1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py 2; RUN: opt < %s -passes='loop-vectorize' -enable-epilogue-vectorization -epilogue-vectorization-force-VF=2 -S | FileCheck %s --check-prefix VF-TWO-CHECK 3; RUN: opt < %s -passes='loop-vectorize' -enable-epilogue-vectorization -epilogue-vectorization-force-VF=4 -S | FileCheck %s --check-prefix VF-FOUR-CHECK 4 5target datalayout = "e-m:e-i64:64-n32:64" 6target triple = "powerpc64le-unknown-linux-gnu" 7 8; Function Attrs: nounwind 9define dso_local void @f1(float* noalias %aa, float* noalias %bb, float* noalias %cc, i32 signext %N) #0 { 10; VF-TWO-CHECK-LABEL: @f1( 11; VF-TWO-CHECK-NEXT: entry: 12; VF-TWO-CHECK-NEXT: [[CMP1:%.*]] = icmp sgt i32 [[N:%.*]], 0 13; VF-TWO-CHECK-NEXT: br i1 [[CMP1]], label [[ITER_CHECK:%.*]], label [[FOR_END:%.*]] 14; VF-TWO-CHECK: iter.check: 15; VF-TWO-CHECK-NEXT: [[WIDE_TRIP_COUNT:%.*]] = zext i32 [[N]] to i64 16; VF-TWO-CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 2 17; VF-TWO-CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]] 18; VF-TWO-CHECK: vector.main.loop.iter.check: 19; VF-TWO-CHECK-NEXT: [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 48 20; VF-TWO-CHECK-NEXT: br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]] 21; VF-TWO-CHECK: vector.ph: 22; VF-TWO-CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 48 23; VF-TWO-CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF]] 24; VF-TWO-CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 25; VF-TWO-CHECK: vector.body: 26; VF-TWO-CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 27; VF-TWO-CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0 28; VF-TWO-CHECK-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 4 29; VF-TWO-CHECK-NEXT: [[TMP2:%.*]] = add i64 [[INDEX]], 8 30; VF-TWO-CHECK-NEXT: [[TMP3:%.*]] = add i64 [[INDEX]], 12 31; VF-TWO-CHECK-NEXT: [[TMP4:%.*]] = add i64 [[INDEX]], 16 32; VF-TWO-CHECK-NEXT: [[TMP5:%.*]] = add i64 [[INDEX]], 20 33; VF-TWO-CHECK-NEXT: [[TMP6:%.*]] = add i64 [[INDEX]], 24 34; VF-TWO-CHECK-NEXT: [[TMP7:%.*]] = add i64 [[INDEX]], 28 35; VF-TWO-CHECK-NEXT: [[TMP8:%.*]] = add i64 [[INDEX]], 32 36; VF-TWO-CHECK-NEXT: [[TMP9:%.*]] = add i64 [[INDEX]], 36 37; VF-TWO-CHECK-NEXT: [[TMP10:%.*]] = add i64 [[INDEX]], 40 38; VF-TWO-CHECK-NEXT: [[TMP11:%.*]] = add i64 [[INDEX]], 44 39; VF-TWO-CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds float, float* [[BB:%.*]], i64 [[TMP0]] 40; VF-TWO-CHECK-NEXT: [[TMP13:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP1]] 41; VF-TWO-CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP2]] 42; VF-TWO-CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP3]] 43; VF-TWO-CHECK-NEXT: [[TMP16:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP4]] 44; VF-TWO-CHECK-NEXT: [[TMP17:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP5]] 45; VF-TWO-CHECK-NEXT: [[TMP18:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP6]] 46; VF-TWO-CHECK-NEXT: [[TMP19:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP7]] 47; VF-TWO-CHECK-NEXT: [[TMP20:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP8]] 48; VF-TWO-CHECK-NEXT: [[TMP21:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP9]] 49; VF-TWO-CHECK-NEXT: [[TMP22:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP10]] 50; VF-TWO-CHECK-NEXT: [[TMP23:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP11]] 51; VF-TWO-CHECK-NEXT: [[TMP24:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 0 52; VF-TWO-CHECK-NEXT: [[TMP25:%.*]] = bitcast float* [[TMP24]] to <4 x float>* 53; VF-TWO-CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, <4 x float>* [[TMP25]], align 4 54; VF-TWO-CHECK-NEXT: [[TMP26:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 4 55; VF-TWO-CHECK-NEXT: [[TMP27:%.*]] = bitcast float* [[TMP26]] to <4 x float>* 56; VF-TWO-CHECK-NEXT: [[WIDE_LOAD2:%.*]] = load <4 x float>, <4 x float>* [[TMP27]], align 4 57; VF-TWO-CHECK-NEXT: [[TMP28:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 8 58; VF-TWO-CHECK-NEXT: [[TMP29:%.*]] = bitcast float* [[TMP28]] to <4 x float>* 59; VF-TWO-CHECK-NEXT: [[WIDE_LOAD3:%.*]] = load <4 x float>, <4 x float>* [[TMP29]], align 4 60; VF-TWO-CHECK-NEXT: [[TMP30:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 12 61; VF-TWO-CHECK-NEXT: [[TMP31:%.*]] = bitcast float* [[TMP30]] to <4 x float>* 62; VF-TWO-CHECK-NEXT: [[WIDE_LOAD4:%.*]] = load <4 x float>, <4 x float>* [[TMP31]], align 4 63; VF-TWO-CHECK-NEXT: [[TMP32:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 16 64; VF-TWO-CHECK-NEXT: [[TMP33:%.*]] = bitcast float* [[TMP32]] to <4 x float>* 65; VF-TWO-CHECK-NEXT: [[WIDE_LOAD5:%.*]] = load <4 x float>, <4 x float>* [[TMP33]], align 4 66; VF-TWO-CHECK-NEXT: [[TMP34:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 20 67; VF-TWO-CHECK-NEXT: [[TMP35:%.*]] = bitcast float* [[TMP34]] to <4 x float>* 68; VF-TWO-CHECK-NEXT: [[WIDE_LOAD6:%.*]] = load <4 x float>, <4 x float>* [[TMP35]], align 4 69; VF-TWO-CHECK-NEXT: [[TMP36:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 24 70; VF-TWO-CHECK-NEXT: [[TMP37:%.*]] = bitcast float* [[TMP36]] to <4 x float>* 71; VF-TWO-CHECK-NEXT: [[WIDE_LOAD7:%.*]] = load <4 x float>, <4 x float>* [[TMP37]], align 4 72; VF-TWO-CHECK-NEXT: [[TMP38:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 28 73; VF-TWO-CHECK-NEXT: [[TMP39:%.*]] = bitcast float* [[TMP38]] to <4 x float>* 74; VF-TWO-CHECK-NEXT: [[WIDE_LOAD8:%.*]] = load <4 x float>, <4 x float>* [[TMP39]], align 4 75; VF-TWO-CHECK-NEXT: [[TMP40:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 32 76; VF-TWO-CHECK-NEXT: [[TMP41:%.*]] = bitcast float* [[TMP40]] to <4 x float>* 77; VF-TWO-CHECK-NEXT: [[WIDE_LOAD9:%.*]] = load <4 x float>, <4 x float>* [[TMP41]], align 4 78; VF-TWO-CHECK-NEXT: [[TMP42:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 36 79; VF-TWO-CHECK-NEXT: [[TMP43:%.*]] = bitcast float* [[TMP42]] to <4 x float>* 80; VF-TWO-CHECK-NEXT: [[WIDE_LOAD10:%.*]] = load <4 x float>, <4 x float>* [[TMP43]], align 4 81; VF-TWO-CHECK-NEXT: [[TMP44:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 40 82; VF-TWO-CHECK-NEXT: [[TMP45:%.*]] = bitcast float* [[TMP44]] to <4 x float>* 83; VF-TWO-CHECK-NEXT: [[WIDE_LOAD11:%.*]] = load <4 x float>, <4 x float>* [[TMP45]], align 4 84; VF-TWO-CHECK-NEXT: [[TMP46:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 44 85; VF-TWO-CHECK-NEXT: [[TMP47:%.*]] = bitcast float* [[TMP46]] to <4 x float>* 86; VF-TWO-CHECK-NEXT: [[WIDE_LOAD12:%.*]] = load <4 x float>, <4 x float>* [[TMP47]], align 4 87; VF-TWO-CHECK-NEXT: [[TMP48:%.*]] = getelementptr inbounds float, float* [[CC:%.*]], i64 [[TMP0]] 88; VF-TWO-CHECK-NEXT: [[TMP49:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP1]] 89; VF-TWO-CHECK-NEXT: [[TMP50:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP2]] 90; VF-TWO-CHECK-NEXT: [[TMP51:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP3]] 91; VF-TWO-CHECK-NEXT: [[TMP52:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP4]] 92; VF-TWO-CHECK-NEXT: [[TMP53:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP5]] 93; VF-TWO-CHECK-NEXT: [[TMP54:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP6]] 94; VF-TWO-CHECK-NEXT: [[TMP55:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP7]] 95; VF-TWO-CHECK-NEXT: [[TMP56:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP8]] 96; VF-TWO-CHECK-NEXT: [[TMP57:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP9]] 97; VF-TWO-CHECK-NEXT: [[TMP58:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP10]] 98; VF-TWO-CHECK-NEXT: [[TMP59:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP11]] 99; VF-TWO-CHECK-NEXT: [[TMP60:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 0 100; VF-TWO-CHECK-NEXT: [[TMP61:%.*]] = bitcast float* [[TMP60]] to <4 x float>* 101; VF-TWO-CHECK-NEXT: [[WIDE_LOAD13:%.*]] = load <4 x float>, <4 x float>* [[TMP61]], align 4 102; VF-TWO-CHECK-NEXT: [[TMP62:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 4 103; VF-TWO-CHECK-NEXT: [[TMP63:%.*]] = bitcast float* [[TMP62]] to <4 x float>* 104; VF-TWO-CHECK-NEXT: [[WIDE_LOAD14:%.*]] = load <4 x float>, <4 x float>* [[TMP63]], align 4 105; VF-TWO-CHECK-NEXT: [[TMP64:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 8 106; VF-TWO-CHECK-NEXT: [[TMP65:%.*]] = bitcast float* [[TMP64]] to <4 x float>* 107; VF-TWO-CHECK-NEXT: [[WIDE_LOAD15:%.*]] = load <4 x float>, <4 x float>* [[TMP65]], align 4 108; VF-TWO-CHECK-NEXT: [[TMP66:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 12 109; VF-TWO-CHECK-NEXT: [[TMP67:%.*]] = bitcast float* [[TMP66]] to <4 x float>* 110; VF-TWO-CHECK-NEXT: [[WIDE_LOAD16:%.*]] = load <4 x float>, <4 x float>* [[TMP67]], align 4 111; VF-TWO-CHECK-NEXT: [[TMP68:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 16 112; VF-TWO-CHECK-NEXT: [[TMP69:%.*]] = bitcast float* [[TMP68]] to <4 x float>* 113; VF-TWO-CHECK-NEXT: [[WIDE_LOAD17:%.*]] = load <4 x float>, <4 x float>* [[TMP69]], align 4 114; VF-TWO-CHECK-NEXT: [[TMP70:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 20 115; VF-TWO-CHECK-NEXT: [[TMP71:%.*]] = bitcast float* [[TMP70]] to <4 x float>* 116; VF-TWO-CHECK-NEXT: [[WIDE_LOAD18:%.*]] = load <4 x float>, <4 x float>* [[TMP71]], align 4 117; VF-TWO-CHECK-NEXT: [[TMP72:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 24 118; VF-TWO-CHECK-NEXT: [[TMP73:%.*]] = bitcast float* [[TMP72]] to <4 x float>* 119; VF-TWO-CHECK-NEXT: [[WIDE_LOAD19:%.*]] = load <4 x float>, <4 x float>* [[TMP73]], align 4 120; VF-TWO-CHECK-NEXT: [[TMP74:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 28 121; VF-TWO-CHECK-NEXT: [[TMP75:%.*]] = bitcast float* [[TMP74]] to <4 x float>* 122; VF-TWO-CHECK-NEXT: [[WIDE_LOAD20:%.*]] = load <4 x float>, <4 x float>* [[TMP75]], align 4 123; VF-TWO-CHECK-NEXT: [[TMP76:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 32 124; VF-TWO-CHECK-NEXT: [[TMP77:%.*]] = bitcast float* [[TMP76]] to <4 x float>* 125; VF-TWO-CHECK-NEXT: [[WIDE_LOAD21:%.*]] = load <4 x float>, <4 x float>* [[TMP77]], align 4 126; VF-TWO-CHECK-NEXT: [[TMP78:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 36 127; VF-TWO-CHECK-NEXT: [[TMP79:%.*]] = bitcast float* [[TMP78]] to <4 x float>* 128; VF-TWO-CHECK-NEXT: [[WIDE_LOAD22:%.*]] = load <4 x float>, <4 x float>* [[TMP79]], align 4 129; VF-TWO-CHECK-NEXT: [[TMP80:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 40 130; VF-TWO-CHECK-NEXT: [[TMP81:%.