1; RUN: opt < %s -passes='loop-vectorize' -enable-epilogue-vectorization -epilogue-vectorization-force-VF=2 -S | FileCheck %s --check-prefix VF-TWO-CHECK 2; RUN: opt < %s -passes='loop-vectorize' -enable-epilogue-vectorization -epilogue-vectorization-force-VF=4 -S | FileCheck %s --check-prefix VF-FOUR-CHECK 3 4target datalayout = "e-m:e-i64:64-n32:64" 5target triple = "powerpc64le-unknown-linux-gnu" 6 7; Function Attrs: nounwind 8define dso_local void @f1(float* noalias %aa, float* noalias %bb, float* noalias %cc, i32 signext %N) #0 { 9; VF-TWO-CHECK-LABEL: @f1( 10; VF-TWO-CHECK-NEXT: entry: 11; VF-TWO-CHECK-NEXT: [[CMP1:%.*]] = icmp sgt i32 [[N:%.*]], 0 12; VF-TWO-CHECK-NEXT: br i1 [[CMP1]], label [[ITER_CHECK:%.*]], label [[FOR_END:%.*]] 13; VF-TWO-CHECK: iter.check: 14; VF-TWO-CHECK-NEXT: [[WIDE_TRIP_COUNT:%.*]] = zext i32 [[N]] to i64 15; VF-TWO-CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 2 16; VF-TWO-CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]] 17; VF-TWO-CHECK: vector.main.loop.iter.check: 18; VF-TWO-CHECK-NEXT: [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 48 19; VF-TWO-CHECK-NEXT: br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]] 20; VF-TWO-CHECK: vector.ph: 21; VF-TWO-CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 48 22; VF-TWO-CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF]] 23; VF-TWO-CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 24; VF-TWO-CHECK: vector.body: 25; VF-TWO-CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 26; VF-TWO-CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0 27; VF-TWO-CHECK-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 4 28; VF-TWO-CHECK-NEXT: [[TMP2:%.*]] = add i64 [[INDEX]], 8 29; VF-TWO-CHECK-NEXT: [[TMP3:%.*]] = add i64 [[INDEX]], 12 30; VF-TWO-CHECK-NEXT: [[TMP4:%.*]] = add i64 [[INDEX]], 16 31; VF-TWO-CHECK-NEXT: [[TMP5:%.*]] = add i64 [[INDEX]], 20 32; VF-TWO-CHECK-NEXT: [[TMP6:%.*]] = add i64 [[INDEX]], 24 33; VF-TWO-CHECK-NEXT: [[TMP7:%.*]] = add i64 [[INDEX]], 28 34; VF-TWO-CHECK-NEXT: [[TMP8:%.*]] = add i64 [[INDEX]], 32 35; VF-TWO-CHECK-NEXT: [[TMP9:%.*]] = add i64 [[INDEX]], 36 36; VF-TWO-CHECK-NEXT: [[TMP10:%.*]] = add i64 [[INDEX]], 40 37; VF-TWO-CHECK-NEXT: [[TMP11:%.*]] = add i64 [[INDEX]], 44 38; VF-TWO-CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds float, float* [[BB:%.*]], i64 [[TMP0]] 39; VF-TWO-CHECK-NEXT: [[TMP13:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP1]] 40; VF-TWO-CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP2]] 41; VF-TWO-CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP3]] 42; VF-TWO-CHECK-NEXT: [[TMP16:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP4]] 43; VF-TWO-CHECK-NEXT: [[TMP17:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP5]] 44; VF-TWO-CHECK-NEXT: [[TMP18:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP6]] 45; VF-TWO-CHECK-NEXT: [[TMP19:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP7]] 46; VF-TWO-CHECK-NEXT: [[TMP20:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP8]] 47; VF-TWO-CHECK-NEXT: [[TMP21:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP9]] 48; VF-TWO-CHECK-NEXT: [[TMP22:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP10]] 49; VF-TWO-CHECK-NEXT: [[TMP23:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP11]] 50; VF-TWO-CHECK-NEXT: [[TMP24:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 0 51; VF-TWO-CHECK-NEXT: [[TMP25:%.