; NOTE: Assertions have been autogenerated by utils/update_test_checks.py

; To test epilogue-vectorization we need to make sure that the vectorizer actually vectorizes the loop.
; Without a target triple this becomes difficult, unless we force vectorization through user hints.
; Currently user provided vectorization hints prevent epilogue vectorization unless the forced
; VF is the same as the epilogue vectorization VF. To make these tests target independent we'll use a
; trick where both VFs are forced to be the same value. Mismatching VFs are tested in target specific tests.
; RUN: opt -passes='loop-vectorize' -force-vector-width=4 -enable-epilogue-vectorization -epilogue-vectorization-force-VF=4 -S %s | FileCheck %s

; Some simpler cases are found profitable even without triple or user hints.
; RUN: opt -passes='loop-vectorize' -enable-epilogue-vectorization -epilogue-vectorization-force-VF=2 -S %s | FileCheck --check-prefix=CHECK-PROFITABLE-BY-DEFAULT %s

target datalayout = "e-m:e-i64:64-n32:64-v128:128:128"

define dso_local void @f1(ptr noalias %aa, ptr noalias %bb, ptr noalias %cc, i32 signext %N) {
; CHECK-LABEL: @f1(
; CHECK-NEXT:  entry:
; CHECK-NEXT:    [[CMP1:%.*]] = icmp sgt i32 [[N:%.*]], 0
; CHECK-NEXT:    br i1 [[CMP1]], label [[ITER_CHECK:%.*]], label [[FOR_END:%.*]]
; CHECK:       iter.check:
; CHECK-NEXT:    [[WIDE_TRIP_COUNT:%.*]] = zext i32 [[N]] to i64
; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 4
; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]]
; CHECK:       vector.main.loop.iter.check:
; CHECK-NEXT:    [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 4
; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK:       vector.ph:
; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 4
; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF]]
; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
; CHECK:       vector.body:
; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[BB:%.*]], i64 [[TMP0]]
; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[CC:%.*]], i64 [[TMP0]]
; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, ptr [[TMP3]], i32 0
; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x float>, ptr [[TMP4]], align 4
; CHECK-NEXT:    [[TMP5:%.*]] = fadd fast <4 x float> [[WIDE_LOAD]], [[WIDE_LOAD2]]
; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[AA:%.*]], i64 [[TMP0]]
; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i32 0
; CHECK-NEXT:    store <4 x float> [[TMP5]], ptr [[TMP7]], align 4
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; CHECK:       middle.block:
; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC]]
; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
; CHECK:       vec.epilog.iter.check:
; CHECK-NEXT:    [[N_VEC_REMAINING:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_VEC]]
; CHECK-NEXT:    [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 4
; CHECK-NEXT:    br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]]
; CHECK:       vec.epilog.ph:
; CHECK-NEXT:    [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
; CHECK-NEXT:    [[N_MOD_VF3:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 4
; CHECK-NEXT:    [[N_VEC4:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF3]]
; CHECK-NEXT:    br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
; CHECK:       vec.epilog.vector.body:
; CHECK-NEXT:    [[INDEX6:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT9:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
; CHECK-NEXT:    [[TMP9:%.*]] = add i64 [[INDEX6]], 0
; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, ptr [[BB]], i64 [[TMP9]]
; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[TMP10]], i32 0
; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x float>, ptr [[TMP11]], align 4
; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[CC]], i64 [[TMP9]]
; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[TMP12]], i32 0
; CHECK-NEXT:    [[WIDE_LOAD8:%.*]] = load <4 x float>, ptr [[TMP13]], align 4
; CHECK-NEXT:    [[TMP14:%.*]] = fadd fast <4 x float> [[WIDE_LOAD7]], [[WIDE_LOAD8]]
; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds float, ptr [[AA]], i64 [[TMP9]]
; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, ptr [[TMP15]], i32 0
; CHECK-NEXT:    store <4 x float> [[TMP14]], ptr [[TMP16]], align 4
; CHECK-NEXT:    [[INDEX_NEXT9]] = add nuw i64 [[INDEX6]], 4
; CHECK-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT9]], [[N_VEC4]]
; CHECK-NEXT:    br i1 [[TMP17]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
; CHECK:       vec.epilog.middle.block:
; CHECK-NEXT:    [[CMP_N5:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC4]]
; CHECK-NEXT:    br i1 [[CMP_N5]], label [[FOR_END_LOOPEXIT]], label [[VEC_EPILOG_SCALAR_PH]]
; CHECK:       vec.epilog.scalar.ph:
; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC4]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK]] ]
; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
; CHECK:       for.body:
; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[BB]], i64 [[INDVARS_IV]]
; CHECK-NEXT:    [[TMP18:%.*]] = load float, ptr [[ARRAYIDX]], align 4
; CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, ptr [[CC]], i64 [[INDVARS_IV]]
; CHECK-NEXT:    [[TMP19:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
; CHECK-NEXT:    [[ADD:%.*]] = fadd fast float [[TMP18]], [[TMP19]]
; CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, ptr [[AA]], i64 [[INDVARS_IV]]
; CHECK-NEXT:    store float [[ADD]], ptr [[ARRAYIDX4]], align 4
; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp ne i64 [[INDVARS_IV_NEXT]], [[WIDE_TRIP_COUNT]]
; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_BODY]], label [[FOR_END_LOOPEXIT]], !llvm.loop [[LOOP4:![0-9]+]]
; CHECK:       for.end.loopexit:
; CHECK-NEXT:    br label [[FOR_END]]
; CHECK:       for.end:
; CHECK-NEXT:    ret void
;
; CHECK-PROFITABLE-BY-DEFAULT-LABEL: @f1(
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:  entry:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[CMP1:%.*]] = icmp sgt i32 [[N:%.*]], 0
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 [[CMP1]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_END:%.*]]
; CHECK-PROFITABLE-BY-DEFAULT:       for.body.preheader:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[WIDE_TRIP_COUNT:%.*]] = zext i32 [[N]] to i64
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br label [[FOR_BODY:%.*]]
; CHECK-PROFITABLE-BY-DEFAULT:       for.body:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ 0, [[FOR_BODY_PREHEADER]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[BB:%.*]], i64 [[INDVARS_IV]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP0:%.*]] = load float, ptr [[ARRAYIDX]], align 4
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, ptr [[CC:%.*]], i64 [[INDVARS_IV]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP1:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[ADD:%.*]] = fadd fast float [[TMP0]], [[TMP1]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, ptr [[AA:%.*]], i64 [[INDVARS_IV]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    store float [[ADD]], ptr [[ARRAYIDX4]], align 4
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[EXITCOND:%.*]] = icmp ne i64 [[INDVARS_IV_NEXT]], [[WIDE_TRIP_COUNT]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 [[EXITCOND]], label [[FOR_BODY]], label [[FOR_END_LOOPEXIT:%.*]]
; CHECK-PROFITABLE-BY-DEFAULT:       for.end.loopexit:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br label [[FOR_END]]
; CHECK-PROFITABLE-BY-DEFAULT:       for.end:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    ret void
;
entry:
  %cmp1 = icmp sgt i32 %N, 0
  br i1 %cmp1, label %for.body.preheader, label %for.end