*]] = bitcast float* [[TMP80]] to <4 x float>* 131; VF-TWO-CHECK-NEXT: [[WIDE_LOAD23:%.*]] = load <4 x float>, <4 x float>* [[TMP81]], align 4 132; VF-TWO-CHECK-NEXT: [[TMP82:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 44 133; VF-TWO-CHECK-NEXT: [[TMP83:%.*]] = bitcast float* [[TMP82]] to <4 x float>* 134; VF-TWO-CHECK-NEXT: [[WIDE_LOAD24:%.*]] = load <4 x float>, <4 x float>* [[TMP83]], align 4 135; VF-TWO-CHECK-NEXT: [[TMP84:%.*]] = fadd fast <4 x float> [[WIDE_LOAD]], [[WIDE_LOAD13]] 136; VF-TWO-CHECK-NEXT: [[TMP85:%.*]] = fadd fast <4 x float> [[WIDE_LOAD2]], [[WIDE_LOAD14]] 137; VF-TWO-CHECK-NEXT: [[TMP86:%.*]] = fadd fast <4 x float> [[WIDE_LOAD3]], [[WIDE_LOAD15]] 138; VF-TWO-CHECK-NEXT: [[TMP87:%.*]] = fadd fast <4 x float> [[WIDE_LOAD4]], [[WIDE_LOAD16]] 139; VF-TWO-CHECK-NEXT: [[TMP88:%.*]] = fadd fast <4 x float> [[WIDE_LOAD5]], [[WIDE_LOAD17]] 140; VF-TWO-CHECK-NEXT: [[TMP89:%.*]] = fadd fast <4 x float> [[WIDE_LOAD6]], [[WIDE_LOAD18]] 141; VF-TWO-CHECK-NEXT: [[TMP90:%.*]] = fadd fast <4 x float> [[WIDE_LOAD7]], [[WIDE_LOAD19]] 142; VF-TWO-CHECK-NEXT: [[TMP91:%.*]] = fadd fast <4 x float> [[WIDE_LOAD8]], [[WIDE_LOAD20]] 143; VF-TWO-CHECK-NEXT: [[TMP92:%.*]] = fadd fast <4 x float> [[WIDE_LOAD9]], [[WIDE_LOAD21]] 144; VF-TWO-CHECK-NEXT: [[TMP93:%.*]] = fadd fast <4 x float> [[WIDE_LOAD10]], [[WIDE_LOAD22]] 145; VF-TWO-CHECK-NEXT: [[TMP94:%.*]] = fadd fast <4 x float> [[WIDE_LOAD11]], [[WIDE_LOAD23]] 146; VF-TWO-CHECK-NEXT: [[TMP95:%.*]] = fadd fast <4 x float> [[WIDE_LOAD12]], [[WIDE_LOAD24]] 147; VF-TWO-CHECK-NEXT: [[TMP96:%.*]] = getelementptr inbounds float, float* [[AA:%.*]], i64 [[TMP0]] 148; VF-TWO-CHECK-NEXT: [[TMP97:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP1]] 149; VF-TWO-CHECK-NEXT: [[TMP98:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP2]] 150; VF-TWO-CHECK-NEXT: [[TMP99:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP3]] 151; VF-TWO-CHECK-NEXT: [[TMP100:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP4]] 152; VF-TWO-CHECK-NEXT: [[TMP101:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP5]] 153; VF-TWO-CHECK-NEXT: [[TMP102:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP6]] 154; VF-TWO-CHECK-NEXT: [[TMP103:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP7]] 155; VF-TWO-CHECK-NEXT: [[TMP104:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP8]] 156; VF-TWO-CHECK-NEXT: [[TMP105:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP9]] 157; VF-TWO-CHECK-NEXT: [[TMP106:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP10]] 158; VF-TWO-CHECK-NEXT: [[TMP107:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP11]] 159; VF-TWO-CHECK-NEXT: [[TMP108:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 0 160; VF-TWO-CHECK-NEXT: [[TMP109:%.*]] = bitcast float* [[TMP108]] to <4 x float>* 161; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP84]], <4 x float>* [[TMP109]], align 4 162; VF-TWO-CHECK-NEXT: [[TMP110:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 4 163; VF-TWO-CHECK-NEXT: [[TMP111:%.*]] = bitcast float* [[TMP110]] to <4 x float>* 164; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP85]], <4 x float>* [[TMP111]], align 4 165; VF-TWO-CHECK-NEXT: [[TMP112:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 8 166; VF-TWO-CHECK-NEXT: [[TMP113:%.*]] = bitcast float* [[TMP112]] to <4 x float>* 167; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP86]], <4 x float>* [[TMP113]], align 4 168; VF-TWO-CHECK-NEXT: [[TMP114:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 12 169; VF-TWO-CHECK-NEXT: [[TMP115:%.*]] = bitcast float* [[TMP114]] to <4 x float>* 170; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP87]], <4 x float>* [[TMP115]], align 4 171; VF-TWO-CHECK-NEXT: [[TMP116:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 16 172; VF-TWO-CHECK-NEXT: [[TMP117:%.*]] = bitcast float* [[TMP116]] to <4 x float>* 173; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP88]], <4 x float>* [[TMP117]], align 4 174; VF-TWO-CHECK-NEXT: [[TMP118:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 20 175; VF-TWO-CHECK-NEXT: [[TMP119:%.*]] = bitcast float* [[TMP118]] to <4 x float>* 176; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP89]], <4 x float>* [[TMP119]], align 4 177; VF-TWO-CHECK-NEXT: [[TMP120:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 24 178; VF-TWO-CHECK-NEXT: [[TMP121:%.*]] = bitcast float* [[TMP120]] to <4 x float>* 179; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP90]], <4 x float>* [[TMP121]], align 4 180; VF-TWO-CHECK-NEXT: [[TMP122:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 28 181; VF-TWO-CHECK-NEXT: [[TMP123:%.*]] = bitcast float* [[TMP122]] to <4 x float>* 182; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP91]], <4 x float>* [[TMP123]], align 4 183; VF-TWO-CHECK-NEXT: [[TMP124:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 32 184; VF-TWO-CHECK-NEXT: [[TMP125:%.*]] = bitcast float* [[TMP124]] to <4 x float>* 185; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP92]], <4 x float>* [[TMP125]], align 4 186; VF-TWO-CHECK-NEXT: [[TMP126:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 36 187; VF-TWO-CHECK-NEXT: [[TMP127:%.*]] = bitcast float* [[TMP126]] to <4 x float>* 188; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP93]], <4 x float>* [[TMP127]], align 4 189; VF-TWO-CHECK-NEXT: [[TMP128:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 40 190; VF-TWO-CHECK-NEXT: [[TMP129:%.*]] = bitcast float* [[TMP128]] to <4 x float>* 191; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP94]], <4 x float>* [[TMP129]], align 4 192; VF-TWO-CHECK-NEXT: [[TMP130:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 44 193; VF-TWO-CHECK-NEXT: [[TMP131:%.*]] = bitcast float* [[TMP130]] to <4 x float>* 194; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP95]], <4 x float>* [[TMP131]], align 4 195; VF-TWO-CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 48 196; VF-TWO-CHECK-NEXT: [[TMP132:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] 197; VF-TWO-CHECK-NEXT: br i1 [[TMP132]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] 198; VF-TWO-CHECK: middle.block: 199; VF-TWO-CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC]] 200; VF-TWO-CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]] 201; VF-TWO-CHECK: vec.epilog.iter.check: 202; VF-TWO-CHECK-NEXT: [[N_VEC_REMAINING:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_VEC]] 203; VF-TWO-CHECK-NEXT: [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 2 204; VF-TWO-CHECK-NEXT: br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]] 205; VF-TWO-CHECK: vec.epilog.ph: 206; VF-TWO-CHECK-NEXT: [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ] 207; VF-TWO-CHECK-NEXT: [[N_MOD_VF25:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 2 208; VF-TWO-CHECK-NEXT: [[N_VEC26:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF25]] 209; VF-TWO-CHECK-NEXT: br label [[VEC_EPILOG_VECTOR_BODY:%.*]] 210; VF-TWO-CHECK: vec.epilog.vector.body: 211; VF-TWO-CHECK-NEXT: [[INDEX27:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT28:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ] 212; VF-TWO-CHECK-NEXT: [[TMP133:%.*]] = add i64 [[INDEX27]], 0 213; VF-TWO-CHECK-NEXT: [[TMP134:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP133]] 214; VF-TWO-CHECK-NEXT: [[TMP135:%.*]] = getelementptr inbounds float, float* [[TMP134]], i32 0 215; VF-TWO-CHECK-NEXT: [[TMP136:%.*]] = bitcast float* [[TMP135]] to <2 x float>* 216; VF-TWO-CHECK-NEXT: [[WIDE_LOAD30:%.*]] = load <2 x float>, <2 x float>* [[TMP136]], align 4 217; VF-TWO-CHECK-NEXT: [[TMP137:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP133]] 218; VF-TWO-CHECK-NEXT: [[TMP138:%.*]] = getelementptr inbounds float, float* [[TMP137]], i32 0 219; VF-TWO-CHECK-NEXT: [[TMP139:%.*]] = bitcast float* [[TMP138]] to <2 x float>* 220; VF-TWO-CHECK-NEXT: [[WIDE_LOAD31:%.*]] = load <2 x float>, <2 x float>* [[TMP139]], align 4 221; VF-TWO-CHECK-NEXT: [[TMP140:%.*]] = fadd fast <2 x float> [[WIDE_LOAD30]], [[WIDE_LOAD31]] 222; VF-TWO-CHECK-NEXT: [[TMP141:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP133]] 223; VF-TWO-CHECK-NEXT: [[TMP142:%.*]] = getelementptr inbounds float, float* [[TMP141]], i32 0 224; VF-TWO-CHECK-NEXT: [[TMP143:%.*]] = bitcast float* [[TMP142]] to <2 x float>* 225; VF-TWO-CHECK-NEXT: store <2 x float> [[TMP140]], <2 x float>* [[TMP143]], align 4 226; VF-TWO-CHECK-NEXT: [[INDEX_NEXT28]] = add nuw i64 [[INDEX27]], 2 227; VF-TWO-CHECK-NEXT: [[TMP144:%.*]] = icmp eq i64 [[INDEX_NEXT28]], [[N_VEC26]] 228; VF-TWO-CHECK-NEXT: br i1 [[TMP144]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP2:![0-9]+]] 229; VF-TWO-CHECK: vec.epilog.middle.block: 230; VF-TWO-CHECK-NEXT: [[CMP_N29:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC26]] 231; VF-TWO-CHECK-NEXT: br i1 [[CMP_N29]], label [[FOR_END_LOOPEXIT_LOOPEXIT:%.*]], label [[VEC_EPILOG_SCALAR_PH]] 232; VF-TWO-CHECK: vec.epilog.scalar.ph: 233; VF-TWO-CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC26]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK]] ] 234; VF-TWO-CHECK-NEXT: br label [[FOR_BODY:%.*]] 235; VF-TWO-CHECK: for.body: 236; VF-TWO-CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ] 237; VF-TWO-CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[INDVARS_IV]] 238; VF-TWO-CHECK-NEXT: [[TMP145:%.*]] = load float, float* [[ARRAYIDX]], align 4 239; VF-TWO-CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[INDVARS_IV]] 240; VF-TWO-CHECK-NEXT: [[TMP146:%.*]] = load float, float* [[ARRAYIDX2]], align 4 241; VF-TWO-CHECK-NEXT: [[ADD:%.*]] = fadd fast float [[TMP145]], [[TMP146]] 242; VF-TWO-CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[INDVARS_IV]] 243; VF-TWO-CHECK-NEXT: store float [[ADD]], float* [[ARRAYIDX4]], align 4 244; VF-TWO-CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1 245; VF-TWO-CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i64 [[INDVARS_IV_NEXT]], [[WIDE_TRIP_COUNT]] 246; VF-TWO-CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_BODY]], label [[FOR_END_LOOPEXIT_LOOPEXIT]], !llvm.loop [[LOOP4:![0-9]+]] 247; VF-TWO-CHECK: for.end.loopexit.loopexit: 248; VF-TWO-CHECK-NEXT: br label [[FOR_END_LOOPEXIT]] 249; VF-TWO-CHECK: for.end.loopexit: 250; VF-TWO-CHECK-NEXT: br label [[FOR_END]] 251; VF-TWO-CHECK: for.end: 252; VF-TWO-CHECK-NEXT: ret void 253; 254; VF-FOUR-CHECK-LABEL: @f1( 255; VF-FOUR-CHECK-NEXT: entry: 256; VF-FOUR-CHECK-NEXT: [[CMP1:%.