*]] = bitcast float* [[TMP24]] to <4 x float>* 52; VF-TWO-CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, <4 x float>* [[TMP25]], align 4 53; VF-TWO-CHECK-NEXT: [[TMP26:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 4 54; VF-TWO-CHECK-NEXT: [[TMP27:%.*]] = bitcast float* [[TMP26]] to <4 x float>* 55; VF-TWO-CHECK-NEXT: [[WIDE_LOAD2:%.*]] = load <4 x float>, <4 x float>* [[TMP27]], align 4 56; VF-TWO-CHECK-NEXT: [[TMP28:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 8 57; VF-TWO-CHECK-NEXT: [[TMP29:%.*]] = bitcast float* [[TMP28]] to <4 x float>* 58; VF-TWO-CHECK-NEXT: [[WIDE_LOAD3:%.*]] = load <4 x float>, <4 x float>* [[TMP29]], align 4 59; VF-TWO-CHECK-NEXT: [[TMP30:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 12 60; VF-TWO-CHECK-NEXT: [[TMP31:%.*]] = bitcast float* [[TMP30]] to <4 x float>* 61; VF-TWO-CHECK-NEXT: [[WIDE_LOAD4:%.*]] = load <4 x float>, <4 x float>* [[TMP31]], align 4 62; VF-TWO-CHECK-NEXT: [[TMP32:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 16 63; VF-TWO-CHECK-NEXT: [[TMP33:%.*]] = bitcast float* [[TMP32]] to <4 x float>* 64; VF-TWO-CHECK-NEXT: [[WIDE_LOAD5:%.*]] = load <4 x float>, <4 x float>* [[TMP33]], align 4 65; VF-TWO-CHECK-NEXT: [[TMP34:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 20 66; VF-TWO-CHECK-NEXT: [[TMP35:%.*]] = bitcast float* [[TMP34]] to <4 x float>* 67; VF-TWO-CHECK-NEXT: [[WIDE_LOAD6:%.*]] = load <4 x float>, <4 x float>* [[TMP35]], align 4 68; VF-TWO-CHECK-NEXT: [[TMP36:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 24 69; VF-TWO-CHECK-NEXT: [[TMP37:%.*]] = bitcast float* [[TMP36]] to <4 x float>* 70; VF-TWO-CHECK-NEXT: [[WIDE_LOAD7:%.*]] = load <4 x float>, <4 x float>* [[TMP37]], align 4 71; VF-TWO-CHECK-NEXT: [[TMP38:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 28 72; VF-TWO-CHECK-NEXT: [[TMP39:%.*]] = bitcast float* [[TMP38]] to <4 x float>* 73; VF-TWO-CHECK-NEXT: [[WIDE_LOAD8:%.*]] = load <4 x float>, <4 x float>* [[TMP39]], align 4 74; VF-TWO-CHECK-NEXT: [[TMP40:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 32 75; VF-TWO-CHECK-NEXT: [[TMP41:%.*]] = bitcast float* [[TMP40]] to <4 x float>* 76; VF-TWO-CHECK-NEXT: [[WIDE_LOAD9:%.*]] = load <4 x float>, <4 x float>* [[TMP41]], align 4 77; VF-TWO-CHECK-NEXT: [[TMP42:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 36 78; VF-TWO-CHECK-NEXT: [[TMP43:%.*]] = bitcast float* [[TMP42]] to <4 x float>* 79; VF-TWO-CHECK-NEXT: [[WIDE_LOAD10:%.*]] = load <4 x float>, <4 x float>* [[TMP43]], align 4 80; VF-TWO-CHECK-NEXT: [[TMP44:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 40 81; VF-TWO-CHECK-NEXT: [[TMP45:%.*]] = bitcast float* [[TMP44]] to <4 x float>* 82; VF-TWO-CHECK-NEXT: [[WIDE_LOAD11:%.*]] = load <4 x float>, <4 x float>* [[TMP45]], align 4 83; VF-TWO-CHECK-NEXT: [[TMP46:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 44 84; VF-TWO-CHECK-NEXT: [[TMP47:%.