for.body.preheader:                               ; preds = %entry
  %wide.trip.count = zext i32 %N to i64
  br label %for.body

for.body:                                         ; preds = %for.body.preheader, %for.body
  %indvars.iv = phi i64 [ 0, %for.body.preheader ], [ %indvars.iv.next, %for.body ]
  %arrayidx = getelementptr inbounds float, ptr %bb, i64 %indvars.iv
  %0 = load float, ptr %arrayidx, align 4
  %arrayidx2 = getelementptr inbounds float, ptr %cc, i64 %indvars.iv
  %1 = load float, ptr %arrayidx2, align 4
  %add = fadd fast float %0, %1
  %arrayidx4 = getelementptr inbounds float, ptr %aa, i64 %indvars.iv
  store float %add, ptr %arrayidx4, align 4
  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
  %exitcond = icmp ne i64 %indvars.iv.next, %wide.trip.count
  br i1 %exitcond, label %for.body, label %for.end.loopexit

for.end.loopexit:                                 ; preds = %for.body
  br label %for.end

for.end:                                          ; preds = %for.end.loopexit, %entry
  ret void
}

define dso_local signext i32 @f2(ptr noalias %A, ptr noalias %B, i32 signext %n) {
; CHECK-LABEL: @f2(
; CHECK-NEXT:  entry:
; CHECK-NEXT:    [[CMP1:%.*]] = icmp sgt i32 [[N:%.*]], 1
; CHECK-NEXT:    br i1 [[CMP1]], label [[ITER_CHECK:%.*]], label [[FOR_END:%.*]]
; CHECK:       iter.check:
; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
; CHECK-NEXT:    [[WIDE_TRIP_COUNT:%.*]] = zext i32 [[TMP0]] to i64
; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 4
; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
; CHECK:       vector.scevcheck:
; CHECK-NEXT:    [[TMP1:%.*]] = add nsw i64 [[WIDE_TRIP_COUNT]], -1
; CHECK-NEXT:    [[TMP2:%.*]] = trunc i64 [[TMP1]] to i32
; CHECK-NEXT:    [[MUL:%.*]] = call { i32, i1 } @llvm.umul.with.overflow.i32(i32 1, i32 [[TMP2]])
; CHECK-NEXT:    [[MUL_RESULT:%.*]] = extractvalue { i32, i1 } [[MUL]], 0
; CHECK-NEXT:    [[MUL_OVERFLOW:%.*]] = extractvalue { i32, i1 } [[MUL]], 1
; CHECK-NEXT:    [[TMP3:%.*]] = sub i32 [[TMP0]], [[MUL_RESULT]]
; CHECK-NEXT:    [[TMP4:%.*]] = icmp sgt i32 [[TMP3]], [[TMP0]]
; CHECK-NEXT:    [[TMP5:%.*]] = or i1 [[TMP4]], [[MUL_OVERFLOW]]
; CHECK-NEXT:    [[TMP6:%.*]] = icmp ugt i64 [[TMP1]], 4294967295
; CHECK-NEXT:    [[TMP7:%.*]] = or i1 [[TMP5]], [[TMP6]]
; CHECK-NEXT:    br i1 [[TMP7]], label [[VEC_EPILOG_SCALAR_PH]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]]
; CHECK:       vector.main.loop.iter.check:
; CHECK-NEXT:    [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 4
; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK:       vector.ph:
; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 4
; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF]]
; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
; CHECK:       vector.body:
; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = trunc i64 [[INDEX]] to i32
; CHECK-NEXT:    [[TMP8:%.*]] = add i32 [[OFFSET_IDX]], 0
; CHECK-NEXT:    [[TMP9:%.*]] = add i64 [[INDEX]], 0
; CHECK-NEXT:    [[TMP10:%.*]] = xor i32 [[TMP8]], -1
; CHECK-NEXT:    [[TMP11:%.*]] = add i32 [[TMP10]], [[N]]
; CHECK-NEXT:    [[TMP12:%.*]] = sext i32 [[TMP11]] to i64
; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[B:%.*]], i64 [[TMP12]]
; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, ptr [[TMP13]], i32 0
; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds float, ptr [[TMP14]], i32 -3
; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP15]], align 4
; CHECK-NEXT:    [[REVERSE:%.*]] = shufflevector <4 x float> [[WIDE_LOAD]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
; CHECK-NEXT:    [[TMP16:%.*]] = fadd fast <4 x float> [[REVERSE]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>
; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[TMP9]]
; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds float, ptr [[TMP17]], i32 0
; CHECK-NEXT:    store <4 x float> [[TMP16]], ptr [[TMP18]], align 4
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
; CHECK:       middle.block:
; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC]]
; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
; CHECK:       vec.epilog.iter.check:
; CHECK-NEXT:    [[IND_END4:%.*]] = trunc i64 [[N_VEC]] to i32
; CHECK-NEXT:    [[N_VEC_REMAINING:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_VEC]]
; CHECK-NEXT:    [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 4
; CHECK-NEXT:    br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]]
; CHECK:       vec.epilog.ph:
; CHECK-NEXT:    [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
; CHECK-NEXT:    [[N_MOD_VF2:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 4
; CHECK-NEXT:    [[N_VEC3:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF2]]
; CHECK-NEXT:    [[IND_END:%.*]] = trunc i64 [[N_VEC3]] to i32
; CHECK-NEXT:    br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
; CHECK:       vec.epilog.vector.body:
; CHECK-NEXT:    [[INDEX7:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT11:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
; CHECK-NEXT:    [[OFFSET_IDX8:%.*]] = trunc i64 [[INDEX7]] to i32
; CHECK-NEXT:    [[TMP20:%.*]] = add i32 [[OFFSET_IDX8]], 0
; CHECK-NEXT:    [[TMP21:%.*]] = add i64 [[INDEX7]], 0
; CHECK-NEXT:    [[TMP22:%.*]] = xor i32 [[TMP20]], -1
; CHECK-NEXT:    [[TMP23:%.*]] = add i32 [[TMP22]], [[N]]