*]] = icmp sgt i32 [[N:%.*]], 0 257; VF-FOUR-CHECK-NEXT: br i1 [[CMP1]], label [[ITER_CHECK:%.*]], label [[FOR_END:%.*]] 258; VF-FOUR-CHECK: iter.check: 259; VF-FOUR-CHECK-NEXT: [[WIDE_TRIP_COUNT:%.*]] = zext i32 [[N]] to i64 260; VF-FOUR-CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 4 261; VF-FOUR-CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]] 262; VF-FOUR-CHECK: vector.main.loop.iter.check: 263; VF-FOUR-CHECK-NEXT: [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 48 264; VF-FOUR-CHECK-NEXT: br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]] 265; VF-FOUR-CHECK: vector.ph: 266; VF-FOUR-CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 48 267; VF-FOUR-CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF]] 268; VF-FOUR-CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 269; VF-FOUR-CHECK: vector.body: 270; VF-FOUR-CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 271; VF-FOUR-CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0 272; VF-FOUR-CHECK-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 4 273; VF-FOUR-CHECK-NEXT: [[TMP2:%.*]] = add i64 [[INDEX]], 8 274; VF-FOUR-CHECK-NEXT: [[TMP3:%.*]] = add i64 [[INDEX]], 12 275; VF-FOUR-CHECK-NEXT: [[TMP4:%.*]] = add i64 [[INDEX]], 16 276; VF-FOUR-CHECK-NEXT: [[TMP5:%.*]] = add i64 [[INDEX]], 20 277; VF-FOUR-CHECK-NEXT: [[TMP6:%.*]] = add i64 [[INDEX]], 24 278; VF-FOUR-CHECK-NEXT: [[TMP7:%.*]] = add i64 [[INDEX]], 28 279; VF-FOUR-CHECK-NEXT: [[TMP8:%.*]] = add i64 [[INDEX]], 32 280; VF-FOUR-CHECK-NEXT: [[TMP9:%.*]] = add i64 [[INDEX]], 36 281; VF-FOUR-CHECK-NEXT: [[TMP10:%.*]] = add i64 [[INDEX]], 40 282; VF-FOUR-CHECK-NEXT: [[TMP11:%.*]] = add i64 [[INDEX]], 44 283; VF-FOUR-CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds float, float* [[BB:%.*]], i64 [[TMP0]] 284; VF-FOUR-CHECK-NEXT: [[TMP13:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP1]] 285; VF-FOUR-CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP2]] 286; VF-FOUR-CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP3]] 287; VF-FOUR-CHECK-NEXT: [[TMP16:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP4]] 288; VF-FOUR-CHECK-NEXT: [[TMP17:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP5]] 289; VF-FOUR-CHECK-NEXT: [[TMP18:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP6]] 290; VF-FOUR-CHECK-NEXT: [[TMP19:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP7]] 291; VF-FOUR-CHECK-NEXT: [[TMP20:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP8]] 292; VF-FOUR-CHECK-NEXT: [[TMP21:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP9]] 293; VF-FOUR-CHECK-NEXT: [[TMP22:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP10]] 294; VF-FOUR-CHECK-NEXT: [[TMP23:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP11]] 295; VF-FOUR-CHECK-NEXT: [[TMP24:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 0 296; VF-FOUR-CHECK-NEXT: [[TMP25:%.*]] = bitcast float* [[TMP24]] to <4 x float>* 297; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, <4 x float>* [[TMP25]], align 4 298; VF-FOUR-CHECK-NEXT: [[TMP26:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 4 299; VF-FOUR-CHECK-NEXT: [[TMP27:%.*]] = bitcast float* [[TMP26]] to <4 x float>* 300; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD2:%.*]] = load <4 x float>, <4 x float>* [[TMP27]], align 4 301; VF-FOUR-CHECK-NEXT: [[TMP28:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 8 302; VF-FOUR-CHECK-NEXT: [[TMP29:%.*]] = bitcast float* [[TMP28]] to <4 x float>* 303; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD3:%.*]] = load <4 x float>, <4 x float>* [[TMP29]], align 4 304; VF-FOUR-CHECK-NEXT: [[TMP30:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 12 305; VF-FOUR-CHECK-NEXT: [[TMP31:%.*]] = bitcast float* [[TMP30]] to <4 x float>* 306; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD4:%.*]] = load <4 x float>, <4 x float>* [[TMP31]], align 4 307; VF-FOUR-CHECK-NEXT: [[TMP32:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 16 308; VF-FOUR-CHECK-NEXT: [[TMP33:%.*]] = bitcast float* [[TMP32]] to <4 x float>* 309; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD5:%.*]] = load <4 x float>, <4 x float>* [[TMP33]], align 4 310; VF-FOUR-CHECK-NEXT: [[TMP34:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 20 311; VF-FOUR-CHECK-NEXT: [[TMP35:%.*]] = bitcast float* [[TMP34]] to <4 x float>* 312; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD6:%.*]] = load <4 x float>, <4 x float>* [[TMP35]], align 4 313; VF-FOUR-CHECK-NEXT: [[TMP36:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 24 314; VF-FOUR-CHECK-NEXT: [[TMP37:%.*]] = bitcast float* [[TMP36]] to <4 x float>* 315; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD7:%.*]] = load <4 x float>, <4 x float>* [[TMP37]], align 4 316; VF-FOUR-CHECK-NEXT: [[TMP38:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 28 317; VF-FOUR-CHECK-NEXT: [[TMP39:%.*]] = bitcast float* [[TMP38]] to <4 x float>* 318; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD8:%.*]] = load <4 x float>, <4 x float>* [[TMP39]], align 4 319; VF-FOUR-CHECK-NEXT: [[TMP40:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 32 320; VF-FOUR-CHECK-NEXT: [[TMP41:%.*]] = bitcast float* [[TMP40]] to <4 x float>* 321; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD9:%.*]] = load <4 x float>, <4 x float>* [[TMP41]], align 4 322; VF-FOUR-CHECK-NEXT: [[TMP42:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 36 323; VF-FOUR-CHECK-NEXT: [[TMP43:%.*]] = bitcast float* [[TMP42]] to <4 x float>* 324; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD10:%.*]] = load <4 x float>, <4 x float>* [[TMP43]], align 4 325; VF-FOUR-CHECK-NEXT: [[TMP44:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 40 326; VF-FOUR-CHECK-NEXT: [[TMP45:%.*]] = bitcast float* [[TMP44]] to <4 x float>* 327; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD11:%.*]] = load <4 x float>, <4 x float>* [[TMP45]], align 4 328; VF-FOUR-CHECK-NEXT: [[TMP46:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 44 329; VF-FOUR-CHECK-NEXT: [[TMP47:%.*]] = bitcast float* [[TMP46]] to <4 x float>* 330; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD12:%.*]] = load <4 x float>, <4 x float>* [[TMP47]], align 4 331; VF-FOUR-CHECK-NEXT: [[TMP48:%.*]] = getelementptr inbounds float, float* [[CC:%.*]], i64 [[TMP0]] 332; VF-FOUR-CHECK-NEXT: [[TMP49:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP1]] 333; VF-FOUR-CHECK-NEXT: [[TMP50:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP2]] 334; VF-FOUR-CHECK-NEXT: [[TMP51:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP3]] 335; VF-FOUR-CHECK-NEXT: [[TMP52:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP4]] 336; VF-FOUR-CHECK-NEXT: [[TMP53:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP5]] 337; VF-FOUR-CHECK-NEXT: [[TMP54:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP6]] 338; VF-FOUR-CHECK-NEXT: [[TMP55:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP7]] 339; VF-FOUR-CHECK-NEXT: [[TMP56:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP8]] 340; VF-FOUR-CHECK-NEXT: [[TMP57:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP9]] 341; VF-FOUR-CHECK-NEXT: [[TMP58:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP10]] 342; VF-FOUR-CHECK-NEXT: [[TMP59:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP11]] 343; VF-FOUR-CHECK-NEXT: [[TMP60:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 0 344; VF-FOUR-CHECK-NEXT: [[TMP61:%.*]] = bitcast float* [[TMP60]] to <4 x float>* 345; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD13:%.*]] = load <4 x float>, <4 x float>* [[TMP61]], align 4 346; VF-FOUR-CHECK-NEXT: [[TMP62:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 4 347; VF-FOUR-CHECK-NEXT: [[TMP63:%.*]] = bitcast float* [[TMP62]] to <4 x float>* 348; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD14:%.*]] = load <4 x float>, <4 x float>* [[TMP63]], align 4 349; VF-FOUR-CHECK-NEXT: [[TMP64:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 8 350; VF-FOUR-CHECK-NEXT: [[TMP65:%.*]] = bitcast float* [[TMP64]] to <4 x float>* 351; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD15:%.*]] = load <4 x float>, <4 x float>* [[TMP65]], align 4 352; VF-FOUR-CHECK-NEXT: [[TMP66:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 12 353; VF-FOUR-CHECK-NEXT: [[TMP67:%.*]] = bitcast float* [[TMP66]] to <4 x float>* 354; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD16:%.*]] = load <4 x float>, <4 x float>* [[TMP67]], align 4 355; VF-FOUR-CHECK-NEXT: [[TMP68:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 16 356; VF-FOUR-CHECK-NEXT: [[TMP69:%.*]] = bitcast float* [[TMP68]] to <4 x float>* 357; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD17:%.*]] = load <4 x float>, <4 x float>* [[TMP69]], align 4 358; VF-FOUR-CHECK-NEXT: [[TMP70:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 20 359; VF-FOUR-CHECK-NEXT: [[TMP71:%.*]] = bitcast float* [[TMP70]] to <4 x float>* 360; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD18:%.*]] = load <4 x float>, <4 x float>* [[TMP71]], align 4 361; VF-FOUR-CHECK-NEXT: [[TMP72:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 24 362; VF-FOUR-CHECK-NEXT: [[TMP73:%.*]] = bitcast float* [[TMP72]] to <4 x float>* 363; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD19:%.*]] = load <4 x float>, <4 x float>* [[TMP73]], align 4 364; VF-FOUR-CHECK-NEXT: [[TMP74:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 28 365; VF-FOUR-CHECK-NEXT: [[TMP75:%.*]] = bitcast float* [[TMP74]] to <4 x float>* 366; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD20:%.*]] = load <4 x float>, <4 x float>* [[TMP75]], align 4 367; VF-FOUR-CHECK-NEXT: [[TMP76:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 32 368; VF-FOUR-CHECK-NEXT: [[TMP77:%.*]] = bitcast float* [[TMP76]] to <4 x float>* 369; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD21:%.*]] = load <4 x float>, <4 x float>* [[TMP77]], align 4 370; VF-FOUR-CHECK-NEXT: [[TMP78:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 36 371; VF-FOUR-CHECK-NEXT: [[TMP79:%.*]] = bitcast float* [[TMP78]] to <4 x float>* 372; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD22:%.