*]] = bitcast float* [[TMP46]] to <4 x float>* 85; VF-TWO-CHECK-NEXT: [[WIDE_LOAD12:%.*]] = load <4 x float>, <4 x float>* [[TMP47]], align 4 86; VF-TWO-CHECK-NEXT: [[TMP48:%.*]] = getelementptr inbounds float, float* [[CC:%.*]], i64 [[TMP0]] 87; VF-TWO-CHECK-NEXT: [[TMP49:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP1]] 88; VF-TWO-CHECK-NEXT: [[TMP50:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP2]] 89; VF-TWO-CHECK-NEXT: [[TMP51:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP3]] 90; VF-TWO-CHECK-NEXT: [[TMP52:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP4]] 91; VF-TWO-CHECK-NEXT: [[TMP53:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP5]] 92; VF-TWO-CHECK-NEXT: [[TMP54:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP6]] 93; VF-TWO-CHECK-NEXT: [[TMP55:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP7]] 94; VF-TWO-CHECK-NEXT: [[TMP56:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP8]] 95; VF-TWO-CHECK-NEXT: [[TMP57:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP9]] 96; VF-TWO-CHECK-NEXT: [[TMP58:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP10]] 97; VF-TWO-CHECK-NEXT: [[TMP59:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP11]] 98; VF-TWO-CHECK-NEXT: [[TMP60:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 0 99; VF-TWO-CHECK-NEXT: [[TMP61:%.*]] = bitcast float* [[TMP60]] to <4 x float>* 100; VF-TWO-CHECK-NEXT: [[WIDE_LOAD13:%.*]] = load <4 x float>, <4 x float>* [[TMP61]], align 4 101; VF-TWO-CHECK-NEXT: [[TMP62:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 4 102; VF-TWO-CHECK-NEXT: [[TMP63:%.*]] = bitcast float* [[TMP62]] to <4 x float>* 103; VF-TWO-CHECK-NEXT: [[WIDE_LOAD14:%.*]] = load <4 x float>, <4 x float>* [[TMP63]], align 4 104; VF-TWO-CHECK-NEXT: [[TMP64:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 8 105; VF-TWO-CHECK-NEXT: [[TMP65:%.*]] = bitcast float* [[TMP64]] to <4 x float>* 106; VF-TWO-CHECK-NEXT: [[WIDE_LOAD15:%.*]] = load <4 x float>, <4 x float>* [[TMP65]], align 4 107; VF-TWO-CHECK-NEXT: [[TMP66:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 12 108; VF-TWO-CHECK-NEXT: [[TMP67:%.*]] = bitcast float* [[TMP66]] to <4 x float>* 109; VF-TWO-CHECK-NEXT: [[WIDE_LOAD16:%.*]] = load <4 x float>, <4 x float>* [[TMP67]], align 4 110; VF-TWO-CHECK-NEXT: [[TMP68:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 16 111; VF-TWO-CHECK-NEXT: [[TMP69:%.*]] = bitcast float* [[TMP68]] to <4 x float>* 112; VF-TWO-CHECK-NEXT: [[WIDE_LOAD17:%.*]] = load <4 x float>, <4 x float>* [[TMP69]], align 4 113; VF-TWO-CHECK-NEXT: [[TMP70:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 20 114; VF-TWO-CHECK-NEXT: [[TMP71:%.*]] = bitcast float* [[TMP70]] to <4 x float>* 115; VF-TWO-CHECK-NEXT: [[WIDE_LOAD18:%.*]] = load <4 x float>, <4 x float>* [[TMP71]], align 4 116; VF-TWO-CHECK-NEXT: [[TMP72:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 24 117; VF-TWO-CHECK-NEXT: [[TMP73:%.*]] = bitcast float* [[TMP72]] to <4 x float>* 118; VF-TWO-CHECK-NEXT: [[WIDE_LOAD19:%.