; CHECK-NEXT:    [[TMP24:%.*]] = sext i32 [[TMP23]] to i64
; CHECK-NEXT:    [[TMP25:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP24]]
; CHECK-NEXT:    [[TMP26:%.*]] = getelementptr inbounds float, ptr [[TMP25]], i32 0
; CHECK-NEXT:    [[TMP27:%.*]] = getelementptr inbounds float, ptr [[TMP26]], i32 -3
; CHECK-NEXT:    [[WIDE_LOAD9:%.*]] = load <4 x float>, ptr [[TMP27]], align 4
; CHECK-NEXT:    [[REVERSE10:%.*]] = shufflevector <4 x float> [[WIDE_LOAD9]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
; CHECK-NEXT:    [[TMP28:%.*]] = fadd fast <4 x float> [[REVERSE10]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>
; CHECK-NEXT:    [[TMP29:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP21]]
; CHECK-NEXT:    [[TMP30:%.*]] = getelementptr inbounds float, ptr [[TMP29]], i32 0
; CHECK-NEXT:    store <4 x float> [[TMP28]], ptr [[TMP30]], align 4
; CHECK-NEXT:    [[INDEX_NEXT11]] = add nuw i64 [[INDEX7]], 4
; CHECK-NEXT:    [[TMP31:%.*]] = icmp eq i64 [[INDEX_NEXT11]], [[N_VEC3]]
; CHECK-NEXT:    br i1 [[TMP31]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
; CHECK:       vec.epilog.middle.block:
; CHECK-NEXT:    [[CMP_N6:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC3]]
; CHECK-NEXT:    br i1 [[CMP_N6]], label [[FOR_END_LOOPEXIT]], label [[VEC_EPILOG_SCALAR_PH]]
; CHECK:       vec.epilog.scalar.ph:
; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC3]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ITER_CHECK]] ]
; CHECK-NEXT:    [[BC_RESUME_VAL5:%.*]] = phi i32 [ [[IND_END]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[IND_END4]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ITER_CHECK]] ]
; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
; CHECK:       for.body:
; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
; CHECK-NEXT:    [[I_014:%.*]] = phi i32 [ [[BC_RESUME_VAL5]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
; CHECK-NEXT:    [[TMP32:%.*]] = xor i32 [[I_014]], -1
; CHECK-NEXT:    [[SUB2:%.*]] = add i32 [[TMP32]], [[N]]
; CHECK-NEXT:    [[IDXPROM:%.*]] = sext i32 [[SUB2]] to i64
; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IDXPROM]]
; CHECK-NEXT:    [[TMP33:%.*]] = load float, ptr [[ARRAYIDX]], align 4
; CHECK-NEXT:    [[CONV3:%.*]] = fadd fast float [[TMP33]], 1.000000e+00
; CHECK-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDVARS_IV]]
; CHECK-NEXT:    store float [[CONV3]], ptr [[ARRAYIDX5]], align 4
; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; CHECK-NEXT:    [[INC]] = add nuw nsw i32 [[I_014]], 1
; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp ne i64 [[INDVARS_IV_NEXT]], [[WIDE_TRIP_COUNT]]
; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_BODY]], label [[FOR_END_LOOPEXIT]], !llvm.loop [[LOOP7:![0-9]+]]
; CHECK:       for.end.loopexit:
; CHECK-NEXT:    br label [[FOR_END]]
; CHECK:       for.end:
; CHECK-NEXT:    ret i32 0
;
; CHECK-PROFITABLE-BY-DEFAULT-LABEL: @f2(
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:  entry:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[CMP1:%.*]] = icmp sgt i32 [[N:%.*]], 1
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 [[CMP1]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_END:%.*]]
; CHECK-PROFITABLE-BY-DEFAULT:       for.body.preheader:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP0:%.*]] = add i32 [[N]], -1
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[WIDE_TRIP_COUNT:%.*]] = zext i32 [[TMP0]] to i64
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br label [[FOR_BODY:%.*]]
; CHECK-PROFITABLE-BY-DEFAULT:       for.body:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ 0, [[FOR_BODY_PREHEADER]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[I_014:%.*]] = phi i32 [ 0, [[FOR_BODY_PREHEADER]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP1:%.*]] = xor i32 [[I_014]], -1
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[SUB2:%.*]] = add i32 [[TMP1]], [[N]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[IDXPROM:%.*]] = sext i32 [[SUB2]] to i64
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[B:%.*]], i64 [[IDXPROM]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP2:%.*]] = load float, ptr [[ARRAYIDX]], align 4
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[CONV3:%.*]] = fadd fast float [[TMP2]], 1.000000e+00
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDVARS_IV]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    store float [[CONV3]], ptr [[ARRAYIDX5]], align 4
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INC]] = add nuw nsw i32 [[I_014]], 1
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[EXITCOND:%.*]] = icmp ne i64 [[INDVARS_IV_NEXT]], [[WIDE_TRIP_COUNT]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 [[EXITCOND]], label [[FOR_BODY]], label [[FOR_END_LOOPEXIT:%.*]]
; CHECK-PROFITABLE-BY-DEFAULT:       for.end.loopexit:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br label [[FOR_END]]
; CHECK-PROFITABLE-BY-DEFAULT:       for.end:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    ret i32 0
;
entry:
  %cmp1 = icmp sgt i32 %n, 1
  br i1 %cmp1, label %for.body.preheader, label %for.end

for.body.preheader:                               ; preds = %entry
  %0 = add i32 %n, -1
  %wide.trip.count = zext i32 %0 to i64
  br label %for.body

