFSET_IDX:%.*]] = sub i32 [[Y]], [[INDEX]]
; UNROLL-NO-VF-NEXT:    [[INDUCTION4:%.*]] = add i32 [[OFFSET_IDX]], 0
; UNROLL-NO-VF-NEXT:    [[INDUCTION5:%.*]] = add i32 [[OFFSET_IDX]], -1
; UNROLL-NO-VF-NEXT:    [[VEC_IV:%.*]] = add i32 [[INDEX]], 0
; UNROLL-NO-VF-NEXT:    [[VEC_IV6:%.*]] = add i32 [[INDEX]], 1
; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = icmp ule i32 [[VEC_IV]], [[TRIP_COUNT_MINUS_1]]
; UNROLL-NO-VF-NEXT:    [[TMP3:%.*]] = icmp ule i32 [[VEC_IV6]], [[TRIP_COUNT_MINUS_1]]
; UNROLL-NO-VF-NEXT:    br i1 [[TMP2]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
; UNROLL-NO-VF:       pred.udiv.if:
; UNROLL-NO-VF-NEXT:    [[TMP4:%.*]] = udiv i32 219220132, [[INDUCTION4]]
; UNROLL-NO-VF-NEXT:    br label [[PRED_UDIV_CONTINUE]]
; UNROLL-NO-VF:       pred.udiv.continue:
; UNROLL-NO-VF-NEXT:    [[TMP5:%.*]] = phi i32 [ poison, [[VECTOR_BODY]] ], [ [[TMP4]], [[PRED_UDIV_IF]] ]
; UNROLL-NO-VF-NEXT:    br i1 [[TMP3]], label [[PRED_UDIV_IF7:%.*]], label [[PRED_UDIV_CONTINUE8:%.*]]
; UNROLL-NO-VF:       pred.udiv.if7:
; UNROLL-NO-VF-NEXT:    [[TMP6:%.*]] = udiv i32 219220132, [[INDUCTION5]]
; UNROLL-NO-VF-NEXT:    br label [[PRED_UDIV_CONTINUE8]]
; UNROLL-NO-VF:       pred.udiv.continue8:
; UNROLL-NO-VF-NEXT:    [[TMP7]] = phi i32 [ poison, [[PRED_UDIV_CONTINUE]] ], [ [[TMP6]], [[PRED_UDIV_IF7]] ]
; UNROLL-NO-VF-NEXT:    [[TMP8]] = add i32 [[VEC_PHI]], [[VECTOR_RECUR]]
; UNROLL-NO-VF-NEXT:    [[TMP9]] = add i32 [[VEC_PHI2]], [[TMP5]]
; UNROLL-NO-VF-NEXT:    br i1 [[TMP2]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
; UNROLL-NO-VF:       pred.store.if:
; UNROLL-NO-VF-NEXT:    [[INDUCTION:%.*]] = add i32 [[INDEX]], 0
; UNROLL-NO-VF-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, i32* [[X:%.*]], i32 [[INDUCTION]]
; UNROLL-NO-VF-NEXT:    store i32 [[INDUCTION4]], i32* [[TMP10]], align 4
; UNROLL-NO-VF-NEXT:    br label [[PRED_STORE_CONTINUE]]
; UNROLL-NO-VF:       pred.store.continue:
; UNROLL-NO-VF-NEXT:    br i1 [[TMP3]], label [[PRED_STORE_IF9:%.*]], label [[PRED_STORE_CONTINUE10]]
; UNROLL-NO-VF:       pred.store.if9:
; UNROLL-NO-VF-NEXT:    [[INDUCTION3:%.*]] = add i32 [[INDEX]], 1
; UNROLL-NO-VF-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[INDUCTION3]]
; UNROLL-NO-VF-NEXT:    store i32 [[INDUCTION5]], i32* [[TMP11]], align 4
; UNROLL-NO-VF-NEXT:    br label [[PRED_STORE_CONTINUE10]]
; UNROLL-NO-VF:       pred.store.continue10:
; UNROLL-NO-VF-NEXT:    [[TMP12:%.*]] = select i1 [[TMP2]], i32 [[TMP8]], i32 [[VEC_PHI]]
; UNROLL-NO-VF-NEXT:    [[TMP13:%.*]] = select i1 [[TMP3]], i32 [[TMP9]], i32 [[VEC_PHI2]]
; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], 2
; UNROLL-NO-VF-NEXT:    [[TMP14:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; UNROLL-NO-VF-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !prof [[PROF23]], !llvm.loop [[LOOP27:![0-9]+]]
; UNROLL-NO-VF:       middle.block:
