; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --filter-out-after "(for.body:|scalar.body:)" --filter-out-after "for.body:" --version 5
; RUN: opt -passes=loop-vectorize -scalable-vectorization=on -force-vector-width=4 -force-vector-interleave=1 -force-target-supports-scalable-vectors=true -S < %s | FileCheck %s --check-prefix=CHECK-VF4UF1
; RUN: opt -passes=loop-vectorize -scalable-vectorization=on -force-vector-width=4 -force-vector-interleave=2 -force-target-supports-scalable-vectors=true -S < %s | FileCheck %s --check-prefix=CHECK-VF4UF2

; void recurrence_1(int *a, int *b, int n) {
;   for(int i = 0; i < n; i++)
;     b[i] =  a[i] + a[i - 1]
; }
;
define i32 @recurrence_1(ptr nocapture readonly %a, ptr nocapture %b, i32 %n) {
; CHECK-VF4UF1-LABEL: define i32 @recurrence_1(
; CHECK-VF4UF1-SAME: ptr readonly captures(none) [[A:%.*]], ptr captures(none) [[B:%.*]], i32 [[N:%.*]]) {
; CHECK-VF4UF1-NEXT:  [[ENTRY:.*:]]
; CHECK-VF4UF1-NEXT:    [[A2:%.*]] = ptrtoint ptr [[A]] to i64
; CHECK-VF4UF1-NEXT:    [[B1:%.*]] = ptrtoint ptr [[B]] to i64
; CHECK-VF4UF1-NEXT:    br label %[[FOR_PREHEADER:.*]]
; CHECK-VF4UF1:       [[FOR_PREHEADER]]:
; CHECK-VF4UF1-NEXT:    [[PRE_LOAD:%.*]] = load i32, ptr [[A]], align 4
; CHECK-VF4UF1-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
; CHECK-VF4UF1-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
; CHECK-VF4UF1-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
; CHECK-VF4UF1-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vscale.i64()
; CHECK-VF4UF1-NEXT:    [[TMP4:%.*]] = shl nuw i64 [[TMP3]], 2
; CHECK-VF4UF1-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], [[TMP4]]
; CHECK-VF4UF1-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
; CHECK-VF4UF1:       [[VECTOR_MEMCHECK]]:
; CHECK-VF4UF1-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
; CHECK-VF4UF1-NEXT:    [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 4
; CHECK-VF4UF1-NEXT:    [[TMP7:%.*]] = mul i64 [[TMP6]], 4
; CHECK-VF4UF1-NEXT:    [[TMP8:%.*]] = add i64 [[B1]], -4
; CHECK-VF4UF1-NEXT:    [[TMP9:%.*]] = sub i64 [[TMP8]], [[A2]]
; CHECK-VF4UF1-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP9]], [[TMP7]]
; CHECK-VF4UF1-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
; CHECK-VF4UF1:       [[VECTOR_PH]]:
; CHECK-VF4UF1-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
; CHECK-VF4UF1-NEXT:    [[TMP11:%.*]] = mul nuw i64 [[TMP10]], 4
; CHECK-VF4UF1-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], [[TMP11]]
; CHECK-VF4UF1-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
; CHECK-VF4UF1-NEXT:    [[TMP14:%.*]] = call i32 @llvm.vscale.i32()
; CHECK-VF4UF1-NEXT:    [[TMP15:%.*]] = mul nuw i32 [[TMP14]], 4
; CHECK-VF4UF1-NEXT:    [[TMP16:%.*]] = sub i32 [[TMP15]], 1
; CHECK-VF4UF1-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[PRE_LOAD]], i32 [[TMP16]]
; CHECK-VF4UF1-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK-VF4UF1:       [[VECTOR_BODY]]:
; CHECK-VF4UF1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-VF4UF1-NEXT:    [[VECTOR_RECUR:%.*]] = phi <vscale x 4 x i32> [ [[VECTOR_RECUR_INIT]], %[[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], %[[VECTOR_BODY]] ]
; CHECK-VF4UF1-NEXT:    [[TMP17:%.*]] = add nuw nsw i64 [[INDEX]], 1
; CHECK-VF4UF1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP17]]
; CHECK-VF4UF1-NEXT:    [[WIDE_LOAD]] = load <vscale x 4 x i32>, ptr [[TMP18]], align 4
; CHECK-VF4UF1-NEXT:    [[TMP20:%.*]] = call <vscale x 4 x i32> @llvm.vector.splice.nxv4i32(<vscale x 4 x i32> [[VECTOR_RECUR]], <vscale x 4 x i32> [[WIDE_LOAD]], i32 -1)
; CHECK-VF4UF1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
; CHECK-VF4UF1-NEXT:    [[TMP22:%.*]] = add <vscale x 4 x i32> [[WIDE_LOAD]], [[TMP20]]
; CHECK-VF4UF1-NEXT:    store <vscale x 4 x i32> [[TMP22]], ptr [[TMP21]], align 4
; CHECK-VF4UF1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP11]]
; CHECK-VF4UF1-NEXT:    [[TMP24:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-VF4UF1-NEXT:    br i1 [[TMP24]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; CHECK-VF4UF1:       [[MIDDLE_BLOCK]]:
; CHECK-VF4UF1-NEXT:    [[TMP25:%.*]] = call i32 @llvm.vscale.i32()
; CHECK-VF4UF1-NEXT:    [[TMP26:%.*]] = mul nuw i32 [[TMP25]], 4
; CHECK-VF4UF1-NEXT:    [[TMP27:%.*]] = sub i32 [[TMP26]], 1
; CHECK-VF4UF1-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <vscale x 4 x i32> [[WIDE_LOAD]], i32 [[TMP27]]
; CHECK-VF4UF1-NEXT:    [[TMP28:%.*]] = call i32 @llvm.vscale.i32()
; CHECK-VF4UF1-NEXT:    [[TMP29:%.*]] = mul nuw i32 [[TMP28]], 4
; CHECK-VF4UF1-NEXT:    [[TMP30:%.*]] = sub i32 [[TMP29]], 2
; CHECK-VF4UF1-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <vscale x 4 x i32> [[WIDE_LOAD]], i32 [[TMP30]]
; CHECK-VF4UF1-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; CHECK-VF4UF1-NEXT:    br i1 [[CMP_N]], [[FOR_EXIT:label %.*]], label %[[SCALAR_PH]]
; CHECK-VF4UF1:       [[SCALAR_PH]]:
; CHECK-VF4UF1-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], %[[MIDDLE_BLOCK]] ], [ [[PRE_LOAD]], %[[FOR_PREHEADER]] ], [ [[PRE_LOAD]], %[[VECTOR_MEMCHECK]] ]
; CHECK-VF4UF1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[FOR_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
; CHECK-VF4UF1-NEXT:    br label %[[SCALAR_BODY:.*]]
; CHECK-VF4UF1:       [[SCALAR_BODY]]:
;
; CHECK-VF4UF2-LABEL: define i32 @recurrence_1(
; CHECK-VF4UF2-SAME: ptr readonly captures(none) [[A:%.*]], ptr captures(none) [[B:%.*]], i32 [[N:%.*]]) {
; CHECK-VF4UF2-NEXT:  [[ENTRY:.*:]]
; CHECK-VF4UF2-NEXT:    [[A2:%.*]] = ptrtoint ptr [[A]] to i64
; CHECK-VF4UF2-NEXT:    [[B1:%.*]] = ptrtoint ptr [[B]] to i64
; CHECK-VF4UF2-NEXT:    br label %[[FOR_PREHEADER:.*]]
; CHECK-VF4UF2:       [[FOR_PREHEADER]]:
; CHECK-VF4UF2-NEXT:    [[PRE_LOAD:%.*]] = load i32, ptr [[A]], align 4
; CHECK-VF4UF2-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
; CHECK-VF4UF2-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
; CHECK-VF4UF2-NEXT:    [[TMP2:%.*]] = add nuw nsw i64 [[TMP1]], 1
; CHECK-VF4UF2-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vscale.i64()
; CHECK-VF4UF2-NEXT:    [[TMP4:%.*]] = shl nuw i64 [[TMP3]], 3
; CHECK-VF4UF2-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], [[TMP4]]
; CHECK-VF4UF2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
; CHECK-VF4UF2:       [[VECTOR_MEMCHECK]]:
; CHECK-VF4UF2-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
; CHECK-VF4UF2-NEXT:    [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 4
; CHECK-VF4UF2-NEXT:    [[TMP7:%.*]] = mul i64 [[TMP6]], 8
; CHECK-VF4UF2-NEXT:    [[TMP8:%.*]] = add i64 [[B1]], -4
; CHECK-VF4UF2-NEXT:    [[TMP9:%.*]] = sub i64 [[TMP8]], [[A2]]
; CHECK-VF4UF2-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP9]], [[TMP7]]
; CHECK-VF4UF2-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
; CHECK-VF4UF2:       [[VECTOR_PH]]:
; CHECK-VF4UF2-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
; CHECK-VF4UF2-NEXT:    [[TMP12:%.*]] = mul nuw i64 [[TMP10]], 4
; CHECK-VF4UF2-NEXT:    [[TMP11:%.*]] = mul nuw i64 [[TMP12]], 2
; CHECK-VF4UF2-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], [[TMP11]]
; CHECK-VF4UF2-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
; CHECK-VF4UF2-NEXT:    [[TMP14:%.*]] = call i32 @llvm.vscale.i32()
; CHECK-VF4UF2-NEXT:    [[TMP15:%.*]] = mul nuw i32 [[TMP14]], 4
; CHECK-VF4UF2-NEXT:    [[TMP16:%.*]] = sub i32 [[TMP15]], 1
; CHECK-VF4UF2-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[PRE_LOAD]], i32 [[TMP16]]
; CHECK-VF4UF2-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK-VF4UF2:       [[VECTOR_BODY]]:
; CHECK-VF4UF2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-VF4UF2-NEXT:    [[VECTOR_RECUR:%.*]] = phi <vscale x 4 x i32> [ [[VECTOR_RECUR_INIT]], %[[VECTOR_PH]] ], [ [[WIDE_LOAD3:%.*]], %[[VECTOR_BODY]] ]
; CHECK-VF4UF2-NEXT:    [[TMP17:%.*]] = add nuw nsw i64 [[INDEX]], 1
; CHECK-VF4UF2-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP17]]
; CHECK-VF4UF2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i32, ptr [[TMP18]], i64 [[TMP12]]
; CHECK-VF4UF2-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP18]], align 4
; CHECK-VF4UF2-NEXT:    [[WIDE_LOAD3]] = load <vscale x 4 x i32>, ptr [[TMP22]], align 4
; CHECK-VF4UF2-NEXT:    [[TMP23:%.*]] = call <vscale x 4 x i32> @llvm.vector.splice.nxv4i32(<vscale x 4 x i32> [[VECTOR_RECUR]], <vscale x 4 x i32> [[WIDE_LOAD]], i32 -1)
; CHECK-VF4UF2-NEXT:    [[TMP24:%.*]] = call <vscale x 4 x i32> @llvm.vector.splice.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]], <vscale x 4 x i32> [[WIDE_LOAD3]], i32 -1)
; CHECK-VF4UF2-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
; CHECK-VF4UF2-NEXT:    [[TMP26:%.*]] = add <vscale x 4 x i32> [[WIDE_LOAD]], [[TMP23]]
; CHECK-VF4UF2-NEXT:    [[TMP27:%.*]] = add <vscale x 4 x i32> [[WIDE_LOAD3]], [[TMP24]]
; CHECK-VF4UF2-NEXT:    [[TMP31:%.*]] = getelementptr inbounds i32, ptr [[TMP25]], i64 [[TMP12]]
; CHECK-VF4UF2-NEXT:    store <vscale x 4 x i32> [[TMP26]], ptr [[TMP25]], align 4
; CHECK-VF4UF2-NEXT:    store <vscale x 4 x i32> [[TMP27]], ptr [[TMP31]], align 4
; CHECK-VF4UF2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP11]]
; CHECK-VF4UF2-NEXT:    [[TMP32:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-VF4UF2-NEXT:    br i1 [[TMP32]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; CHECK-VF4UF2:       [[MIDDLE_BLOCK]]:
; CHECK-VF4UF2-NEXT:    [[TMP33:%.*]] = call i32 @llvm.vscale.i32()
; CHECK-VF4UF2-NEXT:    [[TMP34:%.*]] = mul nuw i32 [[TMP33]], 4
; CHECK-VF4UF2-NEXT:    [[TMP35:%.*]] = sub i32 [[TMP34]], 1
; CHECK-VF4UF2-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <vscale x 4 x i32> [[WIDE_LOAD3]], i32 [[TMP35]]
; CHECK-VF4UF2-NEXT:    [[TMP36:%.*]] = call i32 @llvm.vscale.i32()
; CHECK-VF4UF2-NEXT:    [[TMP37:%.*]] = mul nuw i32 [[TMP36]], 4
; CHECK-VF4UF2-NEXT:    [[TMP38:%.*]] = sub i32 [[TMP37]], 2
; CHECK-VF4UF2-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <vscale x 4 x i32> [[WIDE_LOAD3]], i32 [[TMP38]]
; CHECK-VF4UF2-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; CHECK-VF4UF2-NEXT:    br i1 [[CMP_N]], [[FOR_EXIT:label %.*]], label %[[SCALAR_PH]]
; CHECK-VF4UF2:       [[SCALAR_PH]]:
; CHECK-VF4UF2-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], %[[MIDDLE_BLOCK]] ], [ [[PRE_LOAD]], %[[FOR_PREHEADER]] ], [ [[PRE_LOAD]], %[[VECTOR_MEMCHECK]] ]
; CHECK-VF4UF2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[FOR_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
; CHECK-VF4UF2-NEXT:    br label %[[SCALAR_BODY:.*]]
; CHECK-VF4UF2:       [[SCALAR_BODY]]:
;
entry:
  br label %for.preheader