for.body:                                         ; preds = %for.body.preheader, %for.body
  %indvars.iv = phi i64 [ 0, %for.body.preheader ], [ %indvars.iv.next, %for.body ]
  %i.014 = phi i32 [ 0, %for.body.preheader ], [ %inc, %for.body ]
  %1 = xor i32 %i.014, -1
  %sub2 = add i32 %1, %n
  %idxprom = sext i32 %sub2 to i64
  %arrayidx = getelementptr inbounds float, ptr %B, i64 %idxprom
  %2 = load float, ptr %arrayidx, align 4
  %conv3 = fadd fast float %2, 1.000000e+00
  %arrayidx5 = getelementptr inbounds float, ptr %A, i64 %indvars.iv
  store float %conv3, ptr %arrayidx5, align 4
  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
  %inc = add nuw nsw i32 %i.014, 1
  %exitcond = icmp ne i64 %indvars.iv.next, %wide.trip.count
  br i1 %exitcond, label %for.body, label %for.end.loopexit

for.end.loopexit:                                 ; preds = %for.body
  br label %for.end

for.end:                                          ; preds = %for.end.loopexit, %entry
  ret i32 0
}

define void @f3(ptr noalias %A, i64 %n) {
; CHECK-LABEL: @f3(
; CHECK-NEXT:  iter.check:
; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]]
; CHECK:       vector.main.loop.iter.check:
; CHECK-NEXT:    [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[N]], 4
; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK:       vector.ph:
; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
; CHECK:       vector.body:
; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[TMP0]]
; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
; CHECK-NEXT:    store <4 x i8> <i8 1, i8 1, i8 1, i8 1>, ptr [[TMP2]], align 1
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
; CHECK:       middle.block:
; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
; CHECK:       vec.epilog.iter.check:
; CHECK-NEXT:    [[N_VEC_REMAINING:%.*]] = sub i64 [[N]], [[N_VEC]]
; CHECK-NEXT:    [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 4
; CHECK-NEXT:    br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]]
; CHECK:       vec.epilog.ph:
; CHECK-NEXT:    [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
; CHECK-NEXT:    [[N_MOD_VF2:%.*]] = urem i64 [[N]], 4
; CHECK-NEXT:    [[N_VEC3:%.*]] = sub i64 [[N]], [[N_MOD_VF2]]
; CHECK-NEXT:    br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
; CHECK:       vec.epilog.vector.body:
; CHECK-NEXT:    [[INDEX5:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT6:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
; CHECK-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX5]], 0
; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[TMP4]]
; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[TMP5]], i32 0
; CHECK-NEXT:    store <4 x i8> <i8 1, i8 1, i8 1, i8 1>, ptr [[TMP6]], align 1
; CHECK-NEXT:    [[INDEX_NEXT6]] = add nuw i64 [[INDEX5]], 4
; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT6]], [[N_VEC3]]
; CHECK-NEXT:    br i1 [[TMP7]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
; CHECK:       vec.epilog.middle.block:
; CHECK-NEXT:    [[CMP_N4:%.*]] = icmp eq i64 [[N]], [[N_VEC3]]
; CHECK-NEXT:    br i1 [[CMP_N4]], label [[FOR_END_LOOPEXIT]], label [[VEC_EPILOG_SCALAR_PH]]
; CHECK:       vec.epilog.scalar.ph:
; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC3]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK:%.*]] ]
; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
; CHECK:       for.body:
; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[IV]]
; CHECK-NEXT:    store i8 1, ptr [[ARRAYIDX]], align 1
; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp ne i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_BODY]], label [[FOR_END_LOOPEXIT]], !llvm.loop [[LOOP10:![0-9]+]]
; CHECK:       for.end.loopexit:
; CHECK-NEXT:    br label [[FOR_END:%.*]]
; CHECK:       for.end:
; CHECK-NEXT:    ret void
;
; CHECK-PROFITABLE-BY-DEFAULT-LABEL: @f3(
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:  iter.check:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 2
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]]
; CHECK-PROFITABLE-BY-DEFAULT:       vector.main.loop.iter.check:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[N]], 4
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK-PROFITABLE-BY-DEFAULT:       vector.ph:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br label [[VECTOR_BODY:%.*]]
; CHECK-PROFITABLE-BY-DEFAULT:       vector.body:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[TMP0]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    store <4 x i8> <i8 1, i8 1, i8 1, i8 1>, ptr [[TMP2]], align 1
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; CHECK-PROFITABLE-BY-DEFAULT:       middle.block:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
; CHECK-PROFITABLE-BY-DEFAULT:       vec.epilog.iter.check:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[N_VEC_REMAINING:%.*]] = sub i64 [[N]], [[N_VEC]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 2
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]]
; CHECK-PROFITABLE-BY-DEFAULT:       vec.epilog.ph:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[N_MOD_VF2:%.*]] = urem i64 [[N]], 2
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[N_VEC3:%.*]] = sub i64 [[N]], [[N_MOD_VF2]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
; CHECK-PROFITABLE-BY-DEFAULT:       vec.epilog.vector.body:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDEX5:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT6:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX5]], 0
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[TMP4]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[TMP5]], i32 0
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    store <2 x i8> <i8 1, i8 1>, ptr [[TMP6]], align 1
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDEX_NEXT6]] = add nuw i64 [[INDEX5]], 2
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT6]], [[N_VEC3]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 [[TMP7]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
; CHECK-PROFITABLE-BY-DEFAULT:       vec.epilog.middle.block:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[CMP_N4:%.*]] = icmp eq i64 [[N]], [[N_VEC3]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 [[CMP_N4]], label [[FOR_END_LOOPEXIT]], label [[VEC_EPILOG_SCALAR_PH]]
; CHECK-PROFITABLE-BY-DEFAULT:       vec.epilog.scalar.ph:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC3]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK:%.*]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br label [[FOR_BODY:%.*]]
; CHECK-PROFITABLE-BY-DEFAULT:       for.body:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[IV]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    store i8 1, ptr [[ARRAYIDX]], align 1
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[EXITCOND:%.*]] = icmp ne i64 [[IV_NEXT]], [[N]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 [[EXITCOND]], label [[FOR_BODY]], label [[FOR_END_LOOPEXIT]], !llvm.loop [[LOOP4:![0-9]+]]
; CHECK-PROFITABLE-BY-DEFAULT:       for.end.loopexit:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br label [[FOR_END:%.*]]
; CHECK-PROFITABLE-BY-DEFAULT:       for.end:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    ret void
;
entry:
  br label %for.body

for.body:                                         ; preds = %for.body.preheader, %for.body
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
  %arrayidx = getelementptr inbounds i8, ptr %A, i64 %iv
  store i8 1, ptr %arrayidx, align 1
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond = icmp ne i64 %iv.next, %n
  br i1 %exitcond, label %for.body, label %for.end.loopexit

for.end.loopexit:                                 ; preds = %for.body
  br label %for.end

for.end:                                          ; preds = %for.end.loopexit, %entry
  ret void
}