; UNROLL-NO-VF-NEXT:    [[BIN_RDX:%.*]] = add i32 [[TMP13]], [[TMP12]]
; UNROLL-NO-VF-NEXT:    br i1 true, label [[BB1:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-VF:       scalar.ph:
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[BB:%.*]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[Y]], [[BB]] ]
; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[BB]] ]
; UNROLL-NO-VF-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[BB]] ], [ [[BIN_RDX]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-VF-NEXT:    br label [[BB2:%.*]]
; UNROLL-NO-VF:       bb1:
; UNROLL-NO-VF-NEXT:    [[VAR:%.*]] = phi i32 [ [[VAR6:%.*]], [[BB2]] ], [ [[BIN_RDX]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-VF-NEXT:    ret i32 [[VAR]]
; UNROLL-NO-VF:       bb2:
; UNROLL-NO-VF-NEXT:    [[VAR3:%.*]] = phi i32 [ [[VAR8:%.*]], [[BB2]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; UNROLL-NO-VF-NEXT:    [[IV:%.*]] = phi i32 [ [[IV_NEXT:%.*]], [[BB2]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[VAR7:%.*]], [[BB2]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
; UNROLL-NO-VF-NEXT:    [[VAR5:%.*]] = phi i32 [ [[VAR6]], [[BB2]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
; UNROLL-NO-VF-NEXT:    [[G:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[IV]]
; UNROLL-NO-VF-NEXT:    [[VAR6]] = add i32 [[VAR5]], [[SCALAR_RECUR]]
; UNROLL-NO-VF-NEXT:    [[VAR7]] = udiv i32 219220132, [[VAR3]]
; UNROLL-NO-VF-NEXT:    store i32 [[VAR3]], i32* [[G]], align 4
; UNROLL-NO-VF-NEXT:    [[VAR8]] = add nsw i32 [[VAR3]], -1
; UNROLL-NO-VF-NEXT:    [[IV_NEXT]] = add nsw i32 [[IV]], 1
; UNROLL-NO-VF-NEXT:    [[VAR9:%.*]] = icmp slt i32 [[VAR3]], 2
; UNROLL-NO-VF-NEXT:    br i1 [[VAR9]], label [[BB1]], label [[BB2]], !prof [[PROF25]], !llvm.loop [[LOOP28:![0-9]+]]
;
; SINK-AFTER-LABEL: @sink_into_replication_region_multiple(
; SINK-AFTER-NEXT:  bb:
; SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i32 [[Y:%.*]], 1
; SINK-AFTER-NEXT:    [[SMIN:%.*]] = call i32 @llvm.smin.i32(i32 [[Y]], i32 1)
; SINK-AFTER-NEXT:    [[TMP1:%.*]] = sub i32 [[TMP0]], [[SMIN]]
; SINK-AFTER-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; SINK-AFTER:       vector.ph:
; SINK-AFTER-NEXT:    [[N_RND_UP:%.*]] = add i32 [[TMP1]], 3
; SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[N_RND_UP]], 4
; SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i32 [[N_RND_UP]], [[N_MOD_VF]]
; SINK-AFTER-NEXT:    [[IND_END:%.*]] = sub i32 [[Y]], [[N_VEC]]
; SINK-AFTER-NEXT:    [[TRIP_COUNT_MINUS_1:%.*]] = sub i32 [[TMP1]], 1
; SINK-AFTER-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[TRIP_COUNT_MINUS_1]], i32 0
; SINK-AFTER-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
; SINK-AFTER:       vector.body:
; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE13:%.*]] ]
; SINK-AFTER-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[PRED_STORE_CONTINUE13]] ]
; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP22:%.*]], [[PRED_STORE_CONTINUE13]] ]
; SINK-AFTER-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP24:%.*]], [[PRED_STORE_CONTINUE13]] ]
; SINK-AFTER-NEXT:    [[OFFSET_IDX:%.*]] = sub i32 [[Y]], [[INDEX]]
; SINK-AFTER-NEXT:    [[TMP2:%.*]] = add i32 [[OFFSET_IDX]], 0
; SINK-AFTER-NEXT:    [[TMP3:%.*]] = add i32 [[OFFSET_IDX]], -1
; SINK-AFTER-NEXT:    [[TMP4:%.*]] = add i32 [[OFFSET_IDX]], -2
; SINK-AFTER-NEXT:    [[TMP5:%.*]] = add i32 [[OFFSET_IDX]], -3
; SINK-AFTER-NEXT:    [[TMP6:%.*]] = icmp ule <4 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
; SINK-AFTER-NEXT:    [[TMP7:%.*]] = extractelement <4 x i1> [[TMP6]], i32 0
; SINK-AFTER-NEXT:    br i1 [[TMP7]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
; SINK-AFTER:       pred.udiv.if:
; SINK-AFTER-NEXT:    [[TMP8:%.*]] = udiv i32 219220132, [[TMP2]]
; SINK-AFTER-NEXT:    [[TMP9:%.*]] = insertelement <4 x i32> poison, i32 [[TMP8]], i32 0
; SINK-AFTER-NEXT:    br label [[PRED_UDIV_CONTINUE]]
; SINK-AFTER:       pred.udiv.continue:
; SINK-AFTER-NEXT:    [[TMP10:%.*]] = phi <4 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP9]], [[PRED_UDIV_IF]] ]
; SINK-AFTER-NEXT:    [[TMP11:%.*]] = extractelement <4 x i1> [[TMP6]], i32 1
; SINK-AFTER-NEXT:    br i1 [[TMP11]], label [[PRED_UDIV_IF2:%.*]], label [[PRED_UDIV_CONTINUE3:%.*]]
; SINK-AFTER:       pred.udiv.if2:
; SINK-AFTER-NEXT:    [[TMP12:%.*]] = udiv i32 219220132, [[TMP3]]
; SINK-AFTER-NEXT:    [[TMP13:%.*]] = insertelement <4 x i32> [[TMP10]], i32 [[TMP12]], i32 1
; SINK-AFTER-NEXT:    br label [[PRED_UDIV_CONTINUE3]]
; SINK-AFTER:       pred.udiv.continue3:
; SINK-AFTER-NEXT:    [[TMP14:%.*]] = phi <4 x i32> [ [[TMP10]], [[PRED_UDIV_CONTINUE]] ], [ [[TMP13]], [[PRED_UDIV_IF2]] ]
; SINK-AFTER-NEXT:    [[TMP15:%.*]] = extractelement <4 x i1> [[TMP6]], i32 2
; SINK-AFTER-NEXT:    br i1 [[TMP15]], label [[PRED_UDIV_IF4:%.*]], label [[PRED_UDIV_CONTINUE5:%.*]]
; SINK-AFTER:       pred.udiv.if4:
; SINK-AFTER-NEXT:    [[TMP16:%.*]] = udiv i32 219220132, [[TMP4]]
; SINK-AFTER-NEXT:    [[TMP17:%.*]] = insertelement <4 x i32> [[TMP14]], i32 [[TMP16]], i32 2
; SINK-AFTER-NEXT:    br label [[PRED_UDIV_CONTINUE5]]
; SINK-AFTER:       pred.udiv.continue5:
; SINK-AFTER-NEXT:    [[TMP18:%.*]] = phi <4 x i32> [ [[TMP14]], [[PRED_UDIV_CONTINUE3]] ], [ [[TMP17]], [[PRED_UDIV_IF4]] ]
; SINK-AFTER-NEXT:    [[TMP19:%.*]] = extractelement <4 x i1> [[TMP6]], i32 3
; SINK-AFTER-NEXT:    br i1 [[TMP19]], label [[PRED_UDIV_IF6:%.*]], label [[PRED_UDIV_CONTINUE7:%.*]]
; SINK-AFTER:       pred.udiv.if6:
; SINK-AFTER-NEXT:    [[TMP20:%.*]] = udiv i32 219220132, [[TMP5]]
; SINK-AFTER-NEXT:    [[TMP21:%.*]] = insertelement <4 x i32> [[TMP18]], i32 [[TMP20]], i32 3
; SINK-AFTER-NEXT:    br label [[PRED_UDIV_CONTINUE7]]