*]] = load <4 x float>, <4 x float>* [[TMP73]], align 4 119; VF-TWO-CHECK-NEXT: [[TMP74:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 28 120; VF-TWO-CHECK-NEXT: [[TMP75:%.*]] = bitcast float* [[TMP74]] to <4 x float>* 121; VF-TWO-CHECK-NEXT: [[WIDE_LOAD20:%.*]] = load <4 x float>, <4 x float>* [[TMP75]], align 4 122; VF-TWO-CHECK-NEXT: [[TMP76:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 32 123; VF-TWO-CHECK-NEXT: [[TMP77:%.*]] = bitcast float* [[TMP76]] to <4 x float>* 124; VF-TWO-CHECK-NEXT: [[WIDE_LOAD21:%.*]] = load <4 x float>, <4 x float>* [[TMP77]], align 4 125; VF-TWO-CHECK-NEXT: [[TMP78:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 36 126; VF-TWO-CHECK-NEXT: [[TMP79:%.*]] = bitcast float* [[TMP78]] to <4 x float>* 127; VF-TWO-CHECK-NEXT: [[WIDE_LOAD22:%.*]] = load <4 x float>, <4 x float>* [[TMP79]], align 4 128; VF-TWO-CHECK-NEXT: [[TMP80:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 40 129; VF-TWO-CHECK-NEXT: [[TMP81:%.*]] = bitcast float* [[TMP80]] to <4 x float>* 130; VF-TWO-CHECK-NEXT: [[WIDE_LOAD23:%.*]] = load <4 x float>, <4 x float>* [[TMP81]], align 4 131; VF-TWO-CHECK-NEXT: [[TMP82:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 44 132; VF-TWO-CHECK-NEXT: [[TMP83:%.*]] = bitcast float* [[TMP82]] to <4 x float>* 133; VF-TWO-CHECK-NEXT: [[WIDE_LOAD24:%.*]] = load <4 x float>, <4 x float>* [[TMP83]], align 4 134; VF-TWO-CHECK-NEXT: [[TMP84:%.*]] = fadd fast <4 x float> [[WIDE_LOAD]], [[WIDE_LOAD13]] 135; VF-TWO-CHECK-NEXT: [[TMP85:%.*]] = fadd fast <4 x float> [[WIDE_LOAD2]], [[WIDE_LOAD14]] 136; VF-TWO-CHECK-NEXT: [[TMP86:%.*]] = fadd fast <4 x float> [[WIDE_LOAD3]], [[WIDE_LOAD15]] 137; VF-TWO-CHECK-NEXT: [[TMP87:%.*]] = fadd fast <4 x float> [[WIDE_LOAD4]], [[WIDE_LOAD16]] 138; VF-TWO-CHECK-NEXT: [[TMP88:%.*]] = fadd fast <4 x float> [[WIDE_LOAD5]], [[WIDE_LOAD17]] 139; VF-TWO-CHECK-NEXT: [[TMP89:%.*]] = fadd fast <4 x float> [[WIDE_LOAD6]], [[WIDE_LOAD18]] 140; VF-TWO-CHECK-NEXT: [[TMP90:%.*]] = fadd fast <4 x float> [[WIDE_LOAD7]], [[WIDE_LOAD19]] 141; VF-TWO-CHECK-NEXT: [[TMP91:%.*]] = fadd fast <4 x float> [[WIDE_LOAD8]], [[WIDE_LOAD20]] 142; VF-TWO-CHECK-NEXT: [[TMP92:%.*]] = fadd fast <4 x float> [[WIDE_LOAD9]], [[WIDE_LOAD21]] 143; VF-TWO-CHECK-NEXT: [[TMP93:%.*]] = fadd fast <4 x float> [[WIDE_LOAD10]], [[WIDE_LOAD22]] 144; VF-TWO-CHECK-NEXT: [[TMP94:%.*]] = fadd fast <4 x float> [[WIDE_LOAD11]], [[WIDE_LOAD23]] 145; VF-TWO-CHECK-NEXT: [[TMP95:%.*]] = fadd fast <4 x float> [[WIDE_LOAD12]], [[WIDE_LOAD24]] 146; VF-TWO-CHECK-NEXT: [[TMP96:%.*]] = getelementptr inbounds float, float* [[AA:%.*]], i64 [[TMP0]] 147; VF-TWO-CHECK-NEXT: [[TMP97:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP1]] 148; VF-TWO-CHECK-NEXT: [[TMP98:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP2]] 149; VF-TWO-CHECK-NEXT: [[TMP99:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP3]] 150; VF-TWO-CHECK-NEXT: [[TMP100:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP4]] 151; VF-TWO-CHECK-NEXT: [[TMP101:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP5]] 152; VF-TWO-CHECK-NEXT: [[TMP102:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP6]] 153; VF-TWO-CHECK-NEXT: [[TMP103:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP7]] 