; Test case for PR54745.
define void @induction_resume_value_requires_non_trivial_scev_expansion(ptr %dst) {
; CHECK-LABEL: @induction_resume_value_requires_non_trivial_scev_expansion(
; CHECK-NEXT:  entry:
; CHECK-NEXT:    br label [[ITER_CHECK:%.*]]
; CHECK:       iter.check:
; CHECK-NEXT:    [[INDUCTION_IV:%.*]] = phi i8 [ [[INDUCTION_IV_NEXT:%.*]], [[OUTER_LATCH:%.*]] ], [ -56, [[ENTRY:%.*]] ]
; CHECK-NEXT:    [[INDVAR:%.*]] = phi i8 [ [[INDVAR_NEXT:%.*]], [[OUTER_LATCH]] ], [ 0, [[ENTRY]] ]
; CHECK-NEXT:    [[OUTER_IV:%.*]] = phi i64 [ 2, [[ENTRY]] ], [ [[OUTER_IV_NEXT:%.*]], [[OUTER_LATCH]] ]
; CHECK-NEXT:    [[P2:%.*]] = phi i32 [ -202, [[ENTRY]] ], [ [[ADD:%.*]], [[OUTER_LATCH]] ]
; CHECK-NEXT:    [[TMP0:%.*]] = mul i8 [[INDVAR]], -1
; CHECK-NEXT:    [[TMP1:%.*]] = add i8 [[TMP0]], -3
; CHECK-NEXT:    [[TRUNC_IV:%.*]] = trunc i64 [[OUTER_IV]] to i32
; CHECK-NEXT:    [[ADD]] = add i32 [[P2]], [[TRUNC_IV]]
; CHECK-NEXT:    [[TRUNC_ADD:%.*]] = trunc i32 [[ADD]] to i8
; CHECK-NEXT:    br i1 false, label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]]
; CHECK:       vector.main.loop.iter.check:
; CHECK-NEXT:    br i1 false, label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK:       vector.ph:
; CHECK-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <4 x i8> poison, i8 [[INDUCTION_IV]], i64 0
; CHECK-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <4 x i8> [[DOTSPLATINSERT]], <4 x i8> poison, <4 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP2:%.*]] = mul <4 x i8> <i8 0, i8 1, i8 2, i8 3>, [[DOTSPLAT]]
; CHECK-NEXT:    [[INDUCTION:%.*]] = add <4 x i8> zeroinitializer, [[TMP2]]
; CHECK-NEXT:    [[TMP3:%.*]] = mul i8 [[INDUCTION_IV]], 4
; CHECK-NEXT:    [[DOTSPLATINSERT1:%.*]] = insertelement <4 x i8> poison, i8 [[TMP3]], i64 0
; CHECK-NEXT:    [[DOTSPLAT2:%.*]] = shufflevector <4 x i8> [[DOTSPLATINSERT1]], <4 x i8> poison, <4 x i32> zeroinitializer
; CHECK-NEXT:    [[IND_END:%.*]] = mul i8 84, [[INDUCTION_IV]]
; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
; CHECK:       vector.body:
; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <4 x i8> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
; CHECK-NEXT:    [[TMP4:%.*]] = add i64 [[OFFSET_IDX]], 0
; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[DST:%.*]], i64 [[TMP4]]
; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[TMP5]], i32 0
; CHECK-NEXT:    store <4 x i8> [[VEC_IND]], ptr [[TMP6]], align 1
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i8> [[VEC_IND]], [[DOTSPLAT2]]
; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 84
; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
; CHECK:       middle.block:
; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 84, 84
; CHECK-NEXT:    br i1 [[CMP_N]], label [[OUTER_LATCH]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
; CHECK:       vec.epilog.iter.check:
; CHECK-NEXT:    [[IND_END5:%.*]] = mul i8 84, [[INDUCTION_IV]]
; CHECK-NEXT:    br i1 true, label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]]
; CHECK:       vec.epilog.ph:
; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
; CHECK-NEXT:    [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ 84, [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
; CHECK-NEXT:    [[IND_END4:%.*]] = mul i8 84, [[INDUCTION_IV]]
; CHECK-NEXT:    [[DOTSPLATINSERT9:%.*]] = insertelement <4 x i8> poison, i8 [[BC_RESUME_VAL]], i64 0
; CHECK-NEXT:    [[DOTSPLAT10:%.*]] = shufflevector <4 x i8> [[DOTSPLATINSERT9]], <4 x i8> poison, <4 x i32> zeroinitializer
; CHECK-NEXT:    [[DOTSPLATINSERT11:%.*]] = insertelement <4 x i8> poison, i8 [[INDUCTION_IV]], i64 0
; CHECK-NEXT:    [[DOTSPLAT12:%.*]] = shufflevector <4 x i8> [[DOTSPLATINSERT11]], <4 x i8> poison, <4 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP8:%.*]] = mul <4 x i8> <i8 0, i8 1, i8 2, i8 3>, [[DOTSPLAT12]]
; CHECK-NEXT:    [[INDUCTION13:%.*]] = add <4 x i8> [[DOTSPLAT10]], [[TMP8]]
; CHECK-NEXT:    [[TMP9:%.*]] = mul i8 [[INDUCTION_IV]], 4
; CHECK-NEXT:    [[DOTSPLATINSERT14:%.*]] = insertelement <4 x i8> poison, i8 [[TMP9]], i64 0
; CHECK-NEXT:    [[DOTSPLAT15:%.*]] = shufflevector <4 x i8> [[DOTSPLATINSERT14]], <4 x i8> poison, <4 x i32> zeroinitializer
; CHECK-NEXT:    br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
; CHECK:       vec.epilog.vector.body:
; CHECK-NEXT:    [[INDEX8:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT19:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
; CHECK-NEXT:    [[VEC_IND16:%.*]] = phi <4 x i8> [ [[INDUCTION13]], [[VEC_EPILOG_PH]] ], [ [[VEC_IND_NEXT17:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
; CHECK-NEXT:    [[OFFSET_IDX18:%.*]] = add i64 1, [[INDEX8]]
; CHECK-NEXT:    [[TMP10:%.*]] = add i64 [[OFFSET_IDX18]], 0
; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[TMP10]]
; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 0
; CHECK-NEXT:    store <4 x i8> [[VEC_IND16]], ptr [[TMP12]], align 1
; CHECK-NEXT:    [[INDEX_NEXT19]] = add nuw i64 [[INDEX8]], 4
; CHECK-NEXT:    [[VEC_IND_NEXT17]] = add <4 x i8> [[VEC_IND16]], [[DOTSPLAT15]]
; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT19]], 84