for.preheader:
  %pre_load = load i32, ptr %a
  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, ptr %a, i64 %indvars.iv.next
  %1 = load i32, ptr %arrayidx32
  %arrayidx34 = getelementptr inbounds i32, ptr %b, i64 %indvars.iv
  %add35 = add i32 %1, %0
  store i32 %add35, ptr %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, !llvm.loop !0

for.exit:
  ret i32 %0
}

; 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(ptr nocapture readonly %a, i32 %n) {
; CHECK-VF4UF1-LABEL: define i32 @recurrence_2(
; CHECK-VF4UF1-SAME: ptr readonly captures(none) [[A:%.*]], i32 [[N:%.*]]) {
; CHECK-VF4UF1-NEXT:  [[ENTRY:.*:]]
; CHECK-VF4UF1-NEXT:    br label %[[FOR_PREHEADER:.*]]
; CHECK-VF4UF1:       [[FOR_PREHEADER]]:
; CHECK-VF4UF1-NEXT:    [[ARRAYIDX2_PHI_TRANS_INSERT:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 -1
; CHECK-VF4UF1-NEXT:    [[DOTPRE:%.*]] = load i32, ptr [[ARRAYIDX2_PHI_TRANS_INSERT]], align 4
; CHECK-VF4UF1-NEXT:    [[TMP5:%.*]] = add i32 [[N]], -1
; CHECK-VF4UF1-NEXT:    [[TMP6:%.*]] = zext i32 [[TMP5]] to i64
; CHECK-VF4UF1-NEXT:    [[TMP0:%.*]] = add nuw nsw i64 [[TMP6]], 1
; CHECK-VF4UF1-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
; CHECK-VF4UF1-NEXT:    [[TMP2:%.*]] = shl nuw i64 [[TMP1]], 2
; CHECK-VF4UF1-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP0]], [[TMP2]]
; CHECK-VF4UF1-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK-VF4UF1:       [[VECTOR_PH]]:
; CHECK-VF4UF1-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vscale.i64()
; CHECK-VF4UF1-NEXT:    [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4
; CHECK-VF4UF1-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP0]], [[TMP4]]
; CHECK-VF4UF1-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP0]], [[N_MOD_VF]]
; CHECK-VF4UF1-NEXT:    [[TMP7:%.*]] = call i32 @llvm.vscale.i32()
; CHECK-VF4UF1-NEXT:    [[TMP8:%.*]] = mul nuw i32 [[TMP7]], 4
; CHECK-VF4UF1-NEXT:    [[TMP9:%.*]] = sub i32 [[TMP8]], 1
; CHECK-VF4UF1-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[DOTPRE]], i32 [[TMP9]]
; CHECK-VF4UF1-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK-VF4UF1:       [[VECTOR_BODY]]:
; CHECK-VF4UF1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-VF4UF1-NEXT:    [[VECTOR_RECUR:%.*]] = phi <vscale x 4 x i32> [ [[VECTOR_RECUR_INIT]], %[[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], %[[VECTOR_BODY]] ]
; CHECK-VF4UF1-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP17:%.*]], %[[VECTOR_BODY]] ]
; CHECK-VF4UF1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
; CHECK-VF4UF1-NEXT:    [[WIDE_LOAD]] = load <vscale x 4 x i32>, ptr [[TMP10]], align 4
; CHECK-VF4UF1-NEXT:    [[TMP12:%.*]] = call <vscale x 4 x i32> @llvm.vector.splice.nxv4i32(<vscale x 4 x i32> [[VECTOR_RECUR]], <vscale x 4 x i32> [[WIDE_LOAD]], i32 -1)
; CHECK-VF4UF1-NEXT:    [[TMP13:%.*]] = sub nsw <vscale x 4 x i32> [[WIDE_LOAD]], [[TMP12]]
; CHECK-VF4UF1-NEXT:    [[TMP14:%.*]] = icmp sgt <vscale x 4 x i32> [[TMP13]], zeroinitializer
; CHECK-VF4UF1-NEXT:    [[TMP15:%.*]] = select <vscale x 4 x i1> [[TMP14]], <vscale x 4 x i32> [[TMP13]], <vscale x 4 x i32> zeroinitializer
; CHECK-VF4UF1-NEXT:    [[TMP16:%.*]] = icmp slt <vscale x 4 x i32> [[VEC_PHI]], [[TMP15]]
; CHECK-VF4UF1-NEXT:    [[TMP17]] = select <vscale x 4 x i1> [[TMP16]], <vscale x 4 x i32> [[VEC_PHI]], <vscale x 4 x i32> [[TMP15]]
; CHECK-VF4UF1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP4]]
; CHECK-VF4UF1-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-VF4UF1-NEXT:    br i1 [[TMP18]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
; CHECK-VF4UF1:       [[MIDDLE_BLOCK]]:
; CHECK-VF4UF1-NEXT:    [[TMP19:%.*]] = call i32 @llvm.vector.reduce.smin.nxv4i32(<vscale x 4 x i32> [[TMP17]])
; CHECK-VF4UF1-NEXT:    [[TMP20:%.*]] = call i32 @llvm.vscale.i32()
; CHECK-VF4UF1-NEXT:    [[TMP21:%.*]] = mul nuw i32 [[TMP20]], 4
; CHECK-VF4UF1-NEXT:    [[TMP22:%.*]] = sub i32 [[TMP21]], 1
; CHECK-VF4UF1-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <vscale x 4 x i32> [[WIDE_LOAD]], i32 [[TMP22]]
; CHECK-VF4UF1-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC]]
; CHECK-VF4UF1-NEXT:    br i1 [[CMP_N]], label %[[FOR_COND_CLEANUP_LOOPEXIT:.*]], label %[[SCALAR_PH]]
; CHECK-VF4UF1:       [[SCALAR_PH]]:
; CHECK-VF4UF1-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ [[DOTPRE]], %[[FOR_PREHEADER]] ]
; CHECK-VF4UF1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[FOR_PREHEADER]] ]
; CHECK-VF4UF1-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP19]], %[[MIDDLE_BLOCK]] ], [ 0, %[[FOR_PREHEADER]] ]
; CHECK-VF4UF1-NEXT:    br label %[[SCALAR_BODY:.*]]
; CHECK-VF4UF1:       [[FOR_COND_CLEANUP_LOOPEXIT]]:
; CHECK-VF4UF1-NEXT:    [[MINMAX_0_LCSSA:%.*]] = phi i32 [ [[MINMAX_0_COND:%.*]], %[[SCALAR_BODY]] ], [ [[TMP19]], %[[MIDDLE_BLOCK]] ]
; CHECK-VF4UF1-NEXT:    br label %[[FOR_COND_CLEANUP:.*]]
; CHECK-VF4UF1:       [[FOR_COND_CLEANUP]]:
; CHECK-VF4UF1-NEXT:    ret i32 [[MINMAX_0_LCSSA]]
; CHECK-VF4UF1:       [[SCALAR_BODY]]:
;
; CHECK-VF4UF2-LABEL: define i32 @recurrence_2(
; CHECK-VF4UF2-SAME: ptr readonly captures(none) [[A:%.*]], i32 [[N:%.*]]) {
; CHECK-VF4UF2-NEXT:  [[ENTRY:.*:]]
; CHECK-VF4UF2-NEXT:    br label %[[FOR_PREHEADER:.*]]
; CHECK-VF4UF2:       [[FOR_PREHEADER]]:
; CHECK-VF4UF2-NEXT:    [[ARRAYIDX2_PHI_TRANS_INSERT:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 -1