154; VF-TWO-CHECK-NEXT: [[TMP104:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP8]] 155; VF-TWO-CHECK-NEXT: [[TMP105:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP9]] 156; VF-TWO-CHECK-NEXT: [[TMP106:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP10]] 157; VF-TWO-CHECK-NEXT: [[TMP107:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP11]] 158; VF-TWO-CHECK-NEXT: [[TMP108:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 0 159; VF-TWO-CHECK-NEXT: [[TMP109:%.*]] = bitcast float* [[TMP108]] to <4 x float>* 160; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP84]], <4 x float>* [[TMP109]], align 4 161; VF-TWO-CHECK-NEXT: [[TMP110:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 4 162; VF-TWO-CHECK-NEXT: [[TMP111:%.*]] = bitcast float* [[TMP110]] to <4 x float>* 163; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP85]], <4 x float>* [[TMP111]], align 4 164; VF-TWO-CHECK-NEXT: [[TMP112:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 8 165; VF-TWO-CHECK-NEXT: [[TMP113:%.*]] = bitcast float* [[TMP112]] to <4 x float>* 166; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP86]], <4 x float>* [[TMP113]], align 4 167; VF-TWO-CHECK-NEXT: [[TMP114:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 12 168; VF-TWO-CHECK-NEXT: [[TMP115:%.*]] = bitcast float* [[TMP114]] to <4 x float>* 169; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP87]], <4 x float>* [[TMP115]], align 4 170; VF-TWO-CHECK-NEXT: [[TMP116:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 16 171; VF-TWO-CHECK-NEXT: [[TMP117:%.*]] = bitcast float* [[TMP116]] to <4 x float>* 172; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP88]], <4 x float>* [[TMP117]], align 4 173; VF-TWO-CHECK-NEXT: [[TMP118:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 20 174; VF-TWO-CHECK-NEXT: [[TMP119:%.*]] = bitcast float* [[TMP118]] to <4 x float>* 175; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP89]], <4 x float>* [[TMP119]], align 4 176; VF-TWO-CHECK-NEXT: [[TMP120:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 24 177; VF-TWO-CHECK-NEXT: [[TMP121:%.*]] = bitcast float* [[TMP120]] to <4 x float>* 178; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP90]], <4 x float>* [[TMP121]], align 4 179; VF-TWO-CHECK-NEXT: [[TMP122:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 28 180; VF-TWO-CHECK-NEXT: [[TMP123:%.*]] = bitcast float* [[TMP122]] to <4 x float>* 181; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP91]], <4 x float>* [[TMP123]], align 4 182; VF-TWO-CHECK-NEXT: [[TMP124:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 32 183; VF-TWO-CHECK-NEXT: [[TMP125:%.*]] = bitcast float* [[TMP124]] to <4 x float>* 184; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP92]], <4 x float>* [[TMP125]], align 4 185; VF-TWO-CHECK-NEXT: [[TMP126:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 36 186; VF-TWO-CHECK-NEXT: [[TMP127:%.*]] = bitcast float* [[TMP126]] to <4 x float>* 187; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP93]], <4 x float>* [[TMP127]], align 4 188; VF-TWO-CHECK-NEXT: [[TMP128:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 40 189; VF-TWO-CHECK-NEXT: [[TMP129:%.