; CHECK-NEXT:    br i1 [[TMP13]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
; CHECK:       vec.epilog.middle.block:
; CHECK-NEXT:    [[CMP_N7:%.*]] = icmp eq i64 84, 84
; CHECK-NEXT:    br i1 [[CMP_N7]], label [[OUTER_LATCH]], label [[VEC_EPILOG_SCALAR_PH]]
; CHECK:       vec.epilog.scalar.ph:
; CHECK-NEXT:    [[BC_RESUME_VAL3:%.*]] = phi i64 [ 85, [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ 85, [[VEC_EPILOG_ITER_CHECK]] ], [ 1, [[ITER_CHECK]] ]
; CHECK-NEXT:    [[BC_RESUME_VAL6:%.*]] = phi i8 [ [[IND_END4]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[IND_END5]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK]] ]
; CHECK-NEXT:    br label [[INNER:%.*]]
; CHECK:       inner:
; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL3]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[INNER]] ]
; CHECK-NEXT:    [[IV_2:%.*]] = phi i8 [ [[BC_RESUME_VAL6]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[IV_2_NEXT:%.*]], [[INNER]] ]
; CHECK-NEXT:    [[IV_2_NEXT]] = sub i8 [[IV_2]], [[TRUNC_ADD]]
; CHECK-NEXT:    [[GEP_DST:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[IV]]
; CHECK-NEXT:    store i8 [[IV_2]], ptr [[GEP_DST]], align 1
; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT:    [[EC:%.*]] = icmp ugt i64 [[IV]], 83
; CHECK-NEXT:    br i1 [[EC]], label [[OUTER_LATCH]], label [[INNER]], !llvm.loop [[LOOP13:![0-9]+]]
; CHECK:       outer.latch:
; CHECK-NEXT:    [[OUTER_IV_NEXT]] = add nuw nsw i64 [[OUTER_IV]], 1
; CHECK-NEXT:    [[INDVAR_NEXT]] = add i8 [[INDVAR]], 1
; CHECK-NEXT:    [[INDUCTION_IV_NEXT]] = add i8 [[INDUCTION_IV]], [[TMP1]]
; CHECK-NEXT:    br label [[ITER_CHECK]]
;
; CHECK-PROFITABLE-BY-DEFAULT-LABEL: @induction_resume_value_requires_non_trivial_scev_expansion(
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:  entry:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br label [[ITER_CHECK:%.*]]
; CHECK-PROFITABLE-BY-DEFAULT:       iter.check:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDUCTION_IV:%.*]] = phi i8 [ [[INDUCTION_IV_NEXT:%.*]], [[OUTER_LATCH:%.*]] ], [ -56, [[ENTRY:%.*]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDVAR:%.*]] = phi i8 [ [[INDVAR_NEXT:%.*]], [[OUTER_LATCH]] ], [ 0, [[ENTRY]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[OUTER_IV:%.*]] = phi i64 [ 2, [[ENTRY]] ], [ [[OUTER_IV_NEXT:%.*]], [[OUTER_LATCH]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[P2:%.*]] = phi i32 [ -202, [[ENTRY]] ], [ [[ADD:%.*]], [[OUTER_LATCH]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP0:%.*]] = mul i8 [[INDVAR]], -1
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP1:%.*]] = add i8 [[TMP0]], -3
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TRUNC_IV:%.*]] = trunc i64 [[OUTER_IV]] to i32
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[ADD]] = add i32 [[P2]], [[TRUNC_IV]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TRUNC_ADD:%.*]] = trunc i32 [[ADD]] to i8
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 false, label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]]
; CHECK-PROFITABLE-BY-DEFAULT:       vector.main.loop.iter.check:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 false, label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK-PROFITABLE-BY-DEFAULT:       vector.ph:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <4 x i8> poison, i8 [[INDUCTION_IV]], i64 0
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <4 x i8> [[DOTSPLATINSERT]], <4 x i8> poison, <4 x i32> zeroinitializer
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP2:%.*]] = mul <4 x i8> <i8 0, i8 1, i8 2, i8 3>, [[DOTSPLAT]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDUCTION:%.*]] = add <4 x i8> zeroinitializer, [[TMP2]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP3:%.*]] = mul i8 [[INDUCTION_IV]], 4
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[DOTSPLATINSERT1:%.*]] = insertelement <4 x i8> poison, i8 [[TMP3]], i64 0
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[DOTSPLAT2:%.*]] = shufflevector <4 x i8> [[DOTSPLATINSERT1]], <4 x i8> poison, <4 x i32> zeroinitializer
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[IND_END:%.*]] = mul i8 84, [[INDUCTION_IV]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br label [[VECTOR_BODY:%.*]]
; CHECK-PROFITABLE-BY-DEFAULT:       vector.body:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[VEC_IND:%.*]] = phi <4 x i8> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP4:%.*]] = add i64 [[OFFSET_IDX]], 0
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[DST:%.*]], i64 [[TMP4]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[TMP5]], i32 0
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    store <4 x i8> [[VEC_IND]], ptr [[TMP6]], align 1
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[VEC_IND_NEXT]] = add <4 x i8> [[VEC_IND]], [[DOTSPLAT2]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 84
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
; CHECK-PROFITABLE-BY-DEFAULT:       middle.block:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[CMP_N:%.*]] = icmp eq i64 84, 84