; CHECK-VF4UF2-NEXT:    [[DOTPRE:%.*]] = load i32, ptr [[ARRAYIDX2_PHI_TRANS_INSERT]], align 4
; CHECK-VF4UF2-NEXT:    [[TMP5:%.*]] = add i32 [[N]], -1
; CHECK-VF4UF2-NEXT:    [[TMP6:%.*]] = zext i32 [[TMP5]] to i64
; CHECK-VF4UF2-NEXT:    [[TMP0:%.*]] = add nuw nsw i64 [[TMP6]], 1
; CHECK-VF4UF2-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
; CHECK-VF4UF2-NEXT:    [[TMP2:%.*]] = shl nuw i64 [[TMP1]], 3
; CHECK-VF4UF2-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP0]], [[TMP2]]
; CHECK-VF4UF2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK-VF4UF2:       [[VECTOR_PH]]:
; CHECK-VF4UF2-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vscale.i64()
; CHECK-VF4UF2-NEXT:    [[TMP11:%.*]] = mul nuw i64 [[TMP3]], 4
; CHECK-VF4UF2-NEXT:    [[TMP4:%.*]] = mul nuw i64 [[TMP11]], 2
; CHECK-VF4UF2-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP0]], [[TMP4]]
; CHECK-VF4UF2-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP0]], [[N_MOD_VF]]
; CHECK-VF4UF2-NEXT:    [[TMP7:%.*]] = call i32 @llvm.vscale.i32()
; CHECK-VF4UF2-NEXT:    [[TMP8:%.*]] = mul nuw i32 [[TMP7]], 4
; CHECK-VF4UF2-NEXT:    [[TMP9:%.*]] = sub i32 [[TMP8]], 1
; CHECK-VF4UF2-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[DOTPRE]], i32 [[TMP9]]
; CHECK-VF4UF2-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK-VF4UF2:       [[VECTOR_BODY]]:
; CHECK-VF4UF2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-VF4UF2-NEXT:    [[VECTOR_RECUR:%.*]] = phi <vscale x 4 x i32> [ [[VECTOR_RECUR_INIT]], %[[VECTOR_PH]] ], [ [[WIDE_LOAD2:%.*]], %[[VECTOR_BODY]] ]
; CHECK-VF4UF2-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP25:%.*]], %[[VECTOR_BODY]] ]
; CHECK-VF4UF2-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP26:%.*]], %[[VECTOR_BODY]] ]
; CHECK-VF4UF2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
; CHECK-VF4UF2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[TMP10]], i64 [[TMP11]]
; CHECK-VF4UF2-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP10]], align 4
; CHECK-VF4UF2-NEXT:    [[WIDE_LOAD2]] = load <vscale x 4 x i32>, ptr [[TMP14]], align 4
; CHECK-VF4UF2-NEXT:    [[TMP15:%.*]] = call <vscale x 4 x i32> @llvm.vector.splice.nxv4i32(<vscale x 4 x i32> [[VECTOR_RECUR]], <vscale x 4 x i32> [[WIDE_LOAD]], i32 -1)
; CHECK-VF4UF2-NEXT:    [[TMP16:%.*]] = call <vscale x 4 x i32> @llvm.vector.splice.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]], <vscale x 4 x i32> [[WIDE_LOAD2]], i32 -1)
; CHECK-VF4UF2-NEXT:    [[TMP17:%.*]] = sub nsw <vscale x 4 x i32> [[WIDE_LOAD]], [[TMP15]]
; CHECK-VF4UF2-NEXT:    [[TMP18:%.*]] = sub nsw <vscale x 4 x i32> [[WIDE_LOAD2]], [[TMP16]]
; CHECK-VF4UF2-NEXT:    [[TMP19:%.*]] = icmp sgt <vscale x 4 x i32> [[TMP17]], zeroinitializer
; CHECK-VF4UF2-NEXT:    [[TMP20:%.*]] = icmp sgt <vscale x 4 x i32> [[TMP18]], zeroinitializer
; CHECK-VF4UF2-NEXT:    [[TMP21:%.*]] = select <vscale x 4 x i1> [[TMP19]], <vscale x 4 x i32> [[TMP17]], <vscale x 4 x i32> zeroinitializer
; CHECK-VF4UF2-NEXT:    [[TMP22:%.*]] = select <vscale x 4 x i1> [[TMP20]], <vscale x 4 x i32> [[TMP18]], <vscale x 4 x i32> zeroinitializer
; CHECK-VF4UF2-NEXT:    [[TMP23:%.*]] = icmp slt <vscale x 4 x i32> [[VEC_PHI]], [[TMP21]]
; CHECK-VF4UF2-NEXT:    [[TMP24:%.*]] = icmp slt <vscale x 4 x i32> [[VEC_PHI1]], [[TMP22]]
; CHECK-VF4UF2-NEXT:    [[TMP25]] = select <vscale x 4 x i1> [[TMP23]], <vscale x 4 x i32> [[VEC_PHI]], <vscale x 4 x i32> [[TMP21]]
; CHECK-VF4UF2-NEXT:    [[TMP26]] = select <vscale x 4 x i1> [[TMP24]], <vscale x 4 x i32> [[VEC_PHI1]], <vscale x 4 x i32> [[TMP22]]
; CHECK-VF4UF2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP4]]
; CHECK-VF4UF2-NEXT:    [[TMP27:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-VF4UF2-NEXT:    br i1 [[TMP27]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
; CHECK-VF4UF2:       [[MIDDLE_BLOCK]]:
; CHECK-VF4UF2-NEXT:    [[RDX_MINMAX:%.*]] = call <vscale x 4 x i32> @llvm.smin.nxv4i32(<vscale x 4 x i32> [[TMP25]], <vscale x 4 x i32> [[TMP26]])
; CHECK-VF4UF2-NEXT:    [[TMP28:%.*]] = call i32 @llvm.vector.reduce.smin.nxv4i32(<vscale x 4 x i32> [[RDX_MINMAX]])
; CHECK-VF4UF2-NEXT:    [[TMP29:%.*]] = call i32 @llvm.vscale.i32()
; CHECK-VF4UF2-NEXT:    [[TMP30:%.*]] = mul nuw i32 [[TMP29]], 4
; CHECK-VF4UF2-NEXT:    [[TMP31:%.*]] = sub i32 [[TMP30]], 1
; CHECK-VF4UF2-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <vscale x 4 x i32> [[WIDE_LOAD2]], i32 [[TMP31]]
; CHECK-VF4UF2-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC]]
; CHECK-VF4UF2-NEXT:    br i1 [[CMP_N]], label %[[FOR_COND_CLEANUP_LOOPEXIT:.*]], label %[[SCALAR_PH]]
; CHECK-VF4UF2:       [[SCALAR_PH]]:
; CHECK-VF4UF2-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ [[DOTPRE]], %[[FOR_PREHEADER]] ]
; CHECK-VF4UF2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[FOR_PREHEADER]] ]
; CHECK-VF4UF2-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP28]], %[[MIDDLE_BLOCK]] ], [ 0, %[[FOR_PREHEADER]] ]
; CHECK-VF4UF2-NEXT:    br label %[[SCALAR_BODY:.*]]
; CHECK-VF4UF2:       [[FOR_COND_CLEANUP_LOOPEXIT]]:
; CHECK-VF4UF2-NEXT:    [[MINMAX_0_LCSSA:%.*]] = phi i32 [ [[MINMAX_0_COND:%.*]], %[[SCALAR_BODY]] ], [ [[TMP28]], %[[MIDDLE_BLOCK]] ]
; CHECK-VF4UF2-NEXT:    br label %[[FOR_COND_CLEANUP:.*]]
; CHECK-VF4UF2:       [[FOR_COND_CLEANUP]]:
; CHECK-VF4UF2-NEXT:    ret i32 [[MINMAX_0_LCSSA]]
; CHECK-VF4UF2:       [[SCALAR_BODY]]:
;
entry:
  br label %for.preheader