*]] = bitcast float* [[TMP128]] to <4 x float>* 190; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP94]], <4 x float>* [[TMP129]], align 4 191; VF-TWO-CHECK-NEXT: [[TMP130:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 44 192; VF-TWO-CHECK-NEXT: [[TMP131:%.*]] = bitcast float* [[TMP130]] to <4 x float>* 193; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP95]], <4 x float>* [[TMP131]], align 4 194; VF-TWO-CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 48 195; VF-TWO-CHECK-NEXT: [[TMP132:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] 196; VF-TWO-CHECK-NEXT: br i1 [[TMP132]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOPID_MV:![0-9]+]] 197; VF-TWO-CHECK: middle.block: 198; VF-TWO-CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC]] 199; VF-TWO-CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]] 200; VF-TWO-CHECK: vec.epilog.iter.check: 201; VF-TWO-CHECK-NEXT: [[N_VEC_REMAINING:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_VEC]] 202; VF-TWO-CHECK-NEXT: [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 2 203; VF-TWO-CHECK-NEXT: br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]] 204; VF-TWO-CHECK: vec.epilog.ph: 205; VF-TWO-CHECK-NEXT: [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ] 206; VF-TWO-CHECK-NEXT: [[N_MOD_VF25:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 2 207; VF-TWO-CHECK-NEXT: [[N_VEC26:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF25]] 208; VF-TWO-CHECK-NEXT: br label [[VEC_EPILOG_VECTOR_BODY:%.*]] 209; VF-TWO-CHECK: vec.epilog.vector.body: 210; VF-TWO-CHECK-NEXT: [[OFFSET_IDX:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT31:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ] 211; VF-TWO-CHECK-NEXT: [[TMP133:%.*]] = add i64 [[OFFSET_IDX]], 0 212; VF-TWO-CHECK-NEXT: [[TMP134:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP133]] 213; VF-TWO-CHECK-NEXT: [[TMP135:%.*]] = getelementptr inbounds float, float* [[TMP134]], i32 0 214; VF-TWO-CHECK-NEXT: [[TMP136:%.*]] = bitcast float* [[TMP135]] to <2 x float>* 215; VF-TWO-CHECK-NEXT: [[WIDE_LOAD29:%.*]] = load <2 x float>, <2 x float>* [[TMP136]], align 4 216; VF-TWO-CHECK-NEXT: [[TMP137:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP133]] 217; VF-TWO-CHECK-NEXT: [[TMP138:%.*]] = getelementptr inbounds float, float* [[TMP137]], i32 0 218; VF-TWO-CHECK-NEXT: [[TMP139:%.*]] = bitcast float* [[TMP138]] to <2 x float>* 219; VF-TWO-CHECK-NEXT: [[WIDE_LOAD30:%.*]] = load <2 x float>, <2 x float>* [[TMP139]], align 4 220; VF-TWO-CHECK-NEXT: [[TMP140:%.*]] = fadd fast <2 x float> [[WIDE_LOAD29]], [[WIDE_LOAD30]] 221; VF-TWO-CHECK-NEXT: [[TMP141:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP133]] 222; VF-TWO-CHECK-NEXT: [[TMP142:%.*]] = getelementptr inbounds float, float* [[TMP141]], i32 0 223; VF-TWO-CHECK-NEXT: [[TMP143:%.*]] = bitcast float* [[TMP142]] to <2 x float>* 224; VF-TWO-CHECK-NEXT: store <2 x float> [[TMP140]], <2 x float>* [[TMP143]], align 4 225; VF-TWO-CHECK-NEXT: [[INDEX_NEXT31]] = add nuw i64 [[OFFSET_IDX]], 2 226; VF-TWO-CHECK-NEXT: [[TMP144:%.