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 [[CMP_N]], label [[OUTER_LATCH]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
; CHECK-PROFITABLE-BY-DEFAULT:       vec.epilog.iter.check:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[IND_END5:%.*]] = mul i8 84, [[INDUCTION_IV]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 true, label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]]
; CHECK-PROFITABLE-BY-DEFAULT:       vec.epilog.ph:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ [[IND_END]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ 84, [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[IND_END4:%.*]] = mul i8 84, [[INDUCTION_IV]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[DOTSPLATINSERT9:%.*]] = insertelement <2 x i8> poison, i8 [[BC_RESUME_VAL]], i64 0
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[DOTSPLAT10:%.*]] = shufflevector <2 x i8> [[DOTSPLATINSERT9]], <2 x i8> poison, <2 x i32> zeroinitializer
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[DOTSPLATINSERT11:%.*]] = insertelement <2 x i8> poison, i8 [[INDUCTION_IV]], i64 0
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[DOTSPLAT12:%.*]] = shufflevector <2 x i8> [[DOTSPLATINSERT11]], <2 x i8> poison, <2 x i32> zeroinitializer
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP8:%.*]] = mul <2 x i8> <i8 0, i8 1>, [[DOTSPLAT12]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDUCTION13:%.*]] = add <2 x i8> [[DOTSPLAT10]], [[TMP8]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP9:%.*]] = mul i8 [[INDUCTION_IV]], 2
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[DOTSPLATINSERT14:%.*]] = insertelement <2 x i8> poison, i8 [[TMP9]], i64 0
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[DOTSPLAT15:%.*]] = shufflevector <2 x i8> [[DOTSPLATINSERT14]], <2 x i8> poison, <2 x i32> zeroinitializer
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
; CHECK-PROFITABLE-BY-DEFAULT:       vec.epilog.vector.body:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDEX8:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT19:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[VEC_IND16:%.*]] = phi <2 x i8> [ [[INDUCTION13]], [[VEC_EPILOG_PH]] ], [ [[VEC_IND_NEXT17:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[OFFSET_IDX18:%.*]] = add i64 1, [[INDEX8]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP10:%.*]] = add i64 [[OFFSET_IDX18]], 0
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[TMP10]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 0
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    store <2 x i8> [[VEC_IND16]], ptr [[TMP12]], align 1
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDEX_NEXT19]] = add nuw i64 [[INDEX8]], 2
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[VEC_IND_NEXT17]] = add <2 x i8> [[VEC_IND16]], [[DOTSPLAT15]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT19]], 84
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 [[TMP13]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
; CHECK-PROFITABLE-BY-DEFAULT:       vec.epilog.middle.block:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[CMP_N7:%.*]] = icmp eq i64 84, 84
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 [[CMP_N7]], label [[OUTER_LATCH]], label [[VEC_EPILOG_SCALAR_PH]]
; CHECK-PROFITABLE-BY-DEFAULT:       vec.epilog.scalar.ph:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[BC_RESUME_VAL3:%.*]] = phi i64 [ 85, [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ 85, [[VEC_EPILOG_ITER_CHECK]] ], [ 1, [[ITER_CHECK]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[BC_RESUME_VAL6:%.*]] = phi i8 [ [[IND_END4]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[IND_END5]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br label [[INNER:%.*]]
; CHECK-PROFITABLE-BY-DEFAULT:       inner:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL3]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[INNER]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[IV_2:%.*]] = phi i8 [ [[BC_RESUME_VAL6]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[IV_2_NEXT:%.*]], [[INNER]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[IV_2_NEXT]] = sub i8 [[IV_2]], [[TRUNC_ADD]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[GEP_DST:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[IV]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    store i8 [[IV_2]], ptr [[GEP_DST]], align 1
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[EC:%.*]] = icmp ugt i64 [[IV]], 83
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 [[EC]], label [[OUTER_LATCH]], label [[INNER]], !llvm.loop [[LOOP7:![0-9]+]]
; CHECK-PROFITABLE-BY-DEFAULT:       outer.latch:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[OUTER_IV_NEXT]] = add nuw nsw i64 [[OUTER_IV]], 1
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDVAR_NEXT]] = add i8 [[INDVAR]], 1
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDUCTION_IV_NEXT]] = add i8 [[INDUCTION_IV]], [[TMP1]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br label [[ITER_CHECK]]
;
entry:
  br label %outer.header