for.preheader:
  %arrayidx2.phi.trans.insert = getelementptr inbounds i32, ptr %a, i64 -1
  %.pre = load i32, ptr %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:
  ret i32 %minmax.0.cond.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 [ 0, %for.preheader ], [ %minmax.0.cond, %scalar.body ]
  %arrayidx = getelementptr inbounds i32, ptr %a, i64 %indvars.iv
  %1 = load i32, ptr %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, !llvm.loop !0
}

define void @recurrence_3(ptr nocapture readonly %a, ptr nocapture %b, i32 %n, float %f, i16 %p) {
; CHECK-VF4UF1-LABEL: define void @recurrence_3(
; CHECK-VF4UF1-SAME: ptr readonly captures(none) [[A:%.*]], ptr captures(none) [[B:%.*]], i32 [[N:%.*]], float [[F:%.*]], i16 [[P:%.*]]) {
; CHECK-VF4UF1-NEXT:  [[ENTRY:.*:]]
; CHECK-VF4UF1-NEXT:    [[TMP0:%.*]] = load i16, ptr [[A]], align 2
; CHECK-VF4UF1-NEXT:    [[CONV:%.*]] = sitofp i16 [[TMP0]] to double
; CHECK-VF4UF1-NEXT:    [[CONV1:%.*]] = fpext float [[F]] to double
; CHECK-VF4UF1-NEXT:    [[CONV2:%.*]] = sitofp i16 [[P]] to double
; CHECK-VF4UF1-NEXT:    [[MUL:%.*]] = fmul fast double [[CONV2]], [[CONV1]]
; CHECK-VF4UF1-NEXT:    [[SUB:%.*]] = fsub fast double [[CONV]], [[MUL]]
; CHECK-VF4UF1-NEXT:    store double [[SUB]], ptr [[B]], align 8
; CHECK-VF4UF1-NEXT:    [[CMP25:%.*]] = icmp sgt i32 [[N]], 1
; CHECK-VF4UF1-NEXT:    br i1 [[CMP25]], label %[[FOR_PREHEADER:.*]], [[FOR_END:label %.*]]
; CHECK-VF4UF1:       [[FOR_PREHEADER]]:
; CHECK-VF4UF1-NEXT:    [[TMP1:%.*]] = add i32 [[N]], -1
; CHECK-VF4UF1-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
; CHECK-VF4UF1-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vscale.i64()
; CHECK-VF4UF1-NEXT:    [[TMP4:%.*]] = shl nuw i64 [[TMP3]], 2
; CHECK-VF4UF1-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], [[TMP4]]
; CHECK-VF4UF1-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
; CHECK-VF4UF1:       [[VECTOR_MEMCHECK]]:
; CHECK-VF4UF1-NEXT:    [[SCEVGEP:%.*]] = getelementptr i8, ptr [[B]], i64 8
; CHECK-VF4UF1-NEXT:    [[TMP5:%.*]] = add i32 [[N]], -2
; CHECK-VF4UF1-NEXT:    [[TMP6:%.*]] = zext i32 [[TMP5]] to i64
; CHECK-VF4UF1-NEXT:    [[TMP7:%.*]] = shl nuw nsw i64 [[TMP6]], 3
; CHECK-VF4UF1-NEXT:    [[TMP8:%.*]] = add nuw nsw i64 [[TMP7]], 16
; CHECK-VF4UF1-NEXT:    [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP8]]
; CHECK-VF4UF1-NEXT:    [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[A]], i64 2
; CHECK-VF4UF1-NEXT:    [[TMP9:%.*]] = shl nuw nsw i64 [[TMP6]], 1
; CHECK-VF4UF1-NEXT:    [[TMP10:%.*]] = add nuw nsw i64 [[TMP9]], 4
; CHECK-VF4UF1-NEXT:    [[SCEVGEP3:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP10]]
; CHECK-VF4UF1-NEXT:    [[BOUND0:%.*]] = icmp ult ptr [[SCEVGEP]], [[SCEVGEP3]]
; CHECK-VF4UF1-NEXT:    [[BOUND1:%.*]] = icmp ult ptr [[SCEVGEP2]], [[SCEVGEP1]]
; CHECK-VF4UF1-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
; CHECK-VF4UF1-NEXT:    br i1 [[FOUND_CONFLICT]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
; CHECK-VF4UF1:       [[VECTOR_PH]]:
; CHECK-VF4UF1-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
; CHECK-VF4UF1-NEXT:    [[TMP12:%.*]] = mul nuw i64 [[TMP11]], 4
; CHECK-VF4UF1-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], [[TMP12]]
; CHECK-VF4UF1-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
; CHECK-VF4UF1-NEXT:    [[TMP15:%.*]] = add i64 1, [[N_VEC]]
; CHECK-VF4UF1-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x double> poison, double [[CONV1]], i64 0
; CHECK-VF4UF1-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x double> [[BROADCAST_SPLATINSERT]], <vscale x 4 x double> poison, <vscale x 4 x i32> zeroinitializer
; CHECK-VF4UF1-NEXT:    [[TMP16:%.*]] = call i32 @llvm.vscale.i32()
; CHECK-VF4UF1-NEXT:    [[TMP17:%.*]] = mul nuw i32 [[TMP16]], 4
; CHECK-VF4UF1-NEXT:    [[TMP18:%.*]] = sub i32 [[TMP17]], 1
; CHECK-VF4UF1-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <vscale x 4 x i16> poison, i16 [[TMP0]], i32 [[TMP18]]
; CHECK-VF4UF1-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK-VF4UF1:       [[VECTOR_BODY]]:
; CHECK-VF4UF1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-VF4UF1-NEXT:    [[VECTOR_RECUR:%.*]] = phi <vscale x 4 x i16> [ [[VECTOR_RECUR_INIT]], %[[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], %[[VECTOR_BODY]] ]
; CHECK-VF4UF1-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
; CHECK-VF4UF1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i16, ptr [[A]], i64 [[OFFSET_IDX]]
; CHECK-VF4UF1-NEXT:    [[WIDE_LOAD]] = load <vscale x 4 x i16>, ptr [[TMP19]], align 2, !alias.scope [[META6:![0-9]+]]
; CHECK-VF4UF1-NEXT:    [[TMP21:%.*]] = call <vscale x 4 x i16> @llvm.vector.splice.nxv4i16(<vscale x 4 x i16> [[VECTOR_RECUR]], <vscale x 4 x i16> [[WIDE_LOAD]], i32 -1)
; CHECK-VF4UF1-NEXT:    [[TMP22:%.*]] = sitofp <vscale x 4 x i16> [[WIDE_LOAD]] to <vscale x 4 x double>
; CHECK-VF4UF1-NEXT:    [[TMP23:%.*]] = sitofp <vscale x 4 x i16> [[TMP21]] to <vscale x 4 x double>
; CHECK-VF4UF1-NEXT:    [[TMP24:%.*]] = fmul fast <vscale x 4 x double> [[TMP23]], [[BROADCAST_SPLAT]]
; CHECK-VF4UF1-NEXT:    [[TMP25:%.*]] = fsub fast <vscale x 4 x double> [[TMP22]], [[TMP24]]
; CHECK-VF4UF1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds double, ptr [[B]], i64 [[OFFSET_IDX]]
; CHECK-VF4UF1-NEXT:    store <vscale x 4 x double> [[TMP25]], ptr [[TMP26]], align 8, !alias.scope [[META9:![0-9]+]], !noalias [[META6]]
; CHECK-VF4UF1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP12]]
; CHECK-VF4UF1-NEXT:    [[TMP28:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-VF4UF1-NEXT:    br i1 [[TMP28]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
; CHECK-VF4UF1:       [[MIDDLE_BLOCK]]:
; CHECK-VF4UF1-NEXT:    [[TMP29:%.*]] = call i32 @llvm.vscale.i32()
; CHECK-VF4UF1-NEXT:    [[TMP30:%.*]] = mul nuw i32 [[TMP29]], 4
; CHECK-VF4UF1-NEXT:    [[TMP31:%.*]] = sub i32 [[TMP30]], 1
; CHECK-VF4UF1-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <vscale x 4 x i16> [[WIDE_LOAD]], i32 [[TMP31]]
; CHECK-VF4UF1-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; CHECK-VF4UF1-NEXT:    br i1 [[CMP_N]], [[FOR_END_LOOPEXIT:label %.*]], label %[[SCALAR_PH]]
; CHECK-VF4UF1:       [[SCALAR_PH]]:
; CHECK-VF4UF1-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ [[TMP0]], %[[FOR_PREHEADER]] ], [ [[TMP0]], %[[VECTOR_MEMCHECK]] ]
; CHECK-VF4UF1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[TMP15]], %[[MIDDLE_BLOCK]] ], [ 1, %[[FOR_PREHEADER]] ], [ 1, %[[VECTOR_MEMCHECK]] ]
; CHECK-VF4UF1-NEXT:    br label %[[SCALAR_BODY:.*]]
; CHECK-VF4UF1:       [[SCALAR_BODY]]:
;
; CHECK-VF4UF2-LABEL: define void @recurrence_3(
; CHECK-VF4UF2-SAME: ptr readonly captures(none) [[A:%.*]], ptr captures(none) [[B:%.*]], i32 [[N:%.*]], float [[F:%.*]], i16 [[P:%.*]]) {
; CHECK-VF4UF2-NEXT:  [[ENTRY:.*:]]
; CHECK-VF4UF2-NEXT:    [[TMP0:%.*]] = load i16, ptr [[A]], align 2
; CHECK-VF4UF2-NEXT:    [[CONV:%.*]] = sitofp i16 [[TMP0]] to double
; CHECK-VF4UF2-NEXT:    [[CONV1:%.*]] = fpext float [[F]] to double
; CHECK-VF4UF2-NEXT:    [[CONV2:%.*]] = sitofp i16 [[P]] to double
; CHECK-VF4UF2-NEXT:    [[MUL:%.*]] = fmul fast double [[CONV2]], [[CONV1]]