*]] = load <4 x float>, <4 x float>* [[TMP79]], align 4 373; VF-FOUR-CHECK-NEXT: [[TMP80:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 40 374; VF-FOUR-CHECK-NEXT: [[TMP81:%.*]] = bitcast float* [[TMP80]] to <4 x float>* 375; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD23:%.*]] = load <4 x float>, <4 x float>* [[TMP81]], align 4 376; VF-FOUR-CHECK-NEXT: [[TMP82:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 44 377; VF-FOUR-CHECK-NEXT: [[TMP83:%.*]] = bitcast float* [[TMP82]] to <4 x float>* 378; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD24:%.*]] = load <4 x float>, <4 x float>* [[TMP83]], align 4 379; VF-FOUR-CHECK-NEXT: [[TMP84:%.*]] = fadd fast <4 x float> [[WIDE_LOAD]], [[WIDE_LOAD13]] 380; VF-FOUR-CHECK-NEXT: [[TMP85:%.*]] = fadd fast <4 x float> [[WIDE_LOAD2]], [[WIDE_LOAD14]] 381; VF-FOUR-CHECK-NEXT: [[TMP86:%.*]] = fadd fast <4 x float> [[WIDE_LOAD3]], [[WIDE_LOAD15]] 382; VF-FOUR-CHECK-NEXT: [[TMP87:%.*]] = fadd fast <4 x float> [[WIDE_LOAD4]], [[WIDE_LOAD16]] 383; VF-FOUR-CHECK-NEXT: [[TMP88:%.*]] = fadd fast <4 x float> [[WIDE_LOAD5]], [[WIDE_LOAD17]] 384; VF-FOUR-CHECK-NEXT: [[TMP89:%.*]] = fadd fast <4 x float> [[WIDE_LOAD6]], [[WIDE_LOAD18]] 385; VF-FOUR-CHECK-NEXT: [[TMP90:%.*]] = fadd fast <4 x float> [[WIDE_LOAD7]], [[WIDE_LOAD19]] 386; VF-FOUR-CHECK-NEXT: [[TMP91:%.*]] = fadd fast <4 x float> [[WIDE_LOAD8]], [[WIDE_LOAD20]] 387; VF-FOUR-CHECK-NEXT: [[TMP92:%.*]] = fadd fast <4 x float> [[WIDE_LOAD9]], [[WIDE_LOAD21]] 388; VF-FOUR-CHECK-NEXT: [[TMP93:%.*]] = fadd fast <4 x float> [[WIDE_LOAD10]], [[WIDE_LOAD22]] 389; VF-FOUR-CHECK-NEXT: [[TMP94:%.*]] = fadd fast <4 x float> [[WIDE_LOAD11]], [[WIDE_LOAD23]] 390; VF-FOUR-CHECK-NEXT: [[TMP95:%.*]] = fadd fast <4 x float> [[WIDE_LOAD12]], [[WIDE_LOAD24]] 391; VF-FOUR-CHECK-NEXT: [[TMP96:%.*]] = getelementptr inbounds float, float* [[AA:%.*]], i64 [[TMP0]] 392; VF-FOUR-CHECK-NEXT: [[TMP97:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP1]] 393; VF-FOUR-CHECK-NEXT: [[TMP98:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP2]] 394; VF-FOUR-CHECK-NEXT: [[TMP99:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP3]] 395; VF-FOUR-CHECK-NEXT: [[TMP100:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP4]] 396; VF-FOUR-CHECK-NEXT: [[TMP101:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP5]] 397; VF-FOUR-CHECK-NEXT: [[TMP102:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP6]] 398; VF-FOUR-CHECK-NEXT: [[TMP103:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP7]] 399; VF-FOUR-CHECK-NEXT: [[TMP104:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP8]] 400; VF-FOUR-CHECK-NEXT: [[TMP105:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP9]] 401; VF-FOUR-CHECK-NEXT: [[TMP106:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP10]] 402; VF-FOUR-CHECK-NEXT: [[TMP107:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP11]] 403; VF-FOUR-CHECK-NEXT: [[TMP108:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 0 404; VF-FOUR-CHECK-NEXT: [[TMP109:%.*]] = bitcast float* [[TMP108]] to <4 x float>* 405; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP84]], <4 x float>* [[TMP109]], align 4 406; VF-FOUR-CHECK-NEXT: [[TMP110:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 4 407; VF-FOUR-CHECK-NEXT: [[TMP111:%.*]] = bitcast float* [[TMP110]] to <4 x float>* 408; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP85]], <4 x float>* [[TMP111]], align 4 409; VF-FOUR-CHECK-NEXT: [[TMP112:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 8 410; VF-FOUR-CHECK-NEXT: [[TMP113:%.*]] = bitcast float* [[TMP112]] to <4 x float>* 411; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP86]], <4 x float>* [[TMP113]], align 4 412; VF-FOUR-CHECK-NEXT: [[TMP114:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 12 413; VF-FOUR-CHECK-NEXT: [[TMP115:%.*]] = bitcast float* [[TMP114]] to <4 x float>* 414; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP87]], <4 x float>* [[TMP115]], align 4 415; VF-FOUR-CHECK-NEXT: [[TMP116:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 16 416; VF-FOUR-CHECK-NEXT: [[TMP117:%.*]] = bitcast float* [[TMP116]] to <4 x float>* 417; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP88]], <4 x float>* [[TMP117]], align 4 418; VF-FOUR-CHECK-NEXT: [[TMP118:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 20 419; VF-FOUR-CHECK-NEXT: [[TMP119:%.*]] = bitcast float* [[TMP118]] to <4 x float>* 420; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP89]], <4 x float>* [[TMP119]], align 4 421; VF-FOUR-CHECK-NEXT: [[TMP120:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 24 422; VF-FOUR-CHECK-NEXT: [[TMP121:%.*]] = bitcast float* [[TMP120]] to <4 x float>* 423; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP90]], <4 x float>* [[TMP121]], align 4 424; VF-FOUR-CHECK-NEXT: [[TMP122:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 28 425; VF-FOUR-CHECK-NEXT: [[TMP123:%.*]] = bitcast float* [[TMP122]] to <4 x float>* 426; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP91]], <4 x float>* [[TMP123]], align 4 427; VF-FOUR-CHECK-NEXT: [[TMP124:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 32 428; VF-FOUR-CHECK-NEXT: [[TMP125:%.*]] = bitcast float* [[TMP124]] to <4 x float>* 429; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP92]], <4 x float>* [[TMP125]], align 4 430; VF-FOUR-CHECK-NEXT: [[TMP126:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 36 431; VF-FOUR-CHECK-NEXT: [[TMP127:%.*]] = bitcast float* [[TMP126]] to <4 x float>* 432; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP93]], <4 x float>* [[TMP127]], align 4 433; VF-FOUR-CHECK-NEXT: [[TMP128:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 40 434; VF-FOUR-CHECK-NEXT: [[TMP129:%.*]] = bitcast float* [[TMP128]] to <4 x float>* 435; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP94]], <4 x float>* [[TMP129]], align 4 436; VF-FOUR-CHECK-NEXT: [[TMP130:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 44 437; VF-FOUR-CHECK-NEXT: [[TMP131:%.*]] = bitcast float* [[TMP130]] to <4 x float>* 438; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP95]], <4 x float>* [[TMP131]], align 4 439; VF-FOUR-CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 48 440; VF-FOUR-CHECK-NEXT: [[TMP132:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] 441; VF-FOUR-CHECK-NEXT: br i1 [[TMP132]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] 442; VF-FOUR-CHECK: middle.block: 443; VF-FOUR-CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC]] 444; VF-FOUR-CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]] 445; VF-FOUR-CHECK: vec.epilog.iter.check: 446; VF-FOUR-CHECK-NEXT: [[N_VEC_REMAINING:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_VEC]] 447; VF-FOUR-CHECK-NEXT: [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 4 448; VF-FOUR-CHECK-NEXT: br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]] 449; VF-FOUR-CHECK: vec.epilog.ph: 450; VF-FOUR-CHECK-NEXT: [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ] 451; VF-FOUR-CHECK-NEXT: [[N_MOD_VF25:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 4 452; VF-FOUR-CHECK-NEXT: [[N_VEC26:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF25]] 453; VF-FOUR-CHECK-NEXT: br label [[VEC_EPILOG_VECTOR_BODY:%.*]] 454; VF-FOUR-CHECK: vec.epilog.vector.body: 455; VF-FOUR-CHECK-NEXT: [[INDEX27:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT28:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ] 456; VF-FOUR-CHECK-NEXT: [[TMP133:%.*]] = add i64 [[INDEX27]], 0 457; VF-FOUR-CHECK-NEXT: [[TMP134:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP133]] 458; VF-FOUR-CHECK-NEXT: [[TMP135:%.*]] = getelementptr inbounds float, float* [[TMP134]], i32 0 459; VF-FOUR-CHECK-NEXT: [[TMP136:%.*]] = bitcast float* [[TMP135]] to <4 x float>* 460; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD30:%.*]] = load <4 x float>, <4 x float>* [[TMP136]], align 4 461; VF-FOUR-CHECK-NEXT: [[TMP137:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP133]] 462; VF-FOUR-CHECK-NEXT: [[TMP138:%.*]] = getelementptr inbounds float, float* [[TMP137]], i32 0 463; VF-FOUR-CHECK-NEXT: [[TMP139:%.*]] = bitcast float* [[TMP138]] to <4 x float>* 464; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD31:%.*]] = load <4 x float>, <4 x float>* [[TMP139]], align 4 465; VF-FOUR-CHECK-NEXT: [[TMP140:%.*]] = fadd fast <4 x float> [[WIDE_LOAD30]], [[WIDE_LOAD31]] 466; VF-FOUR-CHECK-NEXT: [[TMP141:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP133]] 467; VF-FOUR-CHECK-NEXT: [[TMP142:%.*]] = getelementptr inbounds float, float* [[TMP141]], i32 0 468; VF-FOUR-CHECK-NEXT: [[TMP143:%.*]] = bitcast float* [[TMP142]] to <4 x float>* 469; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP140]], <4 x float>* [[TMP143]], align 4 470; VF-FOUR-CHECK-NEXT: [[INDEX_NEXT28]] = add nuw i64 [[INDEX27]], 4 471; VF-FOUR-CHECK-NEXT: [[TMP144:%.*]] = icmp eq i64 [[INDEX_NEXT28]], [[N_VEC26]] 472; VF-FOUR-CHECK-NEXT: br i1 [[TMP144]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP2:![0-9]+]] 473; VF-FOUR-CHECK: vec.epilog.middle.block: 474; VF-FOUR-CHECK-NEXT: [[CMP_N29:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC26]] 475; VF-FOUR-CHECK-NEXT: br i1 [[CMP_N29]], label [[FOR_END_LOOPEXIT_LOOPEXIT:%.*]], label [[VEC_EPILOG_SCALAR_PH]] 476; VF-FOUR-CHECK: vec.epilog.scalar.ph: 477; VF-FOUR-CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC26]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK]] ] 478; VF-FOUR-CHECK-NEXT: br label [[FOR_BODY:%.*]] 479; VF-FOUR-CHECK: for.body: 480; VF-FOUR-CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ] 481; VF-FOUR-CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[INDVARS_IV]] 482; VF-FOUR-CHECK-NEXT: [[TMP145:%.*]] = load float, float* [[ARRAYIDX]], align 4 483; VF-FOUR-CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[INDVARS_IV]] 484; VF-FOUR-CHECK-NEXT: [[TMP146:%.*]] = load float, float* [[ARRAYIDX2]], align 4 485; VF-FOUR-CHECK-NEXT: [[ADD:%.