outer.header:
  %outer.iv = phi i64 [ 2, %entry ], [ %outer.iv.next, %outer.latch ]
  %p2 = phi i32 [ -202, %entry ], [ %add, %outer.latch ]
  %trunc.iv  = trunc i64 %outer.iv to i32
  %add = add i32 %p2, %trunc.iv
  %trunc.add = trunc i32 %add to i8
  br label %inner

inner:
  %iv = phi i64 [ 1, %outer.header ], [ %iv.next, %inner ]
  %iv.2 = phi i8 [ 0, %outer.header ], [ %iv.2.next, %inner ]
  %iv.2.next = sub i8 %iv.2, %trunc.add
  %gep.dst = getelementptr inbounds i8, ptr %dst, i64 %iv
  store i8 %iv.2, ptr %gep.dst
  %iv.next = add nuw nsw i64 %iv, 1
  %ec = icmp ugt i64 %iv, 83
  br i1 %ec, label %outer.latch, label %inner

outer.latch:
  %outer.iv.next = add nuw nsw i64 %outer.iv, 1
  br label %outer.header
}

; Check handling of widended/truncated inductions.
define void @f4(ptr noalias %A, i32 signext %n) {
; CHECK-LABEL: @f4(
; CHECK-NEXT:  iter.check:
; CHECK-NEXT:    [[WIDE_TRIP_COUNT:%.*]] = zext i32 [[N:%.*]] to i64
; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 4
; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]]
; CHECK:       vector.main.loop.iter.check:
; CHECK-NEXT:    [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 4
; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK:       vector.ph:
; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 4
; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF]]
; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
; CHECK:       vector.body:
; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
; CHECK-NEXT:    [[TMP1:%.*]] = trunc <4 x i32> [[VEC_IND]] to <4 x i8>
; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[TMP0]]
; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i32 0
; CHECK-NEXT:    store <4 x i8> [[TMP1]], ptr [[TMP3]], align 1
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 4, i32 4, i32 4, i32 4>
; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
; CHECK:       middle.block:
; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC]]
; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
; CHECK:       vec.epilog.iter.check:
; CHECK-NEXT:    [[N_VEC_REMAINING:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_VEC]]
; CHECK-NEXT:    [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 4
; CHECK-NEXT:    br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]]
; CHECK:       vec.epilog.ph:
; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
; CHECK-NEXT:    [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
; CHECK-NEXT:    [[N_MOD_VF2:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 4
; CHECK-NEXT:    [[N_VEC3:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF2]]
; CHECK-NEXT:    [[TMP5:%.*]] = trunc i64 [[BC_RESUME_VAL]] to i32
; CHECK-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[TMP5]], i64 0
; CHECK-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <4 x i32> [[DOTSPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
; CHECK-NEXT:    [[INDUCTION:%.*]] = add <4 x i32> [[DOTSPLAT]], <i32 0, i32 1, i32 2, i32 3>
; CHECK-NEXT:    br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
; CHECK:       vec.epilog.vector.body:
; CHECK-NEXT:    [[INDEX6:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT9:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
; CHECK-NEXT:    [[VEC_IND7:%.*]] = phi <4 x i32> [ [[INDUCTION]], [[VEC_EPILOG_PH]] ], [ [[VEC_IND_NEXT8:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
; CHECK-NEXT:    [[TMP6:%.*]] = add i64 [[INDEX6]], 0
; CHECK-NEXT:    [[TMP7:%.*]] = trunc <4 x i32> [[VEC_IND7]] to <4 x i8>
; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[TMP6]]
; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[TMP8]], i32 0
; CHECK-NEXT:    store <4 x i8> [[TMP7]], ptr [[TMP9]], align 1
; CHECK-NEXT:    [[INDEX_NEXT9]] = add nuw i64 [[INDEX6]], 4
; CHECK-NEXT:    [[VEC_IND_NEXT8]] = add <4 x i32> [[VEC_IND7]], <i32 4, i32 4, i32 4, i32 4>
; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT9]], [[N_VEC3]]
; CHECK-NEXT:    br i1 [[TMP10]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
; CHECK:       vec.epilog.middle.block:
; CHECK-NEXT:    [[CMP_N5:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC3]]
; CHECK-NEXT:    br i1 [[CMP_N5]], label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]]
; CHECK:       vec.epilog.scalar.ph:
; CHECK-NEXT:    [[BC_RESUME_VAL4:%.*]] = phi i64 [ [[N_VEC3]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK:%.*]] ]
; CHECK-NEXT:    br label [[LOOP:%.*]]
; CHECK:       loop:
; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL4]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
; CHECK-NEXT:    [[TMP11:%.*]] = trunc i64 [[IV]] to i32
; CHECK-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP11]] to i8
; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[IV]]
; CHECK-NEXT:    store i8 [[CONV]], ptr [[ARRAYIDX]], align 1
; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp ne i64 [[IV_NEXT]], [[WIDE_TRIP_COUNT]]
; CHECK-NEXT:    br i1 [[EXITCOND]], label [[LOOP]], label [[EXIT]], !llvm.loop [[LOOP16:![0-9]+]]
; CHECK:       exit:
; CHECK-NEXT:    ret void
;
; CHECK-PROFITABLE-BY-DEFAULT-LABEL: @f4(
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:  iter.check:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[WIDE_TRIP_COUNT:%.*]] = zext i32 [[N:%.*]] to i64
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 2
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]]
; CHECK-PROFITABLE-BY-DEFAULT:       vector.main.loop.iter.check:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 4
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK-PROFITABLE-BY-DEFAULT:       vector.ph:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 4
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[N_VEC:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br label [[VECTOR_BODY:%.*]]
; CHECK-PROFITABLE-BY-DEFAULT:       vector.body:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP1:%.*]] = trunc <4 x i32> [[VEC_IND]] to <4 x i8>
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[TMP0]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i32 0
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    store <4 x i8> [[TMP1]], ptr [[TMP3]], align 1
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 4, i32 4, i32 4, i32 4>
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
; CHECK-PROFITABLE-BY-DEFAULT:       middle.block:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
; CHECK-PROFITABLE-BY-DEFAULT:       vec.epilog.iter.check:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[N_VEC_REMAINING:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_VEC]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 2
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]]
; CHECK-PROFITABLE-BY-DEFAULT:       vec.epilog.ph:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[N_MOD_VF2:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 2
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[N_VEC3:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF2]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP5:%.*]] = trunc i64 [[BC_RESUME_VAL]] to i32
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[TMP5]], i64 0
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDUCTION:%.*]] = add <2 x i32> [[DOTSPLAT]], <i32 0, i32 1>
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
; CHECK-PROFITABLE-BY-DEFAULT:       vec.epilog.vector.body:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDEX6:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT9:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[VEC_IND7:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VEC_EPILOG_PH]] ], [ [[VEC_IND_NEXT8:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP6:%.*]] = add i64 [[INDEX6]], 0
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP7:%.*]] = trunc <2 x i32> [[VEC_IND7]] to <2 x i8>
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[TMP6]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[TMP8]], i32 0
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    store <2 x i8> [[TMP7]], ptr [[TMP9]], align 1
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDEX_NEXT9]] = add nuw i64 [[INDEX6]], 2
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[VEC_IND_NEXT8]] = add <2 x i32> [[VEC_IND7]], <i32 2, i32 2>
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT9]], [[N_VEC3]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 [[TMP10]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
; CHECK-PROFITABLE-BY-DEFAULT:       vec.epilog.middle.block:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[CMP_N5:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC3]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 [[CMP_N5]], label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]]
; CHECK-PROFITABLE-BY-DEFAULT:       vec.epilog.scalar.ph:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[BC_RESUME_VAL4:%.*]] = phi i64 [ [[N_VEC3]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK:%.*]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br label [[LOOP:%.*]]
; CHECK-PROFITABLE-BY-DEFAULT:       loop:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL4]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP11:%.*]] = trunc i64 [[IV]] to i32
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP11]] to i8
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[IV]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    store i8 [[CONV]], ptr [[ARRAYIDX]], align 1
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[EXITCOND:%.*]] = icmp ne i64 [[IV_NEXT]], [[WIDE_TRIP_COUNT]]
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 [[EXITCOND]], label [[LOOP]], label [[EXIT]], !llvm.loop [[LOOP10:![0-9]+]]
; CHECK-PROFITABLE-BY-DEFAULT:       exit:
; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    ret void
;
entry:
  %wide.trip.count = zext i32 %n to i64
  br label %loop

loop:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
  %0 = trunc i64 %iv to i32
  %conv = trunc i32 %0 to i8
  %arrayidx = getelementptr inbounds i8, ptr %A, i64 %iv
  store i8 %conv, ptr %arrayidx, align 1
  %iv.next = add nuw nsw i64 %iv, 1
  %exitcond = icmp ne i64 %iv.next, %wide.trip.count
  br i1 %exitcond, label %loop, label %exit

exit:
  ret void
}