; CHECK-VF4UF2-NEXT:    [[SUB:%.*]] = fsub fast double [[CONV]], [[MUL]]
; CHECK-VF4UF2-NEXT:    store double [[SUB]], ptr [[B]], align 8
; CHECK-VF4UF2-NEXT:    [[CMP25:%.*]] = icmp sgt i32 [[N]], 1
; CHECK-VF4UF2-NEXT:    br i1 [[CMP25]], label %[[FOR_PREHEADER:.*]], [[FOR_END:label %.*]]
; CHECK-VF4UF2:       [[FOR_PREHEADER]]:
; CHECK-VF4UF2-NEXT:    [[TMP1:%.*]] = add i32 [[N]], -1
; CHECK-VF4UF2-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
; CHECK-VF4UF2-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vscale.i64()
; CHECK-VF4UF2-NEXT:    [[TMP4:%.*]] = shl nuw i64 [[TMP3]], 3
; CHECK-VF4UF2-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], [[TMP4]]
; CHECK-VF4UF2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
; CHECK-VF4UF2:       [[VECTOR_MEMCHECK]]:
; CHECK-VF4UF2-NEXT:    [[SCEVGEP:%.*]] = getelementptr i8, ptr [[B]], i64 8
; CHECK-VF4UF2-NEXT:    [[TMP5:%.*]] = add i32 [[N]], -2
; CHECK-VF4UF2-NEXT:    [[TMP6:%.*]] = zext i32 [[TMP5]] to i64
; CHECK-VF4UF2-NEXT:    [[TMP7:%.*]] = shl nuw nsw i64 [[TMP6]], 3
; CHECK-VF4UF2-NEXT:    [[TMP8:%.*]] = add nuw nsw i64 [[TMP7]], 16
; CHECK-VF4UF2-NEXT:    [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP8]]
; CHECK-VF4UF2-NEXT:    [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[A]], i64 2
; CHECK-VF4UF2-NEXT:    [[TMP9:%.*]] = shl nuw nsw i64 [[TMP6]], 1
; CHECK-VF4UF2-NEXT:    [[TMP10:%.*]] = add nuw nsw i64 [[TMP9]], 4
; CHECK-VF4UF2-NEXT:    [[SCEVGEP3:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP10]]
; CHECK-VF4UF2-NEXT:    [[BOUND0:%.*]] = icmp ult ptr [[SCEVGEP]], [[SCEVGEP3]]
; CHECK-VF4UF2-NEXT:    [[BOUND1:%.*]] = icmp ult ptr [[SCEVGEP2]], [[SCEVGEP1]]
; CHECK-VF4UF2-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
; CHECK-VF4UF2-NEXT:    br i1 [[FOUND_CONFLICT]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
; CHECK-VF4UF2:       [[VECTOR_PH]]:
; CHECK-VF4UF2-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
; CHECK-VF4UF2-NEXT:    [[TMP13:%.*]] = mul nuw i64 [[TMP11]], 4
; CHECK-VF4UF2-NEXT:    [[TMP12:%.*]] = mul nuw i64 [[TMP13]], 2
; CHECK-VF4UF2-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], [[TMP12]]
; CHECK-VF4UF2-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]]
; CHECK-VF4UF2-NEXT:    [[TMP15:%.*]] = add i64 1, [[N_VEC]]
; CHECK-VF4UF2-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x double> poison, double [[CONV1]], i64 0
; CHECK-VF4UF2-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x double> [[BROADCAST_SPLATINSERT]], <vscale x 4 x double> poison, <vscale x 4 x i32> zeroinitializer
; CHECK-VF4UF2-NEXT:    [[TMP16:%.*]] = call i32 @llvm.vscale.i32()
; CHECK-VF4UF2-NEXT:    [[TMP17:%.*]] = mul nuw i32 [[TMP16]], 4
; CHECK-VF4UF2-NEXT:    [[TMP18:%.*]] = sub i32 [[TMP17]], 1
; CHECK-VF4UF2-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <vscale x 4 x i16> poison, i16 [[TMP0]], i32 [[TMP18]]
; CHECK-VF4UF2-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK-VF4UF2:       [[VECTOR_BODY]]:
; CHECK-VF4UF2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-VF4UF2-NEXT:    [[VECTOR_RECUR:%.*]] = phi <vscale x 4 x i16> [ [[VECTOR_RECUR_INIT]], %[[VECTOR_PH]] ], [ [[WIDE_LOAD4:%.*]], %[[VECTOR_BODY]] ]
; CHECK-VF4UF2-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
; CHECK-VF4UF2-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i16, ptr [[A]], i64 [[OFFSET_IDX]]
; CHECK-VF4UF2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i16, ptr [[TMP19]], i64 [[TMP13]]
; CHECK-VF4UF2-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i16>, ptr [[TMP19]], align 2, !alias.scope [[META6:![0-9]+]]
; CHECK-VF4UF2-NEXT:    [[WIDE_LOAD4]] = load <vscale x 4 x i16>, ptr [[TMP20]], align 2, !alias.scope [[META6]]
; CHECK-VF4UF2-NEXT:    [[TMP24:%.*]] = call <vscale x 4 x i16> @llvm.vector.splice.nxv4i16(<vscale x 4 x i16> [[VECTOR_RECUR]], <vscale x 4 x i16> [[WIDE_LOAD]], i32 -1)
; CHECK-VF4UF2-NEXT:    [[TMP25:%.*]] = call <vscale x 4 x i16> @llvm.vector.splice.nxv4i16(<vscale x 4 x i16> [[WIDE_LOAD]], <vscale x 4 x i16> [[WIDE_LOAD4]], i32 -1)
; CHECK-VF4UF2-NEXT:    [[TMP26:%.*]] = sitofp <vscale x 4 x i16> [[WIDE_LOAD]] to <vscale x 4 x double>
; CHECK-VF4UF2-NEXT:    [[TMP27:%.*]] = sitofp <vscale x 4 x i16> [[WIDE_LOAD4]] to <vscale x 4 x double>
; CHECK-VF4UF2-NEXT:    [[TMP28:%.*]] = sitofp <vscale x 4 x i16> [[TMP24]] to <vscale x 4 x double>
; CHECK-VF4UF2-NEXT:    [[TMP29:%.*]] = sitofp <vscale x 4 x i16> [[TMP25]] to <vscale x 4 x double>
; CHECK-VF4UF2-NEXT:    [[TMP30:%.*]] = fmul fast <vscale x 4 x double> [[TMP28]], [[BROADCAST_SPLAT]]
; CHECK-VF4UF2-NEXT:    [[TMP31:%.*]] = fmul fast <vscale x 4 x double> [[TMP29]], [[BROADCAST_SPLAT]]
; CHECK-VF4UF2-NEXT:    [[TMP32:%.*]] = fsub fast <vscale x 4 x double> [[TMP26]], [[TMP30]]
; CHECK-VF4UF2-NEXT:    [[TMP33:%.*]] = fsub fast <vscale x 4 x double> [[TMP27]], [[TMP31]]
; CHECK-VF4UF2-NEXT:    [[TMP34:%.*]] = getelementptr inbounds double, ptr [[B]], i64 [[OFFSET_IDX]]
; CHECK-VF4UF2-NEXT:    [[TMP35:%.*]] = getelementptr inbounds double, ptr [[TMP34]], i64 [[TMP13]]
; CHECK-VF4UF2-NEXT:    store <vscale x 4 x double> [[TMP32]], ptr [[TMP34]], align 8, !alias.scope [[META9:![0-9]+]], !noalias [[META6]]
; CHECK-VF4UF2-NEXT:    store <vscale x 4 x double> [[TMP33]], ptr [[TMP35]], align 8, !alias.scope [[META9]], !noalias [[META6]]
; CHECK-VF4UF2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP12]]
; CHECK-VF4UF2-NEXT:    [[TMP39:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-VF4UF2-NEXT:    br i1 [[TMP39]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
; CHECK-VF4UF2:       [[MIDDLE_BLOCK]]:
; CHECK-VF4UF2-NEXT:    [[TMP40:%.*]] = call i32 @llvm.vscale.i32()
; CHECK-VF4UF2-NEXT:    [[TMP41:%.*]] = mul nuw i32 [[TMP40]], 4
; CHECK-VF4UF2-NEXT:    [[TMP42:%.*]] = sub i32 [[TMP41]], 1
; CHECK-VF4UF2-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <vscale x 4 x i16> [[WIDE_LOAD4]], i32 [[TMP42]]
; CHECK-VF4UF2-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
; CHECK-VF4UF2-NEXT:    br i1 [[CMP_N]], [[FOR_END_LOOPEXIT:label %.*]], label %[[SCALAR_PH]]
; CHECK-VF4UF2:       [[SCALAR_PH]]:
; CHECK-VF4UF2-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ [[TMP0]], %[[FOR_PREHEADER]] ], [ [[TMP0]], %[[VECTOR_MEMCHECK]] ]
; CHECK-VF4UF2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[TMP15]], %[[MIDDLE_BLOCK]] ], [ 1, %[[FOR_PREHEADER]] ], [ 1, %[[VECTOR_MEMCHECK]] ]
; CHECK-VF4UF2-NEXT:    br label %[[SCALAR_BODY:.*]]
; CHECK-VF4UF2:       [[SCALAR_BODY]]:
;
; Check also that the casts were not moved needlessly.
entry:
  %0 = load i16, ptr %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, ptr %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 ]
  %iv = phi i64 [ %iv.next, %scalar.body ], [ 1, %for.preheader ]
  %arrayidx5 = getelementptr inbounds i16, ptr %a, i64 %iv
  %2 = load i16, ptr %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, ptr %b, i64 %iv
  store double %sub13, ptr %arrayidx15, align 8
  %iv.next = add nuw nsw i64 %iv, 1
  %lftr.wideiv = trunc i64 %iv.next to i32
  %exitcond = icmp eq i32 %lftr.wideiv, %n
  br i1 %exitcond, label %for.end.loopexit, label %scalar.body, !llvm.loop !0