; SINK-AFTER:       pred.udiv.continue7:
; SINK-AFTER-NEXT:    [[TMP22]] = phi <4 x i32> [ [[TMP18]], [[PRED_UDIV_CONTINUE5]] ], [ [[TMP21]], [[PRED_UDIV_IF6]] ]
; SINK-AFTER-NEXT:    [[TMP23:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP22]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; SINK-AFTER-NEXT:    [[TMP24]] = add <4 x i32> [[VEC_PHI]], [[TMP23]]
; SINK-AFTER-NEXT:    [[TMP25:%.*]] = extractelement <4 x i1> [[TMP6]], i32 0
; SINK-AFTER-NEXT:    br i1 [[TMP25]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
; SINK-AFTER:       pred.store.if:
; SINK-AFTER-NEXT:    [[TMP26:%.*]] = add i32 [[INDEX]], 0
; SINK-AFTER-NEXT:    [[TMP27:%.*]] = getelementptr inbounds i32, i32* [[X:%.*]], i32 [[TMP26]]
; SINK-AFTER-NEXT:    store i32 [[TMP2]], i32* [[TMP27]], align 4
; SINK-AFTER-NEXT:    br label [[PRED_STORE_CONTINUE]]
; SINK-AFTER:       pred.store.continue:
; SINK-AFTER-NEXT:    [[TMP28:%.*]] = extractelement <4 x i1> [[TMP6]], i32 1
; SINK-AFTER-NEXT:    br i1 [[TMP28]], label [[PRED_STORE_IF8:%.*]], label [[PRED_STORE_CONTINUE9:%.*]]
; SINK-AFTER:       pred.store.if8:
; SINK-AFTER-NEXT:    [[TMP29:%.*]] = add i32 [[INDEX]], 1
; SINK-AFTER-NEXT:    [[TMP30:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[TMP29]]
; SINK-AFTER-NEXT:    store i32 [[TMP3]], i32* [[TMP30]], align 4
; SINK-AFTER-NEXT:    br label [[PRED_STORE_CONTINUE9]]
; SINK-AFTER:       pred.store.continue9:
; SINK-AFTER-NEXT:    [[TMP31:%.*]] = extractelement <4 x i1> [[TMP6]], i32 2
; SINK-AFTER-NEXT:    br i1 [[TMP31]], label [[PRED_STORE_IF10:%.*]], label [[PRED_STORE_CONTINUE11:%.*]]
; SINK-AFTER:       pred.store.if10:
; SINK-AFTER-NEXT:    [[TMP32:%.*]] = add i32 [[INDEX]], 2
; SINK-AFTER-NEXT:    [[TMP33:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[TMP32]]
; SINK-AFTER-NEXT:    store i32 [[TMP4]], i32* [[TMP33]], align 4
; SINK-AFTER-NEXT:    br label [[PRED_STORE_CONTINUE11]]
; SINK-AFTER:       pred.store.continue11:
; SINK-AFTER-NEXT:    [[TMP34:%.*]] = extractelement <4 x i1> [[TMP6]], i32 3
; SINK-AFTER-NEXT:    br i1 [[TMP34]], label [[PRED_STORE_IF12:%.*]], label [[PRED_STORE_CONTINUE13]]
; SINK-AFTER:       pred.store.if12:
; SINK-AFTER-NEXT:    [[TMP35:%.*]] = add i32 [[INDEX]], 3
; SINK-AFTER-NEXT:    [[TMP36:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[TMP35]]
; SINK-AFTER-NEXT:    store i32 [[TMP5]], i32* [[TMP36]], align 4
; SINK-AFTER-NEXT:    br label [[PRED_STORE_CONTINUE13]]
; SINK-AFTER:       pred.store.continue13:
; SINK-AFTER-NEXT:    [[TMP37:%.*]] = select <4 x i1> [[TMP6]], <4 x i32> [[TMP24]], <4 x i32> [[VEC_PHI]]
; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], 4
; SINK-AFTER-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 4, i32 4, i32 4, i32 4>
; SINK-AFTER-NEXT:    [[TMP38:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; SINK-AFTER-NEXT:    br i1 [[TMP38]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !prof [[PROF24]], !llvm.loop [[LOOP28:![0-9]+]]