*]] = icmp eq i64 [[INDEX_NEXT31]], [[N_VEC26]] 227; VF-TWO-CHECK-NEXT: br i1 [[TMP144]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOPID_EV:![0-9]+]] 228; VF-TWO-CHECK: vec.epilog.middle.block: 229; VF-TWO-CHECK-NEXT: [[CMP_N27:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC26]] 230; VF-TWO-CHECK-NEXT: br i1 [[CMP_N27]], label [[FOR_END_LOOPEXIT]], label [[VEC_EPILOG_SCALAR_PH]] 231; VF-TWO-CHECK: vec.epilog.scalar.ph: 232; VF-TWO-CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC26]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK]] ] 233; VF-TWO-CHECK-NEXT: br label [[FOR_BODY:%.*]] 234; VF-TWO-CHECK: for.body: 235; VF-TWO-CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ] 236; VF-TWO-CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[INDVARS_IV]] 237; VF-TWO-CHECK-NEXT: [[TMP145:%.*]] = load float, float* [[ARRAYIDX]], align 4 238; VF-TWO-CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[INDVARS_IV]] 239; VF-TWO-CHECK-NEXT: [[TMP146:%.*]] = load float, float* [[ARRAYIDX2]], align 4 240; VF-TWO-CHECK-NEXT: [[ADD:%.*]] = fadd fast float [[TMP145]], [[TMP146]] 241; VF-TWO-CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[INDVARS_IV]] 242; VF-TWO-CHECK-NEXT: store float [[ADD]], float* [[ARRAYIDX4]], align 4 243; VF-TWO-CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1 244; VF-TWO-CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i64 [[INDVARS_IV_NEXT]], [[WIDE_TRIP_COUNT]] 245; VF-TWO-CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_BODY]], label [[FOR_END_LOOPEXIT]], !llvm.loop [[LOOP4:![0-9]+]] 246; VF-TWO-CHECK: for.end.loopexit: 247; VF-TWO-CHECK-NEXT: br label [[FOR_END]] 248; VF-TWO-CHECK: for.end: 249; VF-TWO-CHECK-NEXT: ret void 250; 251; VF-FOUR-CHECK-LABEL: @f1( 252; VF-FOUR-CHECK-NEXT: entry: 253; VF-FOUR-CHECK-NEXT: [[CMP1:%.*]] = icmp sgt i32 [[N:%.*]], 0 254; VF-FOUR-CHECK-NEXT: br i1 [[CMP1]], label [[ITER_CHECK:%.*]], label [[FOR_END:%.*]] 255; VF-FOUR-CHECK: iter.check: 256; VF-FOUR-CHECK-NEXT: [[WIDE_TRIP_COUNT:%.*]] = zext i32 [[N]] to i64 257; VF-FOUR-CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 4 258; VF-FOUR-CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]] 259; VF-FOUR-CHECK: vector.main.loop.iter.check: 260; VF-FOUR-CHECK-NEXT: [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 48 261; VF-FOUR-CHECK-NEXT: br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]] 262; VF-FOUR-CHECK: vector.ph: 263; VF-FOUR-CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 48 264; VF-FOUR-CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF]] 265; VF-FOUR-CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 266; VF-FOUR-CHECK: vector.body: 267; VF-FOUR-CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 268; VF-FOUR-CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0 269; VF-FOUR-CHECK-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 4 270; VF-FOUR-CHECK-NEXT: [[TMP2:%.*]] = add i64 [[INDEX]], 8 271; VF-FOUR-CHECK-NEXT: [[TMP3:%.*]] = add i64 [[INDEX]], 12 272; VF-FOUR-CHECK-NEXT: [[TMP4:%.*]] = add i64 [[INDEX]], 16 273; VF-FOUR-CHECK-NEXT: [[TMP5:%.