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

for.end:
  ret void
}

define i64 @constant_folded_previous_value() {
; CHECK-VF4UF1-LABEL: define i64 @constant_folded_previous_value() {
; CHECK-VF4UF1-NEXT:  [[ENTRY:.*]]:
; CHECK-VF4UF1-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
; CHECK-VF4UF1-NEXT:    [[TMP1:%.*]] = shl nuw i64 [[TMP0]], 2
; CHECK-VF4UF1-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1000, [[TMP1]]
; CHECK-VF4UF1-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK-VF4UF1:       [[VECTOR_PH]]:
; CHECK-VF4UF1-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
; CHECK-VF4UF1-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 4
; CHECK-VF4UF1-NEXT:    [[N_MOD_VF:%.*]] = urem i64 1000, [[TMP3]]
; CHECK-VF4UF1-NEXT:    [[N_VEC:%.*]] = sub i64 1000, [[N_MOD_VF]]
; CHECK-VF4UF1-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK-VF4UF1:       [[VECTOR_BODY]]:
; CHECK-VF4UF1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-VF4UF1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP3]]
; CHECK-VF4UF1-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-VF4UF1-NEXT:    br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
; CHECK-VF4UF1:       [[MIDDLE_BLOCK]]:
; CHECK-VF4UF1-NEXT:    [[CMP_N:%.*]] = icmp eq i64 1000, [[N_VEC]]
; CHECK-VF4UF1-NEXT:    br i1 [[CMP_N]], [[FOR_END:label %.*]], label %[[SCALAR_PH]]
; CHECK-VF4UF1:       [[SCALAR_PH]]:
; CHECK-VF4UF1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-VF4UF1-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i64 [ 1, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-VF4UF1-NEXT:    br label %[[SCALAR_BODY:.*]]
; CHECK-VF4UF1:       [[SCALAR_BODY]]:
;
; CHECK-VF4UF2-LABEL: define i64 @constant_folded_previous_value() {
; CHECK-VF4UF2-NEXT:  [[ENTRY:.*]]:
; CHECK-VF4UF2-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
; CHECK-VF4UF2-NEXT:    [[TMP1:%.*]] = shl nuw i64 [[TMP0]], 3
; CHECK-VF4UF2-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1000, [[TMP1]]
; CHECK-VF4UF2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK-VF4UF2:       [[VECTOR_PH]]:
; CHECK-VF4UF2-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
; CHECK-VF4UF2-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 8
; CHECK-VF4UF2-NEXT:    [[N_MOD_VF:%.*]] = urem i64 1000, [[TMP3]]
; CHECK-VF4UF2-NEXT:    [[N_VEC:%.*]] = sub i64 1000, [[N_MOD_VF]]
; CHECK-VF4UF2-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK-VF4UF2:       [[VECTOR_BODY]]:
; CHECK-VF4UF2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-VF4UF2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP3]]
; CHECK-VF4UF2-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-VF4UF2-NEXT:    br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
; CHECK-VF4UF2:       [[MIDDLE_BLOCK]]:
; CHECK-VF4UF2-NEXT:    [[CMP_N:%.*]] = icmp eq i64 1000, [[N_VEC]]
; CHECK-VF4UF2-NEXT:    br i1 [[CMP_N]], [[FOR_END:label %.*]], label %[[SCALAR_PH]]
; CHECK-VF4UF2:       [[SCALAR_PH]]:
; CHECK-VF4UF2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-VF4UF2-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i64 [ 1, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-VF4UF2-NEXT:    br label %[[SCALAR_BODY:.*]]
; CHECK-VF4UF2:       [[SCALAR_BODY]]:
;
entry:
  br label %scalar.body

scalar.body:
  %i = phi i64 [ 0, %entry ], [ %i.next, %scalar.body ]
  %tmp2 = phi i64 [ 0, %entry ], [ %tmp3, %scalar.body ]
  %tmp3 = add i64 0, 1
  %i.next = add nuw nsw i64 %i, 1
  %cond = icmp eq i64 %i.next, 1000
  br i1 %cond, label %for.end, label %scalar.body, !llvm.loop !0