; SINK-AFTER:       middle.block:
; SINK-AFTER-NEXT:    [[TMP39:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP37]])
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP22]], i32 3
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP22]], i32 2
; SINK-AFTER-NEXT:    br i1 true, label [[BB1:%.*]], label [[SCALAR_PH]]
; SINK-AFTER:       scalar.ph:
; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[BB:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[Y]], [[BB]] ]
; SINK-AFTER-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[BB]] ]
; SINK-AFTER-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[BB]] ], [ [[TMP39]], [[MIDDLE_BLOCK]] ]
; SINK-AFTER-NEXT:    br label [[BB2:%.*]]
; SINK-AFTER:       bb1:
; SINK-AFTER-NEXT:    [[VAR:%.*]] = phi i32 [ [[VAR6:%.*]], [[BB2]] ], [ [[TMP39]], [[MIDDLE_BLOCK]] ]
; SINK-AFTER-NEXT:    ret i32 [[VAR]]
; SINK-AFTER:       bb2:
; SINK-AFTER-NEXT:    [[VAR3:%.*]] = phi i32 [ [[VAR8:%.*]], [[BB2]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
; SINK-AFTER-NEXT:    [[IV:%.*]] = phi i32 [ [[IV_NEXT:%.*]], [[BB2]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
; SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[VAR7:%.*]], [[BB2]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
; SINK-AFTER-NEXT:    [[VAR5:%.*]] = phi i32 [ [[VAR6]], [[BB2]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
; SINK-AFTER-NEXT:    [[G:%.*]] = getelementptr inbounds i32, i32* [[X]], i32 [[IV]]
; SINK-AFTER-NEXT:    [[VAR6]] = add i32 [[VAR5]], [[SCALAR_RECUR]]
; SINK-AFTER-NEXT:    [[VAR7]] = udiv i32 219220132, [[VAR3]]
; SINK-AFTER-NEXT:    store i32 [[VAR3]], i32* [[G]], align 4
; SINK-AFTER-NEXT:    [[VAR8]] = add nsw i32 [[VAR3]], -1
; SINK-AFTER-NEXT:    [[IV_NEXT]] = add nsw i32 [[IV]], 1
; SINK-AFTER-NEXT:    [[VAR9:%.*]] = icmp slt i32 [[VAR3]], 2
; SINK-AFTER-NEXT:    br i1 [[VAR9]], label [[BB1]], label [[BB2]], !prof [[PROF26]], !llvm.loop [[LOOP29:![0-9]+]]
;
bb:
  br label %bb2

  bb1:                                              ; preds = %bb2
  %var = phi i32 [ %var6, %bb2 ]
  ret i32 %var

  bb2:                                              ; preds = %bb2, %bb
  %var3 = phi i32 [ %var8, %bb2 ], [ %y, %bb ]
  %iv = phi i32 [ %iv.next, %bb2 ], [ 0, %bb ]
  %var4 = phi i32 [ %var7, %bb2 ], [ 0, %bb ]
  %var5 = phi i32 [ %var6, %bb2 ], [ 0, %bb ]
  %g = getelementptr inbounds i32, i32* %x, i32 %iv
  %var6 = add i32 %var5, %var4
  %var7 = udiv i32 219220132, %var3
  store i32 %var3, i32* %g, align 4
  %var8 = add nsw i32 %var3, -1
  %iv.next = add nsw i32 %iv, 1
  %var9 = icmp slt i32 %var3, 2
  br i1 %var9, label %bb1, label %bb2, !prof !2
}

; %vec.dead will be marked as dead instruction in the vector loop and no recipe
; will be created for it. Make sure a valid sink target is used.