*]] = fadd fast float [[TMP145]], [[TMP146]] 486; VF-FOUR-CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[INDVARS_IV]] 487; VF-FOUR-CHECK-NEXT: store float [[ADD]], float* [[ARRAYIDX4]], align 4 488; VF-FOUR-CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1 489; VF-FOUR-CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i64 [[INDVARS_IV_NEXT]], [[WIDE_TRIP_COUNT]] 490; VF-FOUR-CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_BODY]], label [[FOR_END_LOOPEXIT_LOOPEXIT]], !llvm.loop [[LOOP4:![0-9]+]] 491; VF-FOUR-CHECK: for.end.loopexit.loopexit: 492; VF-FOUR-CHECK-NEXT: br label [[FOR_END_LOOPEXIT]] 493; VF-FOUR-CHECK: for.end.loopexit: 494; VF-FOUR-CHECK-NEXT: br label [[FOR_END]] 495; VF-FOUR-CHECK: for.end: 496; VF-FOUR-CHECK-NEXT: ret void 497; 498 499 500entry: 501 %cmp1 = icmp sgt i32 %N, 0 502 br i1 %cmp1, label %for.body.preheader, label %for.end 503 504for.body.preheader: ; preds = %entry 505 %wide.trip.count = zext i32 %N to i64 506 br label %for.body 507 508for.body: ; preds = %for.body.preheader, %for.body 509 %indvars.iv = phi i64 [ 0, %for.body.preheader ], [ %indvars.iv.next, %for.body ] 510 %arrayidx = getelementptr inbounds float, float* %bb, i64 %indvars.iv 511 %0 = load float, float* %arrayidx, align 4 512 %arrayidx2 = getelementptr inbounds float, float* %cc, i64 %indvars.iv 513 %1 = load float, float* %arrayidx2, align 4 514 %add = fadd fast float %0, %1 515 %arrayidx4 = getelementptr inbounds float, float* %aa, i64 %indvars.iv 516 store float %add, float* %arrayidx4, align 4 517 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 518 %exitcond = icmp ne i64 %indvars.iv.next, %wide.trip.count 519 br i1 %exitcond, label %for.body, label %for.end.loopexit 520 521for.end.loopexit: ; preds = %for.body 522 br label %for.end 523 524for.end: ; preds = %for.end.loopexit, %entry 525 ret void 526} 527 528define dso_local signext i32 @f2(float* noalias %A, float* noalias %B, i32 signext %n) #0 { 529; VF-TWO-CHECK-LABEL: @f2( 530; VF-TWO-CHECK-NEXT: entry: 531; VF-TWO-CHECK-NEXT: [[CMP1:%.*]] = icmp sgt i32 [[N:%.*]], 1 532; VF-TWO-CHECK-NEXT: br i1 [[CMP1]], label [[ITER_CHECK:%.*]], label [[FOR_END:%.*]] 533; VF-TWO-CHECK: iter.check: 534; VF-TWO-CHECK-NEXT: [[TMP0:%.*]] = add i32 [[N]], -1 535; VF-TWO-CHECK-NEXT: [[WIDE_TRIP_COUNT:%.*]] = zext i32 [[TMP0]] to i64 536; VF-TWO-CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 2 537; VF-TWO-CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]] 538; VF-TWO-CHECK: vector.scevcheck: 539; VF-TWO-CHECK-NEXT: [[TMP1:%.*]] = add nsw i64 [[WIDE_TRIP_COUNT]], -1 540; VF-TWO-CHECK-NEXT: [[TMP2:%.*]] = trunc i64 [[TMP1]] to i32 541; VF-TWO-CHECK-NEXT: [[MUL:%.*]] = call { i32, i1 } @llvm.umul.with.overflow.i32(i32 1, i32 [[TMP2]]) 542; VF-TWO-CHECK-NEXT: [[MUL_RESULT:%.*]] = extractvalue { i32, i1 } [[MUL]], 0 543; VF-TWO-CHECK-NEXT: [[MUL_OVERFLOW:%.*]] = extractvalue { i32, i1 } [[MUL]], 1 544; VF-TWO-CHECK-NEXT: [[TMP3:%.*]] = add i32 [[TMP0]], [[MUL_RESULT]] 545; VF-TWO-CHECK-NEXT: [[TMP4:%.*]] = sub i32 [[TMP0]], [[MUL_RESULT]] 546; VF-TWO-CHECK-NEXT: [[TMP5:%.*]] = icmp sgt i32 [[TMP4]], [[TMP0]] 547; VF-TWO-CHECK-NEXT: [[TMP6:%.*]] = icmp slt i32 [[TMP3]], [[TMP0]] 548; VF-TWO-CHECK-NEXT: [[TMP7:%.*]] = icmp ugt i64 [[TMP1]], 4294967295 549; VF-TWO-CHECK-NEXT: [[TMP8:%.*]] = or i1 [[TMP5]], [[TMP7]] 550; VF-TWO-CHECK-NEXT: [[TMP9:%.*]] = or i1 [[TMP8]], [[MUL_OVERFLOW]] 551; VF-TWO-CHECK-NEXT: br i1 [[TMP9]], label [[VEC_EPILOG_SCALAR_PH]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]] 552; VF-TWO-CHECK: vector.main.loop.iter.check: 553; VF-TWO-CHECK-NEXT: [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 32 554; VF-TWO-CHECK-NEXT: br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]] 555; VF-TWO-CHECK: vector.ph: 556; VF-TWO-CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 32 557; VF-TWO-CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF]] 558; VF-TWO-CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 559; VF-TWO-CHECK: vector.body: 560; VF-TWO-CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 561; VF-TWO-CHECK-NEXT: [[TMP10:%.*]] = add i64 [[INDEX]], 0 562; VF-TWO-CHECK-NEXT: [[TMP11:%.*]] = add i64 [[INDEX]], 4 563; VF-TWO-CHECK-NEXT: [[TMP12:%.*]] = add i64 [[INDEX]], 8 564; VF-TWO-CHECK-NEXT: [[TMP13:%.*]] = add i64 [[INDEX]], 12 565; VF-TWO-CHECK-NEXT: [[TMP14:%.*]] = add i64 [[INDEX]], 16 566; VF-TWO-CHECK-NEXT: [[TMP15:%.*]] = add i64 [[INDEX]], 20 567; VF-TWO-CHECK-NEXT: [[TMP16:%.*]] = add i64 [[INDEX]], 24 568; VF-TWO-CHECK-NEXT: [[TMP17:%.*]] = add i64 [[INDEX]], 28 569; VF-TWO-CHECK-NEXT: [[OFFSET_IDX:%.*]] = trunc i64 [[INDEX]] to i32 570; VF-TWO-CHECK-NEXT: [[TMP18:%.*]] = add i32 [[OFFSET_IDX]], 0 571; VF-TWO-CHECK-NEXT: [[TMP19:%.*]] = add i32 [[OFFSET_IDX]], 4 572; VF-TWO-CHECK-NEXT: [[TMP20:%.*]] = add i32 [[OFFSET_IDX]], 8 573; VF-TWO-CHECK-NEXT: [[TMP21:%.*]] = add i32 [[OFFSET_IDX]], 12 574; VF-TWO-CHECK-NEXT: [[TMP22:%.*]] = add i32 [[OFFSET_IDX]], 16 575; VF-TWO-CHECK-NEXT: [[TMP23:%.*]] = add i32 [[OFFSET_IDX]], 20 576; VF-TWO-CHECK-NEXT: [[TMP24:%.*]] = add i32 [[OFFSET_IDX]], 24 577; VF-TWO-CHECK-NEXT: [[TMP25:%.*]] = add i32 [[OFFSET_IDX]], 28 578; VF-TWO-CHECK-NEXT: [[TMP26:%.*]] = xor i32 [[TMP18]], -1 579; VF-TWO-CHECK-NEXT: [[TMP27:%.*]] = xor i32 [[TMP19]], -1 580; VF-TWO-CHECK-NEXT: [[TMP28:%.*]] = xor i32 [[TMP20]], -1 581; VF-TWO-CHECK-NEXT: [[TMP29:%.*]] = xor i32 [[TMP21]], -1 582; VF-TWO-CHECK-NEXT: [[TMP30:%.*]] = xor i32 [[TMP22]], -1 583; VF-TWO-CHECK-NEXT: [[TMP31:%.*]] = xor i32 [[TMP23]], -1 584; VF-TWO-CHECK-NEXT: [[TMP32:%.*]] = xor i32 [[TMP24]], -1 585; VF-TWO-CHECK-NEXT: [[TMP33:%.*]] = xor i32 [[TMP25]], -1 586; VF-TWO-CHECK-NEXT: [[TMP34:%.*]] = add i32 [[TMP26]], [[N]] 587; VF-TWO-CHECK-NEXT: [[TMP35:%.*]] = add i32 [[TMP27]], [[N]] 588; VF-TWO-CHECK-NEXT: [[TMP36:%.*]] = add i32 [[TMP28]], [[N]] 589; VF-TWO-CHECK-NEXT: [[TMP37:%.*]] = add i32 [[TMP29]], [[N]] 590; VF-TWO-CHECK-NEXT: [[TMP38:%.*]] = add i32 [[TMP30]], [[N]] 591; VF-TWO-CHECK-NEXT: [[TMP39:%.*]] = add i32 [[TMP31]], [[N]] 592; VF-TWO-CHECK-NEXT: [[TMP40:%.*]] = add i32 [[TMP32]], [[N]] 593; VF-TWO-CHECK-NEXT: [[TMP41:%.*]] = add i32 [[TMP33]], [[N]] 594; VF-TWO-CHECK-NEXT: [[TMP42:%.*]] = sext i32 [[TMP34]] to i64 595; VF-TWO-CHECK-NEXT: [[TMP43:%.*]] = sext i32 [[TMP35]] to i64 596; VF-TWO-CHECK-NEXT: [[TMP44:%.*]] = sext i32 [[TMP36]] to i64 597; VF-TWO-CHECK-NEXT: [[TMP45:%.*]] = sext i32 [[TMP37]] to i64 598; VF-TWO-CHECK-NEXT: [[TMP46:%.*]] = sext i32 [[TMP38]] to i64 599; VF-TWO-CHECK-NEXT: [[TMP47:%.*]] = sext i32 [[TMP39]] to i64 600; VF-TWO-CHECK-NEXT: [[TMP48:%.*]] = sext i32 [[TMP40]] to i64 601; VF-TWO-CHECK-NEXT: [[TMP49:%.*]] = sext i32 [[TMP41]] to i64 602; VF-TWO-CHECK-NEXT: [[TMP50:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 [[TMP42]] 603; VF-TWO-CHECK-NEXT: [[TMP51:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP43]] 604; VF-TWO-CHECK-NEXT: [[TMP52:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP44]] 605; VF-TWO-CHECK-NEXT: [[TMP53:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP45]] 606; VF-TWO-CHECK-NEXT: [[TMP54:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP46]] 607; VF-TWO-CHECK-NEXT: [[TMP55:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP47]] 608; VF-TWO-CHECK-NEXT: [[TMP56:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP48]] 609; VF-TWO-CHECK-NEXT: [[TMP57:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP49]] 610; VF-TWO-CHECK-NEXT: [[TMP58:%.*]] = getelementptr inbounds float, float* [[TMP50]], i32 0 611; VF-TWO-CHECK-NEXT: [[TMP59:%.*]] = getelementptr inbounds float, float* [[TMP58]], i32 -3 612; VF-TWO-CHECK-NEXT: [[TMP60:%.*]] = bitcast float* [[TMP59]] to <4 x float>* 613; VF-TWO-CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, <4 x float>* [[TMP60]], align 4 614; VF-TWO-CHECK-NEXT: [[REVERSE:%.*]] = shufflevector <4 x float> [[WIDE_LOAD]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 615; VF-TWO-CHECK-NEXT: [[TMP61:%.*]] = getelementptr inbounds float, float* [[TMP50]], i32 -4 616; VF-TWO-CHECK-NEXT: [[TMP62:%.*]] = getelementptr inbounds float, float* [[TMP61]], i32 -3 617; VF-TWO-CHECK-NEXT: [[TMP63:%.*]] = bitcast float* [[TMP62]] to <4 x float>* 618; VF-TWO-CHECK-NEXT: [[WIDE_LOAD2:%.*]] = load <4 x float>, <4 x float>* [[TMP63]], align 4 619; VF-TWO-CHECK-NEXT: [[REVERSE3:%.*]] = shufflevector <4 x float> [[WIDE_LOAD2]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 620; VF-TWO-CHECK-NEXT: [[TMP64:%.*]] = getelementptr inbounds float, float* [[TMP50]], i32 -8 621; VF-TWO-CHECK-NEXT: [[TMP65:%.*]] = getelementptr inbounds float, float* [[TMP64]], i32 -3 622; VF-TWO-CHECK-NEXT: [[TMP66:%.*]] = bitcast float* [[TMP65]] to <4 x float>* 623; VF-TWO-CHECK-NEXT: [[WIDE_LOAD4:%.*]] = load <4 x float>, <4 x float>* [[TMP66]], align 4 624; VF-TWO-CHECK-NEXT: [[REVERSE5:%.*]] = shufflevector <4 x float> [[WIDE_LOAD4]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 625; VF-TWO-CHECK-NEXT: [[TMP67:%.*]] = getelementptr inbounds float, float* [[TMP50]], i32 -12 626; VF-TWO-CHECK-NEXT: [[TMP68:%.*]] = getelementptr inbounds float, float* [[TMP67]], i32 -3 627; VF-TWO-CHECK-NEXT: [[TMP69:%.*]] = bitcast float* [[TMP68]] to <4 x float>* 628; VF-TWO-CHECK-NEXT: [[WIDE_LOAD6:%.*]] = load <4 x float>, <4 x float>* [[TMP69]], align 4 629; VF-TWO-CHECK-NEXT: [[REVERSE7:%.*]] = shufflevector <4 x float> [[WIDE_LOAD6]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 630; VF-TWO-CHECK-NEXT: [[TMP70:%.*]] = getelementptr inbounds float, float* [[TMP50]], i32 -16 631; VF-TWO-CHECK-NEXT: [[TMP71:%.*]] = getelementptr inbounds float, float* [[TMP70]], i32 -3 632; VF-TWO-CHECK-NEXT: [[TMP72:%.*]] = bitcast float* [[TMP71]] to <4 x float>* 633; VF-TWO-CHECK-NEXT: [[WIDE_LOAD8:%.*]] = load <4 x float>, <4 x float>* [[TMP72]], align 4 634; VF-TWO-CHECK-NEXT: [[REVERSE9:%.*]] = shufflevector <4 x float> [[WIDE_LOAD8]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 635; VF-TWO-CHECK-NEXT: [[TMP73:%.*]] = getelementptr inbounds float, float* [[TMP50]], i32 -20 636; VF-TWO-CHECK-NEXT: [[TMP74:%.*]] = getelementptr inbounds float, float* [[TMP73]], i32 -3 637; VF-TWO-CHECK-NEXT: [[TMP75:%.*]] = bitcast float* [[TMP74]] to <4 x float>* 638; VF-TWO-CHECK-NEXT: [[WIDE_LOAD10:%.*]] = load <4 x float>, <4 x float>* [[TMP75]], align 4 639; VF-TWO-CHECK-NEXT: [[REVERSE11:%.*]] = shufflevector <4 x float> [[WIDE_LOAD10]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 640; VF-TWO-CHECK-NEXT: [[TMP76:%.*]] = getelementptr inbounds float, float* [[TMP50]], i32 -24 641; VF-TWO-CHECK-NEXT: [[TMP77:%.*]] = getelementptr inbounds float, float* [[TMP76]], i32 -3 642; VF-TWO-CHECK-NEXT: [[TMP78:%.