for.end:
  ret i64 %tmp2
}

; 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).
define i32 @extract_second_last_iteration(ptr %cval, i32 %x)  {
; CHECK-VF4UF1-LABEL: define i32 @extract_second_last_iteration(
; CHECK-VF4UF1-SAME: ptr [[CVAL:%.*]], i32 [[X:%.*]]) {
; CHECK-VF4UF1-NEXT:  [[ENTRY:.*]]:
; CHECK-VF4UF1-NEXT:    [[TMP0:%.*]] = call i32 @llvm.vscale.i32()
; CHECK-VF4UF1-NEXT:    [[TMP1:%.*]] = shl nuw i32 [[TMP0]], 2
; CHECK-VF4UF1-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 96, [[TMP1]]
; CHECK-VF4UF1-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK-VF4UF1:       [[VECTOR_PH]]:
; CHECK-VF4UF1-NEXT:    [[TMP2:%.*]] = call i32 @llvm.vscale.i32()
; CHECK-VF4UF1-NEXT:    [[TMP3:%.*]] = mul nuw i32 [[TMP2]], 4
; CHECK-VF4UF1-NEXT:    [[N_MOD_VF:%.*]] = urem i32 96, [[TMP3]]
; CHECK-VF4UF1-NEXT:    [[N_VEC:%.*]] = sub i32 96, [[N_MOD_VF]]
; CHECK-VF4UF1-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[X]], i64 0
; CHECK-VF4UF1-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
; CHECK-VF4UF1-NEXT:    [[TMP6:%.*]] = call <vscale x 4 x i32> @llvm.stepvector.nxv4i32()
; CHECK-VF4UF1-NEXT:    [[TMP7:%.*]] = mul <vscale x 4 x i32> [[TMP6]], splat (i32 1)
; CHECK-VF4UF1-NEXT:    [[INDUCTION:%.*]] = add <vscale x 4 x i32> zeroinitializer, [[TMP7]]
; CHECK-VF4UF1-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[TMP3]], i64 0
; CHECK-VF4UF1-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[DOTSPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
; CHECK-VF4UF1-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK-VF4UF1:       [[VECTOR_BODY]]:
; CHECK-VF4UF1-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-VF4UF1-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 4 x i32> [ [[INDUCTION]], %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-VF4UF1-NEXT:    [[TMP9:%.*]] = add <vscale x 4 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
; CHECK-VF4UF1-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP3]]
; CHECK-VF4UF1-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 4 x i32> [[VEC_IND]], [[DOTSPLAT]]
; CHECK-VF4UF1-NEXT:    [[TMP10:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-VF4UF1-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
; CHECK-VF4UF1:       [[MIDDLE_BLOCK]]:
; CHECK-VF4UF1-NEXT:    [[TMP11:%.*]] = call i32 @llvm.vscale.i32()
; CHECK-VF4UF1-NEXT:    [[TMP12:%.*]] = mul nuw i32 [[TMP11]], 4
; CHECK-VF4UF1-NEXT:    [[TMP13:%.*]] = sub i32 [[TMP12]], 1
; CHECK-VF4UF1-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <vscale x 4 x i32> [[TMP9]], i32 [[TMP13]]
; CHECK-VF4UF1-NEXT:    [[TMP14:%.*]] = call i32 @llvm.vscale.i32()
; CHECK-VF4UF1-NEXT:    [[TMP15:%.*]] = mul nuw i32 [[TMP14]], 4
; CHECK-VF4UF1-NEXT:    [[TMP16:%.*]] = sub i32 [[TMP15]], 2
; CHECK-VF4UF1-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <vscale x 4 x i32> [[TMP9]], i32 [[TMP16]]
; CHECK-VF4UF1-NEXT:    [[CMP_N:%.*]] = icmp eq i32 96, [[N_VEC]]
; CHECK-VF4UF1-NEXT:    br i1 [[CMP_N]], [[FOR_END:label %.*]], label %[[SCALAR_PH]]
; CHECK-VF4UF1:       [[SCALAR_PH]]:
; CHECK-VF4UF1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-VF4UF1-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-VF4UF1-NEXT:    br label %[[FOR_BODY:.*]]
; CHECK-VF4UF1:       [[FOR_BODY]]:
;
; CHECK-VF4UF2-LABEL: define i32 @extract_second_last_iteration(
; CHECK-VF4UF2-SAME: ptr [[CVAL:%.*]], i32 [[X:%.*]]) {
; CHECK-VF4UF2-NEXT:  [[ENTRY:.*]]:
; CHECK-VF4UF2-NEXT:    [[TMP0:%.*]] = call i32 @llvm.vscale.i32()
; CHECK-VF4UF2-NEXT:    [[TMP1:%.*]] = shl nuw i32 [[TMP0]], 3
; CHECK-VF4UF2-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 96, [[TMP1]]
; CHECK-VF4UF2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK-VF4UF2:       [[VECTOR_PH]]:
; CHECK-VF4UF2-NEXT:    [[TMP4:%.*]] = call i32 @llvm.vscale.i32()
; CHECK-VF4UF2-NEXT:    [[TMP5:%.*]] = mul nuw i32 [[TMP4]], 4
; CHECK-VF4UF2-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[TMP5]], i64 0
; CHECK-VF4UF2-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT1]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
; CHECK-VF4UF2-NEXT:    [[TMP6:%.*]] = mul nuw i32 [[TMP5]], 2
; CHECK-VF4UF2-NEXT:    [[N_MOD_VF:%.*]] = urem i32 96, [[TMP6]]
; CHECK-VF4UF2-NEXT:    [[N_VEC:%.*]] = sub i32 96, [[N_MOD_VF]]
; CHECK-VF4UF2-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[X]], i64 0
; CHECK-VF4UF2-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
; CHECK-VF4UF2-NEXT:    [[TMP7:%.*]] = call <vscale x 4 x i32> @llvm.stepvector.nxv4i32()
; CHECK-VF4UF2-NEXT:    [[TMP8:%.*]] = mul <vscale x 4 x i32> [[TMP7]], splat (i32 1)
; CHECK-VF4UF2-NEXT:    [[INDUCTION:%.*]] = add <vscale x 4 x i32> zeroinitializer, [[TMP8]]
; CHECK-VF4UF2-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK-VF4UF2:       [[VECTOR_BODY]]:
; CHECK-VF4UF2-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-VF4UF2-NEXT:    [[STEP_ADD:%.*]] = phi <vscale x 4 x i32> [ [[INDUCTION]], %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT1:%.*]], %[[VECTOR_BODY]] ]
; CHECK-VF4UF2-NEXT:    [[VEC_IND_NEXT:%.*]] = add <vscale x 4 x i32> [[STEP_ADD]], [[BROADCAST_SPLAT2]]
; CHECK-VF4UF2-NEXT:    [[TMP9:%.*]] = add <vscale x 4 x i32> [[VEC_IND_NEXT]], [[BROADCAST_SPLAT]]
; CHECK-VF4UF2-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP6]]
; CHECK-VF4UF2-NEXT:    [[VEC_IND_NEXT1]] = add <vscale x 4 x i32> [[VEC_IND_NEXT]], [[BROADCAST_SPLAT2]]
; CHECK-VF4UF2-NEXT:    [[TMP10:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-VF4UF2-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
; CHECK-VF4UF2:       [[MIDDLE_BLOCK]]:
; CHECK-VF4UF2-NEXT:    [[TMP11:%.*]] = call i32 @llvm.vscale.i32()
; CHECK-VF4UF2-NEXT:    [[TMP12:%.*]] = mul nuw i32 [[TMP11]], 4
; CHECK-VF4UF2-NEXT:    [[TMP13:%.*]] = sub i32 [[TMP12]], 1
; CHECK-VF4UF2-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <vscale x 4 x i32> [[TMP9]], i32 [[TMP13]]
; CHECK-VF4UF2-NEXT:    [[TMP14:%.*]] = call i32 @llvm.vscale.i32()
; CHECK-VF4UF2-NEXT:    [[TMP15:%.*]] = mul nuw i32 [[TMP14]], 4
; CHECK-VF4UF2-NEXT:    [[TMP16:%.*]] = sub i32 [[TMP15]], 2
; CHECK-VF4UF2-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <vscale x 4 x i32> [[TMP9]], i32 [[TMP16]]
; CHECK-VF4UF2-NEXT:    [[CMP_N:%.*]] = icmp eq i32 96, [[N_VEC]]
; CHECK-VF4UF2-NEXT:    br i1 [[CMP_N]], [[FOR_END:label %.*]], label %[[SCALAR_PH]]
; CHECK-VF4UF2:       [[SCALAR_PH]]:
; CHECK-VF4UF2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-VF4UF2-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-VF4UF2-NEXT:    br label %[[FOR_BODY:.*]]
; CHECK-VF4UF2:       [[FOR_BODY]]:
;
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, !llvm.loop !0