define void @sink_after_dead_inst(i32* %A.ptr) {
; UNROLL-NO-IC-LABEL: @sink_after_dead_inst(
; UNROLL-NO-IC-NEXT:  entry:
; UNROLL-NO-IC-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-IC:       vector.ph:
; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-IC:       vector.body:
; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[VEC_IND:%.*]] = phi <4 x i16> [ <i16 0, i16 1, i16 2, i16 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-IC-NEXT:    [[STEP_ADD:%.*]] = add <4 x i16> [[VEC_IND]], <i16 4, i16 4, i16 4, i16 4>
; UNROLL-NO-IC-NEXT:    [[OFFSET_IDX:%.*]] = trunc i32 [[INDEX]] to i16
; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i16 [[OFFSET_IDX]], 0
; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = add i16 [[OFFSET_IDX]], 4
; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add <4 x i16> [[VEC_IND]], <i16 1, i16 1, i16 1, i16 1>
; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add <4 x i16> [[STEP_ADD]], <i16 1, i16 1, i16 1, i16 1>
; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = or <4 x i16> [[TMP2]], [[TMP2]]
; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = or <4 x i16> [[TMP3]], [[TMP3]]
; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = zext <4 x i16> [[TMP4]] to <4 x i32>
; UNROLL-NO-IC-NEXT:    [[TMP7]] = zext <4 x i16> [[TMP5]] to <4 x i32>
; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP6]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = shufflevector <4 x i32> [[TMP6]], <4 x i32> [[TMP7]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = getelementptr i32, i32* [[A_PTR:%.*]], i16 [[TMP0]]
; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = getelementptr i32, i32* [[A_PTR]], i16 [[TMP1]]
; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = getelementptr i32, i32* [[TMP10]], i32 0
; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = bitcast i32* [[TMP12]] to <4 x i32>*
; UNROLL-NO-IC-NEXT:    store <4 x i32> zeroinitializer, <4 x i32>* [[TMP13]], align 4
; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = getelementptr i32, i32* [[TMP10]], i32 4
; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = bitcast i32* [[TMP14]] to <4 x i32>*
; UNROLL-NO-IC-NEXT:    store <4 x i32> zeroinitializer, <4 x i32>* [[TMP15]], align 4
; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <4 x i16> [[STEP_ADD]], <i16 4, i16 4, i16 4, i16 4>
; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = icmp eq i32 [[INDEX_NEXT]], 16
; UNROLL-NO-IC-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
; UNROLL-NO-IC:       middle.block:
; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i32 16, 16
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP7]], i32 3
; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP7]], i32 2
; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-IC:       scalar.ph:
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 16, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; UNROLL-NO-IC-NEXT:    br label [[LOOP:%.*]]
; UNROLL-NO-IC:       loop:
; UNROLL-NO-IC-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[FOR_PREV:%.*]], [[LOOP]] ]
; UNROLL-NO-IC-NEXT:    [[CMP:%.*]] = icmp eq i32 [[SCALAR_RECUR]], 15
; UNROLL-NO-IC-NEXT:    [[C:%.*]] = icmp eq i1 [[CMP]], true
; UNROLL-NO-IC-NEXT:    [[VEC_DEAD:%.*]] = and i1 [[C]], true
; UNROLL-NO-IC-NEXT:    [[IV_NEXT]] = add i16 [[IV]], 1
; UNROLL-NO-IC-NEXT:    [[B1:%.*]] = or i16 [[IV_NEXT]], [[IV_NEXT]]
; UNROLL-NO-IC-NEXT:    [[B3:%.*]] = and i1 [[CMP]], [[C]]