*]] = bitcast float* [[TMP77]] to <4 x float>* 643; VF-TWO-CHECK-NEXT: [[WIDE_LOAD12:%.*]] = load <4 x float>, <4 x float>* [[TMP78]], align 4 644; VF-TWO-CHECK-NEXT: [[REVERSE13:%.*]] = shufflevector <4 x float> [[WIDE_LOAD12]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 645; VF-TWO-CHECK-NEXT: [[TMP79:%.*]] = getelementptr inbounds float, float* [[TMP50]], i32 -28 646; VF-TWO-CHECK-NEXT: [[TMP80:%.*]] = getelementptr inbounds float, float* [[TMP79]], i32 -3 647; VF-TWO-CHECK-NEXT: [[TMP81:%.*]] = bitcast float* [[TMP80]] to <4 x float>* 648; VF-TWO-CHECK-NEXT: [[WIDE_LOAD14:%.*]] = load <4 x float>, <4 x float>* [[TMP81]], align 4 649; VF-TWO-CHECK-NEXT: [[REVERSE15:%.*]] = shufflevector <4 x float> [[WIDE_LOAD14]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 650; VF-TWO-CHECK-NEXT: [[TMP82:%.*]] = fadd fast <4 x float> [[REVERSE]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 651; VF-TWO-CHECK-NEXT: [[TMP83:%.*]] = fadd fast <4 x float> [[REVERSE3]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 652; VF-TWO-CHECK-NEXT: [[TMP84:%.*]] = fadd fast <4 x float> [[REVERSE5]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 653; VF-TWO-CHECK-NEXT: [[TMP85:%.*]] = fadd fast <4 x float> [[REVERSE7]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 654; VF-TWO-CHECK-NEXT: [[TMP86:%.*]] = fadd fast <4 x float> [[REVERSE9]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 655; VF-TWO-CHECK-NEXT: [[TMP87:%.*]] = fadd fast <4 x float> [[REVERSE11]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 656; VF-TWO-CHECK-NEXT: [[TMP88:%.*]] = fadd fast <4 x float> [[REVERSE13]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 657; VF-TWO-CHECK-NEXT: [[TMP89:%.*]] = fadd fast <4 x float> [[REVERSE15]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 658; VF-TWO-CHECK-NEXT: [[TMP90:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[TMP10]] 659; VF-TWO-CHECK-NEXT: [[TMP91:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP11]] 660; VF-TWO-CHECK-NEXT: [[TMP92:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP12]] 661; VF-TWO-CHECK-NEXT: [[TMP93:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP13]] 662; VF-TWO-CHECK-NEXT: [[TMP94:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP14]] 663; VF-TWO-CHECK-NEXT: [[TMP95:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP15]] 664; VF-TWO-CHECK-NEXT: [[TMP96:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP16]] 665; VF-TWO-CHECK-NEXT: [[TMP97:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP17]] 666; VF-TWO-CHECK-NEXT: [[TMP98:%.*]] = getelementptr inbounds float, float* [[TMP90]], i32 0 667; VF-TWO-CHECK-NEXT: [[TMP99:%.*]] = bitcast float* [[TMP98]] to <4 x float>* 668; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP82]], <4 x float>* [[TMP99]], align 4 669; VF-TWO-CHECK-NEXT: [[TMP100:%.*]] = getelementptr inbounds float, float* [[TMP90]], i32 4 670; VF-TWO-CHECK-NEXT: [[TMP101:%.*]] = bitcast float* [[TMP100]] to <4 x float>* 671; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP83]], <4 x float>* [[TMP101]], align 4 672; VF-TWO-CHECK-NEXT: [[TMP102:%.*]] = getelementptr inbounds float, float* [[TMP90]], i32 8 673; VF-TWO-CHECK-NEXT: [[TMP103:%.*]] = bitcast float* [[TMP102]] to <4 x float>* 674; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP84]], <4 x float>* [[TMP103]], align 4 675; VF-TWO-CHECK-NEXT: [[TMP104:%.*]] = getelementptr inbounds float, float* [[TMP90]], i32 12 676; VF-TWO-CHECK-NEXT: [[TMP105:%.*]] = bitcast float* [[TMP104]] to <4 x float>* 677; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP85]], <4 x float>* [[TMP105]], align 4 678; VF-TWO-CHECK-NEXT: [[TMP106:%.*]] = getelementptr inbounds float, float* [[TMP90]], i32 16 679; VF-TWO-CHECK-NEXT: [[TMP107:%.*]] = bitcast float* [[TMP106]] to <4 x float>* 680; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP86]], <4 x float>* [[TMP107]], align 4 681; VF-TWO-CHECK-NEXT: [[TMP108:%.*]] = getelementptr inbounds float, float* [[TMP90]], i32 20 682; VF-TWO-CHECK-NEXT: [[TMP109:%.*]] = bitcast float* [[TMP108]] to <4 x float>* 683; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP87]], <4 x float>* [[TMP109]], align 4 684; VF-TWO-CHECK-NEXT: [[TMP110:%.*]] = getelementptr inbounds float, float* [[TMP90]], i32 24 685; VF-TWO-CHECK-NEXT: [[TMP111:%.*]] = bitcast float* [[TMP110]] to <4 x float>* 686; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP88]], <4 x float>* [[TMP111]], align 4 687; VF-TWO-CHECK-NEXT: [[TMP112:%.*]] = getelementptr inbounds float, float* [[TMP90]], i32 28 688; VF-TWO-CHECK-NEXT: [[TMP113:%.*]] = bitcast float* [[TMP112]] to <4 x float>* 689; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP89]], <4 x float>* [[TMP113]], align 4 690; VF-TWO-CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 32 691; VF-TWO-CHECK-NEXT: [[TMP114:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] 692; VF-TWO-CHECK-NEXT: br i1 [[TMP114]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]] 693; VF-TWO-CHECK: middle.block: 694; VF-TWO-CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC]] 695; VF-TWO-CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]] 696; VF-TWO-CHECK: vec.epilog.iter.check: 697; VF-TWO-CHECK-NEXT: [[IND_END21:%.*]] = trunc i64 [[N_VEC]] to i32 698; VF-TWO-CHECK-NEXT: [[N_VEC_REMAINING:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_VEC]] 699; VF-TWO-CHECK-NEXT: [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 2 700; VF-TWO-CHECK-NEXT: br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]] 701; VF-TWO-CHECK: vec.epilog.ph: 702; VF-TWO-CHECK-NEXT: [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ] 703; VF-TWO-CHECK-NEXT: [[N_MOD_VF16:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 2 704; VF-TWO-CHECK-NEXT: [[N_VEC17:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF16]] 705; VF-TWO-CHECK-NEXT: [[IND_END:%.*]] = trunc i64 [[N_VEC17]] to i32 706; VF-TWO-CHECK-NEXT: br label [[VEC_EPILOG_VECTOR_BODY:%.*]] 707; VF-TWO-CHECK: vec.epilog.vector.body: 708; VF-TWO-CHECK-NEXT: [[INDEX18:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT19:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ] 709; VF-TWO-CHECK-NEXT: [[TMP115:%.*]] = add i64 [[INDEX18]], 0 710; VF-TWO-CHECK-NEXT: [[OFFSET_IDX23:%.*]] = trunc i64 [[INDEX18]] to i32 711; VF-TWO-CHECK-NEXT: [[TMP116:%.*]] = add i32 [[OFFSET_IDX23]], 0 712; VF-TWO-CHECK-NEXT: [[TMP117:%.*]] = xor i32 [[TMP116]], -1 713; VF-TWO-CHECK-NEXT: [[TMP118:%.*]] = add i32 [[TMP117]], [[N]] 714; VF-TWO-CHECK-NEXT: [[TMP119:%.*]] = sext i32 [[TMP118]] to i64 715; VF-TWO-CHECK-NEXT: [[TMP120:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP119]] 716; VF-TWO-CHECK-NEXT: [[TMP121:%.*]] = getelementptr inbounds float, float* [[TMP120]], i32 0 717; VF-TWO-CHECK-NEXT: [[TMP122:%.*]] = getelementptr inbounds float, float* [[TMP121]], i32 -1 718; VF-TWO-CHECK-NEXT: [[TMP123:%.*]] = bitcast float* [[TMP122]] to <2 x float>* 719; VF-TWO-CHECK-NEXT: [[WIDE_LOAD24:%.*]] = load <2 x float>, <2 x float>* [[TMP123]], align 4 720; VF-TWO-CHECK-NEXT: [[REVERSE25:%.*]] = shufflevector <2 x float> [[WIDE_LOAD24]], <2 x float> poison, <2 x i32> <i32 1, i32 0> 721; VF-TWO-CHECK-NEXT: [[TMP124:%.*]] = fadd fast <2 x float> [[REVERSE25]], <float 1.000000e+00, float 1.000000e+00> 722; VF-TWO-CHECK-NEXT: [[TMP125:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP115]] 723; VF-TWO-CHECK-NEXT: [[TMP126:%.*]] = getelementptr inbounds float, float* [[TMP125]], i32 0 724; VF-TWO-CHECK-NEXT: [[TMP127:%.*]] = bitcast float* [[TMP126]] to <2 x float>* 725; VF-TWO-CHECK-NEXT: store <2 x float> [[TMP124]], <2 x float>* [[TMP127]], align 4 726; VF-TWO-CHECK-NEXT: [[INDEX_NEXT19]] = add nuw i64 [[INDEX18]], 2 727; VF-TWO-CHECK-NEXT: [[TMP128:%.*]] = icmp eq i64 [[INDEX_NEXT19]], [[N_VEC17]] 728; VF-TWO-CHECK-NEXT: br i1 [[TMP128]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]] 729; VF-TWO-CHECK: vec.epilog.middle.block: 730; VF-TWO-CHECK-NEXT: [[CMP_N22:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC17]] 731; VF-TWO-CHECK-NEXT: br i1 [[CMP_N22]], label [[FOR_END_LOOPEXIT_LOOPEXIT:%.*]], label [[VEC_EPILOG_SCALAR_PH]] 732; VF-TWO-CHECK: vec.epilog.scalar.ph: 733; VF-TWO-CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC17]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ITER_CHECK]] ] 734; VF-TWO-CHECK-NEXT: [[BC_RESUME_VAL20:%.*]] = phi i32 [ [[IND_END]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[IND_END21]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ITER_CHECK]] ] 735; VF-TWO-CHECK-NEXT: br label [[FOR_BODY:%.*]] 736; VF-TWO-CHECK: for.body: 737; VF-TWO-CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ] 738; VF-TWO-CHECK-NEXT: [[I_014:%.*]] = phi i32 [ [[BC_RESUME_VAL20]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ] 739; VF-TWO-CHECK-NEXT: [[TMP129:%.*]] = xor i32 [[I_014]], -1 740; VF-TWO-CHECK-NEXT: [[SUB2:%.*]] = add i32 [[TMP129]], [[N]] 741; VF-TWO-CHECK-NEXT: [[IDXPROM:%.*]] = sext i32 [[SUB2]] to i64 742; VF-TWO-CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[IDXPROM]] 743; VF-TWO-CHECK-NEXT: [[TMP130:%.*]] = load float, float* [[ARRAYIDX]], align 4 744; VF-TWO-CHECK-NEXT: [[CONV3:%.*]] = fadd fast float [[TMP130]], 1.000000e+00 745; VF-TWO-CHECK-NEXT: [[ARRAYIDX5:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDVARS_IV]] 746; VF-TWO-CHECK-NEXT: store float [[CONV3]], float* [[ARRAYIDX5]], align 4 747; VF-TWO-CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1 748; VF-TWO-CHECK-NEXT: [[INC]] = add nuw nsw i32 [[I_014]], 1 749; VF-TWO-CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i64 [[INDVARS_IV_NEXT]], [[WIDE_TRIP_COUNT]] 750; VF-TWO-CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_BODY]], label [[FOR_END_LOOPEXIT_LOOPEXIT]], !llvm.loop [[LOOP7:![0-9]+]] 751; VF-TWO-CHECK: for.end.loopexit.loopexit: 752; VF-TWO-CHECK-NEXT: br label [[FOR_END_LOOPEXIT]] 753; VF-TWO-CHECK: for.end.loopexit: 754; VF-TWO-CHECK-NEXT: br label [[FOR_END]] 755; VF-TWO-CHECK: for.end: 756; VF-TWO-CHECK-NEXT: ret i32 0 757; 758; VF-FOUR-CHECK-LABEL: @f2( 759; VF-FOUR-CHECK-NEXT: entry: 760; VF-FOUR-CHECK-NEXT: [[CMP1:%.*]] = icmp sgt i32 [[N:%.*]], 1 761; VF-FOUR-CHECK-NEXT: br i1 [[CMP1]], label [[ITER_CHECK:%.*]], label [[FOR_END:%.