for.end:
  ret i32 %val.phi
}

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

; Check that the sext sank after the load in the vector loop.
define void @sink_after(ptr %a, ptr %b, i64 %n) {
; CHECK-VF4UF1-LABEL: define void @sink_after(
; CHECK-VF4UF1-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) {
; CHECK-VF4UF1-NEXT:  [[ENTRY:.*]]:
; CHECK-VF4UF1-NEXT:    [[DOTPRE:%.*]] = load i16, ptr [[A]], align 2
; CHECK-VF4UF1-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
; CHECK-VF4UF1-NEXT:    [[TMP1:%.*]] = shl nuw i64 [[TMP0]], 2
; CHECK-VF4UF1-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP1]]
; CHECK-VF4UF1-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
; CHECK-VF4UF1:       [[VECTOR_MEMCHECK]]:
; CHECK-VF4UF1-NEXT:    [[TMP2:%.*]] = shl i64 [[N]], 2
; CHECK-VF4UF1-NEXT:    [[SCEVGEP:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP2]]
; CHECK-VF4UF1-NEXT:    [[SCEVGEP1:%.*]] = getelementptr nuw i8, ptr [[A]], i64 2
; CHECK-VF4UF1-NEXT:    [[TMP3:%.*]] = shl i64 [[N]], 1
; CHECK-VF4UF1-NEXT:    [[TMP4:%.*]] = add i64 [[TMP3]], 2
; CHECK-VF4UF1-NEXT:    [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP4]]
; CHECK-VF4UF1-NEXT:    [[BOUND0:%.*]] = icmp ult ptr [[B]], [[SCEVGEP2]]
; CHECK-VF4UF1-NEXT:    [[BOUND1:%.*]] = icmp ult ptr [[SCEVGEP1]], [[SCEVGEP]]
; CHECK-VF4UF1-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
; CHECK-VF4UF1-NEXT:    br i1 [[FOUND_CONFLICT]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
; CHECK-VF4UF1:       [[VECTOR_PH]]:
; CHECK-VF4UF1-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
; CHECK-VF4UF1-NEXT:    [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 4
; CHECK-VF4UF1-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP6]]
; CHECK-VF4UF1-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; CHECK-VF4UF1-NEXT:    [[TMP9:%.*]] = call i32 @llvm.vscale.i32()
; CHECK-VF4UF1-NEXT:    [[TMP10:%.*]] = mul nuw i32 [[TMP9]], 4
; CHECK-VF4UF1-NEXT:    [[TMP11:%.*]] = sub i32 [[TMP10]], 1
; CHECK-VF4UF1-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <vscale x 4 x i16> poison, i16 [[DOTPRE]], i32 [[TMP11]]
; CHECK-VF4UF1-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK-VF4UF1:       [[VECTOR_BODY]]:
; CHECK-VF4UF1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-VF4UF1-NEXT:    [[VECTOR_RECUR:%.*]] = phi <vscale x 4 x i16> [ [[VECTOR_RECUR_INIT]], %[[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], %[[VECTOR_BODY]] ]
; CHECK-VF4UF1-NEXT:    [[TMP12:%.*]] = add nuw nsw i64 [[INDEX]], 1
; CHECK-VF4UF1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i16, ptr [[A]], i64 [[TMP12]]
; CHECK-VF4UF1-NEXT:    [[WIDE_LOAD]] = load <vscale x 4 x i16>, ptr [[TMP13]], align 2, !alias.scope [[META17:![0-9]+]]
; CHECK-VF4UF1-NEXT:    [[TMP15:%.*]] = call <vscale x 4 x i16> @llvm.vector.splice.nxv4i16(<vscale x 4 x i16> [[VECTOR_RECUR]], <vscale x 4 x i16> [[WIDE_LOAD]], i32 -1)
; CHECK-VF4UF1-NEXT:    [[TMP16:%.*]] = sext <vscale x 4 x i16> [[TMP15]] to <vscale x 4 x i32>
; CHECK-VF4UF1-NEXT:    [[TMP17:%.*]] = sext <vscale x 4 x i16> [[WIDE_LOAD]] to <vscale x 4 x i32>
; CHECK-VF4UF1-NEXT:    [[TMP18:%.*]] = mul nsw <vscale x 4 x i32> [[TMP17]], [[TMP16]]
; CHECK-VF4UF1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
; CHECK-VF4UF1-NEXT:    store <vscale x 4 x i32> [[TMP18]], ptr [[TMP19]], align 4, !alias.scope [[META20:![0-9]+]], !noalias [[META17]]
; CHECK-VF4UF1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP6]]
; CHECK-VF4UF1-NEXT:    [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-VF4UF1-NEXT:    br i1 [[TMP21]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
; CHECK-VF4UF1:       [[MIDDLE_BLOCK]]:
; CHECK-VF4UF1-NEXT:    [[TMP22:%.*]] = call i32 @llvm.vscale.i32()
; CHECK-VF4UF1-NEXT:    [[TMP23:%.*]] = mul nuw i32 [[TMP22]], 4
; CHECK-VF4UF1-NEXT:    [[TMP24:%.*]] = sub i32 [[TMP23]], 1
; CHECK-VF4UF1-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <vscale x 4 x i16> [[WIDE_LOAD]], i32 [[TMP24]]
; CHECK-VF4UF1-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; CHECK-VF4UF1-NEXT:    br i1 [[CMP_N]], [[FOR_END:label %.*]], label %[[SCALAR_PH]]
; CHECK-VF4UF1:       [[SCALAR_PH]]:
; CHECK-VF4UF1-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ [[DOTPRE]], %[[ENTRY]] ], [ [[DOTPRE]], %[[VECTOR_MEMCHECK]] ]
; CHECK-VF4UF1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
; CHECK-VF4UF1-NEXT:    br label %[[FOR_BODY:.*]]
; CHECK-VF4UF1:       [[FOR_BODY]]:
;
; CHECK-VF4UF2-LABEL: define void @sink_after(
; CHECK-VF4UF2-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) {
; CHECK-VF4UF2-NEXT:  [[ENTRY:.*]]:
; CHECK-VF4UF2-NEXT:    [[DOTPRE:%.*]] = load i16, ptr [[A]], align 2
; CHECK-VF4UF2-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
; CHECK-VF4UF2-NEXT:    [[TMP1:%.*]] = shl nuw i64 [[TMP0]], 3
; CHECK-VF4UF2-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP1]]
; CHECK-VF4UF2-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
; CHECK-VF4UF2:       [[VECTOR_MEMCHECK]]:
; CHECK-VF4UF2-NEXT:    [[TMP2:%.*]] = shl i64 [[N]], 2
; CHECK-VF4UF2-NEXT:    [[SCEVGEP:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP2]]
; CHECK-VF4UF2-NEXT:    [[SCEVGEP1:%.*]] = getelementptr nuw i8, ptr [[A]], i64 2
; CHECK-VF4UF2-NEXT:    [[TMP3:%.*]] = shl i64 [[N]], 1
; CHECK-VF4UF2-NEXT:    [[TMP4:%.*]] = add i64 [[TMP3]], 2
; CHECK-VF4UF2-NEXT:    [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP4]]
; CHECK-VF4UF2-NEXT:    [[BOUND0:%.*]] = icmp ult ptr [[B]], [[SCEVGEP2]]
; CHECK-VF4UF2-NEXT:    [[BOUND1:%.*]] = icmp ult ptr [[SCEVGEP1]], [[SCEVGEP]]
; CHECK-VF4UF2-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
; CHECK-VF4UF2-NEXT:    br i1 [[FOUND_CONFLICT]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
; CHECK-VF4UF2:       [[VECTOR_PH]]:
; CHECK-VF4UF2-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
; CHECK-VF4UF2-NEXT:    [[TMP7:%.*]] = mul nuw i64 [[TMP5]], 4
; CHECK-VF4UF2-NEXT:    [[TMP6:%.*]] = mul nuw i64 [[TMP7]], 2
; CHECK-VF4UF2-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP6]]
; CHECK-VF4UF2-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; CHECK-VF4UF2-NEXT:    [[TMP9:%.*]] = call i32 @llvm.vscale.i32()
; CHECK-VF4UF2-NEXT:    [[TMP10:%.*]] = mul nuw i32 [[TMP9]], 4
; CHECK-VF4UF2-NEXT:    [[TMP11:%.*]] = sub i32 [[TMP10]], 1
; CHECK-VF4UF2-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <vscale x 4 x i16> poison, i16 [[DOTPRE]], i32 [[TMP11]]
; CHECK-VF4UF2-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK-VF4UF2:       [[VECTOR_BODY]]:
; CHECK-VF4UF2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-VF4UF2-NEXT:    [[VECTOR_RECUR:%.*]] = phi <vscale x 4 x i16> [ [[VECTOR_RECUR_INIT]], %[[VECTOR_PH]] ], [ [[WIDE_LOAD3:%.*]], %[[VECTOR_BODY]] ]
; CHECK-VF4UF2-NEXT:    [[TMP12:%.*]] = add nuw nsw i64 [[INDEX]], 1
; CHECK-VF4UF2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i16, ptr [[A]], i64 [[TMP12]]
; CHECK-VF4UF2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i16, ptr [[TMP13]], i64 [[TMP7]]
; CHECK-VF4UF2-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i16>, ptr [[TMP13]], align 2, !alias.scope [[META17:![0-9]+]]
; CHECK-VF4UF2-NEXT:    [[WIDE_LOAD3]] = load <vscale x 4 x i16>, ptr [[TMP14]], align 2, !alias.scope [[META17]]
; CHECK-VF4UF2-NEXT:    [[TMP18:%.*]] = call <vscale x 4 x i16> @llvm.vector.splice.nxv4i16(<vscale x 4 x i16> [[VECTOR_RECUR]], <vscale x 4 x i16> [[WIDE_LOAD]], i32 -1)
; CHECK-VF4UF2-NEXT:    [[TMP19:%.*]] = call <vscale x 4 x i16> @llvm.vector.splice.nxv4i16(<vscale x 4 x i16> [[WIDE_LOAD]], <vscale x 4 x i16> [[WIDE_LOAD3]], i32 -1)
; CHECK-VF4UF2-NEXT:    [[TMP20:%.*]] = sext <vscale x 4 x i16> [[TMP18]] to <vscale x 4 x i32>
; CHECK-VF4UF2-NEXT:    [[TMP21:%.*]] = sext <vscale x 4 x i16> [[TMP19]] to <vscale x 4 x i32>
; CHECK-VF4UF2-NEXT:    [[TMP22:%.*]] = sext <vscale x 4 x i16> [[WIDE_LOAD]] to <vscale x 4 x i32>
; CHECK-VF4UF2-NEXT:    [[TMP23:%.*]] = sext <vscale x 4 x i16> [[WIDE_LOAD3]] to <vscale x 4 x i32>
; CHECK-VF4UF2-NEXT:    [[TMP24:%.*]] = mul nsw <vscale x 4 x i32> [[TMP22]], [[TMP20]]
; CHECK-VF4UF2-NEXT:    [[TMP25:%.*]] = mul nsw <vscale x 4 x i32> [[TMP23]], [[TMP21]]
; CHECK-VF4UF2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
; CHECK-VF4UF2-NEXT:    [[TMP27:%.*]] = getelementptr inbounds i32, ptr [[TMP26]], i64 [[TMP7]]
; CHECK-VF4UF2-NEXT:    store <vscale x 4 x i32> [[TMP24]], ptr [[TMP26]], align 4, !alias.scope [[META20:![0-9]+]], !noalias [[META17]]
; CHECK-VF4UF2-NEXT:    store <vscale x 4 x i32> [[TMP25]], ptr [[TMP27]], align 4, !alias.scope [[META20]], !noalias [[META17]]
; CHECK-VF4UF2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP6]]
; CHECK-VF4UF2-NEXT:    [[TMP31:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-VF4UF2-NEXT:    br i1 [[TMP31]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
; CHECK-VF4UF2:       [[MIDDLE_BLOCK]]:
; CHECK-VF4UF2-NEXT:    [[TMP32:%.*]] = call i32 @llvm.vscale.i32()
; CHECK-VF4UF2-NEXT:    [[TMP33:%.*]] = mul nuw i32 [[TMP32]], 4
; CHECK-VF4UF2-NEXT:    [[TMP34:%.*]] = sub i32 [[TMP33]], 1
; CHECK-VF4UF2-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <vscale x 4 x i16> [[WIDE_LOAD3]], i32 [[TMP34]]
; CHECK-VF4UF2-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; CHECK-VF4UF2-NEXT:    br i1 [[CMP_N]], [[FOR_END:label %.*]], label %[[SCALAR_PH]]
; CHECK-VF4UF2:       [[SCALAR_PH]]:
; CHECK-VF4UF2-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ [[DOTPRE]], %[[ENTRY]] ], [ [[DOTPRE]], %[[VECTOR_MEMCHECK]] ]
; CHECK-VF4UF2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
; CHECK-VF4UF2-NEXT:    br label %[[FOR_BODY:.*]]
; CHECK-VF4UF2:       [[FOR_BODY]]:
;
entry:
  %.pre = load i16, ptr %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, ptr %a, i64 %indvars.iv.next
  %1 = load i16, ptr %arrayidx2
  %conv3 = sext i16 %1 to i32
  %mul = mul nsw i32 %conv3, %conv
  %arrayidx5 = getelementptr inbounds i32, ptr %b, i64 %indvars.iv
  store i32 %mul, ptr %arrayidx5
  %exitcond = icmp eq i64 %indvars.iv.next, %n
  br i1 %exitcond, label %for.end, label %for.body, !llvm.loop !0

for.end:
  ret void
}

!0 = distinct !{!0, !1}
!1 = !{!"llvm.loop.vectorize.scalable.enable", i1 true}