; UNROLL-NO-IC-NEXT:    [[FOR_PREV]] = zext i16 [[B1]] to i32
; UNROLL-NO-IC-NEXT:    [[EXT:%.*]] = zext i1 [[B3]] to i32
; UNROLL-NO-IC-NEXT:    [[A_GEP:%.*]] = getelementptr i32, i32* [[A_PTR]], i16 [[IV]]
; UNROLL-NO-IC-NEXT:    store i32 0, i32* [[A_GEP]], align 4
; UNROLL-NO-IC-NEXT:    br i1 [[VEC_DEAD]], label [[FOR_END]], label [[LOOP]], !llvm.loop [[LOOP31:![0-9]+]]
; UNROLL-NO-IC:       for.end:
; UNROLL-NO-IC-NEXT:    ret void
;
; UNROLL-NO-VF-LABEL: @sink_after_dead_inst(
; UNROLL-NO-VF-NEXT:  entry:
; UNROLL-NO-VF-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; UNROLL-NO-VF:       vector.ph:
; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
; UNROLL-NO-VF:       vector.body:
; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
; UNROLL-NO-VF-NEXT:    [[OFFSET_IDX:%.*]] = trunc i32 [[INDEX]] to i16
; UNROLL-NO-VF-NEXT:    [[INDUCTION:%.*]] = add i16 [[OFFSET_IDX]], 0
; UNROLL-NO-VF-NEXT:    [[INDUCTION1:%.*]] = add i16 [[OFFSET_IDX]], 1
; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = add i16 [[INDUCTION]], 1
; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = add i16 [[INDUCTION1]], 1
; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = or i16 [[TMP0]], [[TMP0]]
; UNROLL-NO-VF-NEXT:    [[TMP3:%.*]] = or i16 [[TMP1]], [[TMP1]]
; UNROLL-NO-VF-NEXT:    [[TMP4:%.*]] = zext i16 [[TMP2]] to i32
; UNROLL-NO-VF-NEXT:    [[TMP5]] = zext i16 [[TMP3]] to i32
; UNROLL-NO-VF-NEXT:    [[TMP6:%.*]] = getelementptr i32, i32* [[A_PTR:%.*]], i16 [[INDUCTION]]
; UNROLL-NO-VF-NEXT:    [[TMP7:%.*]] = getelementptr i32, i32* [[A_PTR]], i16 [[INDUCTION1]]
; UNROLL-NO-VF-NEXT:    store i32 0, i32* [[TMP6]], align 4
; UNROLL-NO-VF-NEXT:    store i32 0, i32* [[TMP7]], align 4
; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
; UNROLL-NO-VF-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[INDEX_NEXT]], 16
; UNROLL-NO-VF-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP29:![0-9]+]]
; UNROLL-NO-VF:       middle.block:
; UNROLL-NO-VF-NEXT:    [[CMP_N:%.*]] = icmp eq i32 16, 16
; UNROLL-NO-VF-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; UNROLL-NO-VF:       scalar.ph:
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[TMP5]], [[MIDDLE_BLOCK]] ]
; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 16, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; UNROLL-NO-VF-NEXT:    br label [[LOOP:%.*]]
; UNROLL-NO-VF:       loop:
; UNROLL-NO-VF-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[FOR_PREV:%.*]], [[LOOP]] ]
; UNROLL-NO-VF-NEXT:    [[CMP:%.*]] = icmp eq i32 [[SCALAR_RECUR]], 15
; UNROLL-NO-VF-NEXT:    [[C:%.*]] = icmp eq i1 [[CMP]], true
; UNROLL-NO-VF-NEXT:    [[VEC_DEAD:%.*]] = and i1 [[C]], true
; UNROLL-NO-VF-NEXT:    [[IV_NEXT]] = add i16 [[IV]], 1
; UNROLL-NO-VF-NEXT:    [[B1:%.*]] = or i16 [[IV_NEXT]], [[IV_NEXT]]
; UNROLL-NO-VF-NEXT:    [[B3:%.*]] = and i1 [[CMP]], [[C]]
; UNROLL-NO-VF-NEXT:    [[FOR_PREV]] = zext i16 [[B1]] to i32
; UNROLL-NO-VF-NEXT:    [[EXT:%.*]] = zext i1 [[B3]] to i32
; UNROLL-NO-VF-NEXT:    [[A_GEP:%.*]] = getelementptr i32, i32* [[A_PTR]], i16 [[IV]]
; UNROLL-NO-VF-NEXT:    store i32 0, i32* [[A_GEP]], align 4
; UNROLL-NO-VF-NEXT:    br i1 [[VEC_DEAD]], label [[FOR_END]], label [[LOOP]], !llvm.loop [[LOOP30:![0-9]+]]
; UNROLL-NO-VF:       for.end:
; UNROLL-NO-VF-NEXT:    ret void
;
; SINK-AFTER-LABEL: @sink_after_dead_inst(
; SINK-AFTER-NEXT:  entry:
; SINK-AFTER-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; SINK-AFTER:       vector.ph:
; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
; SINK-AFTER:       vector.body:
; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[VEC_IND:%.*]] = phi <4 x i16> [ <i16 0, i16 1, i16 2, i16 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP3:%.*]], [[VECTOR_BODY]] ]
; SINK-AFTER-NEXT:    [[OFFSET_IDX:%.*]] = trunc i32 [[INDEX]] to i16
; SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i16 [[OFFSET_IDX]], 0
; SINK-AFTER-NEXT:    [[TMP1:%.*]] = add <4 x i16> [[VEC_IND]], <i16 1, i16 1, i16 1, i16 1>
; SINK-AFTER-NEXT:    [[TMP2:%.*]] = or <4 x i16> [[TMP1]], [[TMP1]]
; SINK-AFTER-NEXT:    [[TMP3]] = zext <4 x i16> [[TMP2]] to <4 x i32>
; SINK-AFTER-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP3]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
; SINK-AFTER-NEXT:    [[TMP5:%.*]] = getelementptr i32, i32* [[A_PTR:%.*]], i16 [[TMP0]]
; SINK-AFTER-NEXT:    [[TMP6:%.*]] = getelementptr i32, i32* [[TMP5]], i32 0
; SINK-AFTER-NEXT:    [[TMP7:%.*]] = bitcast i32* [[TMP6]] to <4 x i32>*
; SINK-AFTER-NEXT:    store <4 x i32> zeroinitializer, <4 x i32>* [[TMP7]], align 4
; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; SINK-AFTER-NEXT:    [[VEC_IND_NEXT]] = add <4 x i16> [[VEC_IND]], <i16 4, i16 4, i16 4, i16 4>
; SINK-AFTER-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[INDEX_NEXT]], 16
; SINK-AFTER-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
; SINK-AFTER:       middle.block:
; SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i32 16, 16
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP3]], i32 3
; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP3]], i32 2
; SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
; SINK-AFTER:       scalar.ph:
; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ]
; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 16, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
; SINK-AFTER-NEXT:    br label [[LOOP:%.*]]
; SINK-AFTER:       loop:
; SINK-AFTER-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
; SINK-AFTER-NEXT:    [[SCALAR_RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[FOR_PREV:%.*]], [[LOOP]] ]
; SINK-AFTER-NEXT:    [[CMP:%.*]] = icmp eq i32 [[SCALAR_RECUR]], 15
; SINK-AFTER-NEXT:    [[C:%.*]] = icmp eq i1 [[CMP]], true
; SINK-AFTER-NEXT:    [[VEC_DEAD:%.*]] = and i1 [[C]], true
; SINK-AFTER-NEXT:    [[IV_NEXT]] = add i16 [[IV]], 1
; SINK-AFTER-NEXT:    [[B1:%.*]] = or i16 [[IV_NEXT]], [[IV_NEXT]]
; SINK-AFTER-NEXT:    [[B3:%.*]] = and i1 [[CMP]], [[C]]
; SINK-AFTER-NEXT:    [[FOR_PREV]] = zext i16 [[B1]] to i32
; SINK-AFTER-NEXT:    [[EXT:%.*]] = zext i1 [[B3]] to i32
; SINK-AFTER-NEXT:    [[A_GEP:%.*]] = getelementptr i32, i32* [[A_PTR]], i16 [[IV]]
; SINK-AFTER-NEXT:    store i32 0, i32* [[A_GEP]], align 4
; SINK-AFTER-NEXT:    br i1 [[VEC_DEAD]], label [[FOR_END]], label [[LOOP]], !llvm.loop [[LOOP31:![0-9]+]]
; SINK-AFTER:       for.end:
; SINK-AFTER-NEXT:    ret void
;
entry:
  br label %loop

loop:
  %iv = phi i16 [ 0, %entry ], [ %iv.next, %loop ]
  %for = phi i32 [ 0, %entry ], [ %for.prev, %loop ]
  %cmp = icmp eq i32 %for, 15
  %C = icmp eq i1 %cmp, true
  %vec.dead = and i1 %C, 1
  %iv.next = add i16 %iv, 1
  %B1 = or i16 %iv.next, %iv.next
  %B3 = and i1 %cmp, %C
  %for.prev = zext i16 %B1 to i32

  %ext = zext i1 %B3 to i32
  %A.gep = getelementptr i32, i32* %A.ptr, i16 %iv
  store i32 0, i32* %A.gep
  br i1 %vec.dead, label %for.end, label %loop

for.end:
  ret void
}

!2 = !{!"branch_weights", i32 1, i32 1}