*]] 762; VF-FOUR-CHECK: iter.check: 763; VF-FOUR-CHECK-NEXT: [[TMP0:%.*]] = add i32 [[N]], -1 764; VF-FOUR-CHECK-NEXT: [[WIDE_TRIP_COUNT:%.*]] = zext i32 [[TMP0]] to i64 765; VF-FOUR-CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 4 766; VF-FOUR-CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]] 767; VF-FOUR-CHECK: vector.scevcheck: 768; VF-FOUR-CHECK-NEXT: [[TMP1:%.*]] = add nsw i64 [[WIDE_TRIP_COUNT]], -1 769; VF-FOUR-CHECK-NEXT: [[TMP2:%.*]] = trunc i64 [[TMP1]] to i32 770; VF-FOUR-CHECK-NEXT: [[MUL:%.*]] = call { i32, i1 } @llvm.umul.with.overflow.i32(i32 1, i32 [[TMP2]]) 771; VF-FOUR-CHECK-NEXT: [[MUL_RESULT:%.*]] = extractvalue { i32, i1 } [[MUL]], 0 772; VF-FOUR-CHECK-NEXT: [[MUL_OVERFLOW:%.*]] = extractvalue { i32, i1 } [[MUL]], 1 773; VF-FOUR-CHECK-NEXT: [[TMP3:%.*]] = add i32 [[TMP0]], [[MUL_RESULT]] 774; VF-FOUR-CHECK-NEXT: [[TMP4:%.*]] = sub i32 [[TMP0]], [[MUL_RESULT]] 775; VF-FOUR-CHECK-NEXT: [[TMP5:%.*]] = icmp sgt i32 [[TMP4]], [[TMP0]] 776; VF-FOUR-CHECK-NEXT: [[TMP6:%.*]] = icmp slt i32 [[TMP3]], [[TMP0]] 777; VF-FOUR-CHECK-NEXT: [[TMP7:%.*]] = icmp ugt i64 [[TMP1]], 4294967295 778; VF-FOUR-CHECK-NEXT: [[TMP8:%.*]] = or i1 [[TMP5]], [[TMP7]] 779; VF-FOUR-CHECK-NEXT: [[TMP9:%.*]] = or i1 [[TMP8]], [[MUL_OVERFLOW]] 780; VF-FOUR-CHECK-NEXT: br i1 [[TMP9]], label [[VEC_EPILOG_SCALAR_PH]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]] 781; VF-FOUR-CHECK: vector.main.loop.iter.check: 782; VF-FOUR-CHECK-NEXT: [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 32 783; VF-FOUR-CHECK-NEXT: br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]] 784; VF-FOUR-CHECK: vector.ph: 785; VF-FOUR-CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 32 786; VF-FOUR-CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF]] 787; VF-FOUR-CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 788; VF-FOUR-CHECK: vector.body: 789; VF-FOUR-CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 790; VF-FOUR-CHECK-NEXT: [[TMP10:%.*]] = add i64 [[INDEX]], 0 791; VF-FOUR-CHECK-NEXT: [[TMP11:%.*]] = add i64 [[INDEX]], 4 792; VF-FOUR-CHECK-NEXT: [[TMP12:%.*]] = add i64 [[INDEX]], 8 793; VF-FOUR-CHECK-NEXT: [[TMP13:%.*]] = add i64 [[INDEX]], 12 794; VF-FOUR-CHECK-NEXT: [[TMP14:%.*]] = add i64 [[INDEX]], 16 795; VF-FOUR-CHECK-NEXT: [[TMP15:%.*]] = add i64 [[INDEX]], 20 796; VF-FOUR-CHECK-NEXT: [[TMP16:%.*]] = add i64 [[INDEX]], 24 797; VF-FOUR-CHECK-NEXT: [[TMP17:%.*]] = add i64 [[INDEX]], 28 798; VF-FOUR-CHECK-NEXT: [[OFFSET_IDX:%.*]] = trunc i64 [[INDEX]] to i32 799; VF-FOUR-CHECK-NEXT: [[TMP18:%.*]] = add i32 [[OFFSET_IDX]], 0 800; VF-FOUR-CHECK-NEXT: [[TMP19:%.*]] = add i32 [[OFFSET_IDX]], 4 801; VF-FOUR-CHECK-NEXT: [[TMP20:%.*]] = add i32 [[OFFSET_IDX]], 8 802; VF-FOUR-CHECK-NEXT: [[TMP21:%.*]] = add i32 [[OFFSET_IDX]], 12 803; VF-FOUR-CHECK-NEXT: [[TMP22:%.*]] = add i32 [[OFFSET_IDX]], 16 804; VF-FOUR-CHECK-NEXT: [[TMP23:%.*]] = add i32 [[OFFSET_IDX]], 20 805; VF-FOUR-CHECK-NEXT: [[TMP24:%.*]] = add i32 [[OFFSET_IDX]], 24 806; VF-FOUR-CHECK-NEXT: [[TMP25:%.*]] = add i32 [[OFFSET_IDX]], 28 807; VF-FOUR-CHECK-NEXT: [[TMP26:%.*]] = xor i32 [[TMP18]], -1 808; VF-FOUR-CHECK-NEXT: [[TMP27:%.*]] = xor i32 [[TMP19]], -1 809; VF-FOUR-CHECK-NEXT: [[TMP28:%.*]] = xor i32 [[TMP20]], -1 810; VF-FOUR-CHECK-NEXT: [[TMP29:%.*]] = xor i32 [[TMP21]], -1 811; VF-FOUR-CHECK-NEXT: [[TMP30:%.*]] = xor i32 [[TMP22]], -1 812; VF-FOUR-CHECK-NEXT: [[TMP31:%.*]] = xor i32 [[TMP23]], -1 813; VF-FOUR-CHECK-NEXT: [[TMP32:%.*]] = xor i32 [[TMP24]], -1 814; VF-FOUR-CHECK-NEXT: [[TMP33:%.*]] = xor i32 [[TMP25]], -1 815; VF-FOUR-CHECK-NEXT: [[TMP34:%.*]] = add i32 [[TMP26]], [[N]] 816; VF-FOUR-CHECK-NEXT: [[TMP35:%.*]] = add i32 [[TMP27]], [[N]] 817; VF-FOUR-CHECK-NEXT: [[TMP36:%.*]] = add i32 [[TMP28]], [[N]] 818; VF-FOUR-CHECK-NEXT: [[TMP37:%.*]] = add i32 [[TMP29]], [[N]] 819; VF-FOUR-CHECK-NEXT: [[TMP38:%.*]] = add i32 [[TMP30]], [[N]] 820; VF-FOUR-CHECK-NEXT: [[TMP39:%.*]] = add i32 [[TMP31]], [[N]] 821; VF-FOUR-CHECK-NEXT: [[TMP40:%.*]] = add i32 [[TMP32]], [[N]] 822; VF-FOUR-CHECK-NEXT: [[TMP41:%.*]] = add i32 [[TMP33]], [[N]] 823; VF-FOUR-CHECK-NEXT: [[TMP42:%.*]] = sext i32 [[TMP34]] to i64 824; VF-FOUR-CHECK-NEXT: [[TMP43:%.*]] = sext i32 [[TMP35]] to i64 825; VF-FOUR-CHECK-NEXT: [[TMP44:%.*]] = sext i32 [[TMP36]] to i64 826; VF-FOUR-CHECK-NEXT: [[TMP45:%.*]] = sext i32 [[TMP37]] to i64 827; VF-FOUR-CHECK-NEXT: [[TMP46:%.*]] = sext i32 [[TMP38]] to i64 828; VF-FOUR-CHECK-NEXT: [[TMP47:%.*]] = sext i32 [[TMP39]] to i64 829; VF-FOUR-CHECK-NEXT: [[TMP48:%.*]] = sext i32 [[TMP40]] to i64 830; VF-FOUR-CHECK-NEXT: [[TMP49:%.*]] = sext i32 [[TMP41]] to i64 831; VF-FOUR-CHECK-NEXT: [[TMP50:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 [[TMP42]] 832; VF-FOUR-CHECK-NEXT: [[TMP51:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP43]] 833; VF-FOUR-CHECK-NEXT: [[TMP52:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP44]] 834; VF-FOUR-CHECK-NEXT: [[TMP53:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP45]] 835; VF-FOUR-CHECK-NEXT: [[TMP54:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP46]] 836; VF-FOUR-CHECK-NEXT: [[TMP55:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP47]] 837; VF-FOUR-CHECK-NEXT: [[TMP56:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP48]] 838; VF-FOUR-CHECK-NEXT: [[TMP57:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP49]] 839; VF-FOUR-CHECK-NEXT: [[TMP58:%.*]] = getelementptr inbounds float, float* [[TMP50]], i32 0 840; VF-FOUR-CHECK-NEXT: [[TMP59:%.*]] = getelementptr inbounds float, float* [[TMP58]], i32 -3 841; VF-FOUR-CHECK-NEXT: [[TMP60:%.*]] = bitcast float* [[TMP59]] to <4 x float>* 842; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, <4 x float>* [[TMP60]], align 4 843; VF-FOUR-CHECK-NEXT: [[REVERSE:%.*]] = shufflevector <4 x float> [[WIDE_LOAD]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 844; VF-FOUR-CHECK-NEXT: [[TMP61:%.*]] = getelementptr inbounds float, float* [[TMP50]], i32 -4 845; VF-FOUR-CHECK-NEXT: [[TMP62:%.*]] = getelementptr inbounds float, float* [[TMP61]], i32 -3 846; VF-FOUR-CHECK-NEXT: [[TMP63:%.*]] = bitcast float* [[TMP62]] to <4 x float>* 847; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD2:%.*]] = load <4 x float>, <4 x float>* [[TMP63]], align 4 848; VF-FOUR-CHECK-NEXT: [[REVERSE3:%.*]] = shufflevector <4 x float> [[WIDE_LOAD2]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 849; VF-FOUR-CHECK-NEXT: [[TMP64:%.*]] = getelementptr inbounds float, float* [[TMP50]], i32 -8 850; VF-FOUR-CHECK-NEXT: [[TMP65:%.*]] = getelementptr inbounds float, float* [[TMP64]], i32 -3 851; VF-FOUR-CHECK-NEXT: [[TMP66:%.*]] = bitcast float* [[TMP65]] to <4 x float>* 852; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD4:%.*]] = load <4 x float>, <4 x float>* [[TMP66]], align 4 853; VF-FOUR-CHECK-NEXT: [[REVERSE5:%.*]] = shufflevector <4 x float> [[WIDE_LOAD4]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 854; VF-FOUR-CHECK-NEXT: [[TMP67:%.*]] = getelementptr inbounds float, float* [[TMP50]], i32 -12 855; VF-FOUR-CHECK-NEXT: [[TMP68:%.*]] = getelementptr inbounds float, float* [[TMP67]], i32 -3 856; VF-FOUR-CHECK-NEXT: [[TMP69:%.*]] = bitcast float* [[TMP68]] to <4 x float>* 857; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD6:%.*]] = load <4 x float>, <4 x float>* [[TMP69]], align 4 858; VF-FOUR-CHECK-NEXT: [[REVERSE7:%.*]] = shufflevector <4 x float> [[WIDE_LOAD6]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 859; VF-FOUR-CHECK-NEXT: [[TMP70:%.*]] = getelementptr inbounds float, float* [[TMP50]], i32 -16 860; VF-FOUR-CHECK-NEXT: [[TMP71:%.*]] = getelementptr inbounds float, float* [[TMP70]], i32 -3 861; VF-FOUR-CHECK-NEXT: [[TMP72:%.*]] = bitcast float* [[TMP71]] to <4 x float>* 862; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD8:%.*]] = load <4 x float>, <4 x float>* [[TMP72]], align 4 863; VF-FOUR-CHECK-NEXT: [[REVERSE9:%.*]] = shufflevector <4 x float> [[WIDE_LOAD8]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 864; VF-FOUR-CHECK-NEXT: [[TMP73:%.*]] = getelementptr inbounds float, float* [[TMP50]], i32 -20 865; VF-FOUR-CHECK-NEXT: [[TMP74:%.*]] = getelementptr inbounds float, float* [[TMP73]], i32 -3 866; VF-FOUR-CHECK-NEXT: [[TMP75:%.*]] = bitcast float* [[TMP74]] to <4 x float>* 867; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD10:%.*]] = load <4 x float>, <4 x float>* [[TMP75]], align 4 868; VF-FOUR-CHECK-NEXT: [[REVERSE11:%.*]] = shufflevector <4 x float> [[WIDE_LOAD10]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 869; VF-FOUR-CHECK-NEXT: [[TMP76:%.*]] = getelementptr inbounds float, float* [[TMP50]], i32 -24 870; VF-FOUR-CHECK-NEXT: [[TMP77:%.*]] = getelementptr inbounds float, float* [[TMP76]], i32 -3 871; VF-FOUR-CHECK-NEXT: [[TMP78:%.*]] = bitcast float* [[TMP77]] to <4 x float>* 872; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD12:%.*]] = load <4 x float>, <4 x float>* [[TMP78]], align 4 873; VF-FOUR-CHECK-NEXT: [[REVERSE13:%.*]] = shufflevector <4 x float> [[WIDE_LOAD12]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 874; VF-FOUR-CHECK-NEXT: [[TMP79:%.*]] = getelementptr inbounds float, float* [[TMP50]], i32 -28 875; VF-FOUR-CHECK-NEXT: [[TMP80:%.*]] = getelementptr inbounds float, float* [[TMP79]], i32 -3 876; VF-FOUR-CHECK-NEXT: [[TMP81:%.*]] = bitcast float* [[TMP80]] to <4 x float>* 877; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD14:%.*]] = load <4 x float>, <4 x float>* [[TMP81]], align 4 878; VF-FOUR-CHECK-NEXT: [[REVERSE15:%.*]] = shufflevector <4 x float> [[WIDE_LOAD14]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 879; VF-FOUR-CHECK-NEXT: [[TMP82:%.*]] = fadd fast <4 x float> [[REVERSE]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 880; VF-FOUR-CHECK-NEXT: [[TMP83:%.*]] = fadd fast <4 x float> [[REVERSE3]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 881; VF-FOUR-CHECK-NEXT: [[TMP84:%.*]] = fadd fast <4 x float> [[REVERSE5]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 882; VF-FOUR-CHECK-NEXT: [[TMP85:%.*]] = fadd fast <4 x float> [[REVERSE7]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 883; VF-FOUR-CHECK-NEXT: [[TMP86:%.*]] = fadd fast <4 x float> [[REVERSE9]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 884; VF-FOUR-CHECK-NEXT: [[TMP87:%.*]] = fadd fast <4 x float> [[REVERSE11]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 885; VF-FOUR-CHECK-NEXT: [[TMP88:%.*]] = fadd fast <4 x float> [[REVERSE13]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 886; VF-FOUR-CHECK-NEXT: [[TMP89:%.*]] = fadd fast <4 x float> [[REVERSE15]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 887; VF-FOUR-CHECK-NEXT: [[TMP90:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[TMP10]] 888; VF-FOUR-CHECK-NEXT: [[TMP91:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP11]] 889; VF-FOUR-CHECK-NEXT: [[TMP92:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP12]] 890; VF-FOUR-CHECK-NEXT: [[TMP93:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP13]] 891; VF-FOUR-CHECK-NEXT: [[TMP94:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP14]] 892; VF-FOUR-CHECK-NEXT: [[TMP95:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP15]] 893; VF-FOUR-CHECK-NEXT: [[TMP96:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP16]] 894; VF-FOUR-CHECK-NEXT: [[TMP97:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP17]] 895; VF-FOUR-CHECK-NEXT: [[TMP98:%.*]] = getelementptr inbounds float, float* [[TMP90]], i32 0 896; VF-FOUR-CHECK-NEXT: [[TMP99:%.*]] = bitcast float* [[TMP98]] to <4 x float>* 897; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP82]], <4 x float>* [[TMP99]], align 4 898; VF-FOUR-CHECK-NEXT: [[TMP100:%.*]] = getelementptr inbounds float, float* [[TMP90]], i32 4 899; VF-FOUR-CHECK-NEXT: [[TMP101:%.*]] = bitcast float* [[TMP100]] to <4 x float>* 900; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP83]], <4 x float>* [[TMP101]], align 4 901; VF-FOUR-CHECK-NEXT: [[TMP102:%.*]] = getelementptr inbounds float, float* [[TMP90]], i32 8 902; VF-FOUR-CHECK-NEXT: [[TMP103:%.*]] = bitcast float* [[TMP102]] to <4 x float>* 903; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP84]], <4 x float>* [[TMP103]], align 4 904; VF-FOUR-CHECK-NEXT: [[TMP104:%.*]] = getelementptr inbounds float, float* [[TMP90]], i32 12 905; VF-FOUR-CHECK-NEXT: [[TMP105:%.*]] = bitcast float* [[TMP104]] to <4 x float>* 906; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP85]], <4 x float>* [[TMP105]], align 4 907; VF-FOUR-CHECK-NEXT: [[TMP106:%.*]] = getelementptr inbounds float, float* [[TMP90]], i32 16 908; VF-FOUR-CHECK-NEXT: [[TMP107:%.*]] = bitcast float* [[TMP106]] to <4 x float>* 909; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP86]], <4 x float>* [[TMP107]], align 4 910; VF-FOUR-CHECK-NEXT: [[TMP108:%.*]] = getelementptr inbounds float, float* [[TMP90]], i32 20 911; VF-FOUR-CHECK-NEXT: [[TMP109:%.*]] = bitcast float* [[TMP108]] to <4 x float>* 912; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP87]], <4 x float>* [[TMP109]], align 4 913; VF-FOUR-CHECK-NEXT: [[TMP110:%.*]] = getelementptr inbounds float, float* [[TMP90]], i32 24 914; VF-FOUR-CHECK-NEXT: [[TMP111:%.*]] = bitcast float* [[TMP110]] to <4 x float>* 915; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP88]], <4 x float>* [[TMP111]], align 4 916; VF-FOUR-CHECK-NEXT: [[TMP112:%.*]] = getelementptr inbounds float, float* [[TMP90]], i32 28 917; VF-FOUR-CHECK-NEXT: [[TMP113:%.*]] = bitcast float* [[TMP112]] to <4 x float>* 918; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP89]], <4 x float>* [[TMP113]], align 4 919; VF-FOUR-CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 32 920; VF-FOUR-CHECK-NEXT: [[TMP114:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] 921; VF-FOUR-CHECK-NEXT: br i1 [[TMP114]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]] 922; VF-FOUR-CHECK: middle.block: 923; VF-FOUR-CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC]] 924; VF-FOUR-CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]] 925; VF-FOUR-CHECK: vec.epilog.iter.check: 926; VF-FOUR-CHECK-NEXT: [[IND_END21:%.*]] = trunc i64 [[N_VEC]] to i32 927; VF-FOUR-CHECK-NEXT: [[N_VEC_REMAINING:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_VEC]] 928; VF-FOUR-CHECK-NEXT: [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 4 929; VF-FOUR-CHECK-NEXT: br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]] 930; VF-FOUR-CHECK: vec.epilog.ph: 931; VF-FOUR-CHECK-NEXT: [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ] 932; VF-FOUR-CHECK-NEXT: [[N_MOD_VF16:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 4 933; VF-FOUR-CHECK-NEXT: [[N_VEC17:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF16]] 934; VF-FOUR-CHECK-NEXT: [[IND_END:%.*]] = trunc i64 [[N_VEC17]] to i32 935; VF-FOUR-CHECK-NEXT: br label [[VEC_EPILOG_VECTOR_BODY:%.*]] 936; VF-FOUR-CHECK: vec.epilog.vector.body: 937; VF-FOUR-CHECK-NEXT: [[INDEX18:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT19:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ] 938; VF-FOUR-CHECK-NEXT: [[TMP115:%.*]] = add i64 [[INDEX18]], 0 939; VF-FOUR-CHECK-NEXT: [[OFFSET_IDX23:%.*]] = trunc i64 [[INDEX18]] to i32 940; VF-FOUR-CHECK-NEXT: [[TMP116:%.*]] = add i32 [[OFFSET_IDX23]], 0 941; VF-FOUR-CHECK-NEXT: [[TMP117:%.*]] = xor i32 [[TMP116]], -1 942; VF-FOUR-CHECK-NEXT: [[TMP118:%.*]] = add i32 [[TMP117]], [[N]] 943; VF-FOUR-CHECK-NEXT: [[TMP119:%.*]] = sext i32 [[TMP118]] to i64 944; VF-FOUR-CHECK-NEXT: [[TMP120:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP119]] 945; VF-FOUR-CHECK-NEXT: [[TMP121:%.*]] = getelementptr inbounds float, float* [[TMP120]], i32 0 946; VF-FOUR-CHECK-NEXT: [[TMP122:%.*]] = getelementptr inbounds float, float* [[TMP121]], i32 -3 947; VF-FOUR-CHECK-NEXT: [[TMP123:%.*]] = bitcast float* [[TMP122]] to <4 x float>* 948; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD24:%.*]] = load <4 x float>, <4 x float>* [[TMP123]], align 4 949; VF-FOUR-CHECK-NEXT: [[REVERSE25:%.*]] = shufflevector <4 x float> [[WIDE_LOAD24]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 950; VF-FOUR-CHECK-NEXT: [[TMP124:%.*]] = fadd fast <4 x float> [[REVERSE25]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 951; VF-FOUR-CHECK-NEXT: [[TMP125:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP115]] 952; VF-FOUR-CHECK-NEXT: [[TMP126:%.*]] = getelementptr inbounds float, float* [[TMP125]], i32 0 953; VF-FOUR-CHECK-NEXT: [[TMP127:%.*]] = bitcast float* [[TMP126]] to <4 x float>* 954; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP124]], <4 x float>* [[TMP127]], align 4 955; VF-FOUR-CHECK-NEXT: [[INDEX_NEXT19]] = add nuw i64 [[INDEX18]], 4 956; VF-FOUR-CHECK-NEXT: [[TMP128:%.*]] = icmp eq i64 [[INDEX_NEXT19]], [[N_VEC17]] 957; VF-FOUR-CHECK-NEXT: br i1 [[TMP128]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]] 958; VF-FOUR-CHECK: vec.epilog.middle.block: 959; VF-FOUR-CHECK-NEXT: [[CMP_N22:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC17]] 960; VF-FOUR-CHECK-NEXT: br i1 [[CMP_N22]], label [[FOR_END_LOOPEXIT_LOOPEXIT:%.*]], label [[VEC_EPILOG_SCALAR_PH]] 961; VF-FOUR-CHECK: vec.epilog.scalar.ph: 962; VF-FOUR-CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC17]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ITER_CHECK]] ] 963; VF-FOUR-CHECK-NEXT: [[BC_RESUME_VAL20:%.*]] = phi i32 [ [[IND_END]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[IND_END21]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ITER_CHECK]] ] 964; VF-FOUR-CHECK-NEXT: br label [[FOR_BODY:%.*]] 965; VF-FOUR-CHECK: for.body: 966; VF-FOUR-CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ] 967; VF-FOUR-CHECK-NEXT: [[I_014:%.*]] = phi i32 [ [[BC_RESUME_VAL20]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ] 968; VF-FOUR-CHECK-NEXT: [[TMP129:%.*]] = xor i32 [[I_014]], -1 969; VF-FOUR-CHECK-NEXT: [[SUB2:%.*]] = add i32 [[TMP129]], [[N]] 970; VF-FOUR-CHECK-NEXT: [[IDXPROM:%.*]] = sext i32 [[SUB2]] to i64 971; VF-FOUR-CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[IDXPROM]] 972; VF-FOUR-CHECK-NEXT: [[TMP130:%.*]] = load float, float* [[ARRAYIDX]], align 4 973; VF-FOUR-CHECK-NEXT: [[CONV3:%.*]] = fadd fast float [[TMP130]], 1.000000e+00 974; VF-FOUR-CHECK-NEXT: [[ARRAYIDX5:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDVARS_IV]] 975; VF-FOUR-CHECK-NEXT: store float [[CONV3]], float* [[ARRAYIDX5]], align 4 976; VF-FOUR-CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1 977; VF-FOUR-CHECK-NEXT: [[INC]] = add nuw nsw i32 [[I_014]], 1 978; VF-FOUR-CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i64 [[INDVARS_IV_NEXT]], [[WIDE_TRIP_COUNT]] 979; VF-FOUR-CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_BODY]], label [[FOR_END_LOOPEXIT_LOOPEXIT]], !llvm.loop [[LOOP7:![0-9]+]] 980; VF-FOUR-CHECK: for.end.loopexit.loopexit: 981; VF-FOUR-CHECK-NEXT: br label [[FOR_END_LOOPEXIT]] 982; VF-FOUR-CHECK: for.end.loopexit: 983; VF-FOUR-CHECK-NEXT: br label [[FOR_END]] 984; VF-FOUR-CHECK: for.end: 985; VF-FOUR-CHECK-NEXT: ret i32 0 986; 987 988 989entry: 990 %cmp1 = icmp sgt i32 %n, 1 991 br i1 %cmp1, label %for.body.preheader, label %for.end 992 993for.body.preheader: ; preds = %entry 994 %0 = add i32 %n, -1 995 %wide.trip.count = zext i32 %0 to i64 996 br label %for.body 997 998for.body: ; preds = %for.body.preheader, %for.body 999 %indvars.iv = phi i64 [ 0, %for.body.preheader ], [ %indvars.iv.next, %for.body ] 1000 %i.014 = phi i32 [ 0, %for.body.preheader ], [ %inc, %for.body ] 1001 %1 = xor i32 %i.014, -1 1002 %sub2 = add i32 %1, %n 1003 %idxprom = sext i32 %sub2 to i64 1004 %arrayidx = getelementptr inbounds float, float* %B, i64 %idxprom 1005 %2 = load float, float* %arrayidx, align 4 1006 %conv3 = fadd fast float %2, 1.000000e+00 1007 %arrayidx5 = getelementptr inbounds float, float* %A, i64 %indvars.iv 1008 store float %conv3, float* %arrayidx5, align 4 1009 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 1010 %inc = add nuw nsw i32 %i.014, 1 1011 %exitcond = icmp ne i64 %indvars.iv.next, %wide.trip.count 1012 br i1 %exitcond, label %for.body, label %for.end.loopexit 1013 1014for.end.loopexit: ; preds = %for.body 1015 br label %for.end 1016 1017for.end: ; preds = %for.end.loopexit, %entry 1018 ret i32 0 1019} 1020 1021attributes #0 = { nounwind "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "frame-pointer"="none" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-infs-fp-math"="true" "no-jump-tables"="false" "no-nans-fp-math"="true" "no-signed-zeros-fp-math"="true" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="ppc64le" "target-features"="+altivec,+bpermd,+crypto,+direct-move,+extdiv,+htm,+power8-vector,+vsx,-power9-vector,-spe" "unsafe-fp-math"="true" "use-soft-float"="false" } 1022