*]] = add i64 [[INDEX]], 20 274; VF-FOUR-CHECK-NEXT: [[TMP6:%.*]] = add i64 [[INDEX]], 24 275; VF-FOUR-CHECK-NEXT: [[TMP7:%.*]] = add i64 [[INDEX]], 28 276; VF-FOUR-CHECK-NEXT: [[TMP8:%.*]] = add i64 [[INDEX]], 32 277; VF-FOUR-CHECK-NEXT: [[TMP9:%.*]] = add i64 [[INDEX]], 36 278; VF-FOUR-CHECK-NEXT: [[TMP10:%.*]] = add i64 [[INDEX]], 40 279; VF-FOUR-CHECK-NEXT: [[TMP11:%.*]] = add i64 [[INDEX]], 44 280; VF-FOUR-CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds float, float* [[BB:%.*]], i64 [[TMP0]] 281; VF-FOUR-CHECK-NEXT: [[TMP13:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP1]] 282; VF-FOUR-CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP2]] 283; VF-FOUR-CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP3]] 284; VF-FOUR-CHECK-NEXT: [[TMP16:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP4]] 285; VF-FOUR-CHECK-NEXT: [[TMP17:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP5]] 286; VF-FOUR-CHECK-NEXT: [[TMP18:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP6]] 287; VF-FOUR-CHECK-NEXT: [[TMP19:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP7]] 288; VF-FOUR-CHECK-NEXT: [[TMP20:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP8]] 289; VF-FOUR-CHECK-NEXT: [[TMP21:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP9]] 290; VF-FOUR-CHECK-NEXT: [[TMP22:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP10]] 291; VF-FOUR-CHECK-NEXT: [[TMP23:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP11]] 292; VF-FOUR-CHECK-NEXT: [[TMP24:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 0 293; VF-FOUR-CHECK-NEXT: [[TMP25:%.*]] = bitcast float* [[TMP24]] to <4 x float>* 294; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, <4 x float>* [[TMP25]], align 4 295; VF-FOUR-CHECK-NEXT: [[TMP26:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 4 296; VF-FOUR-CHECK-NEXT: [[TMP27:%.*]] = bitcast float* [[TMP26]] to <4 x float>* 297; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD2:%.*]] = load <4 x float>, <4 x float>* [[TMP27]], align 4 298; VF-FOUR-CHECK-NEXT: [[TMP28:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 8 299; VF-FOUR-CHECK-NEXT: [[TMP29:%.*]] = bitcast float* [[TMP28]] to <4 x float>* 300; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD3:%.*]] = load <4 x float>, <4 x float>* [[TMP29]], align 4 301; VF-FOUR-CHECK-NEXT: [[TMP30:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 12 302; VF-FOUR-CHECK-NEXT: [[TMP31:%.*]] = bitcast float* [[TMP30]] to <4 x float>* 303; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD4:%.*]] = load <4 x float>, <4 x float>* [[TMP31]], align 4 304; VF-FOUR-CHECK-NEXT: [[TMP32:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 16 305; VF-FOUR-CHECK-NEXT: [[TMP33:%.*]] = bitcast float* [[TMP32]] to <4 x float>* 306; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD5:%.*]] = load <4 x float>, <4 x float>* [[TMP33]], align 4 307; VF-FOUR-CHECK-NEXT: [[TMP34:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 20 308; VF-FOUR-CHECK-NEXT: [[TMP35:%.*]] = bitcast float* [[TMP34]] to <4 x float>* 309; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD6:%.*]] = load <4 x float>, <4 x float>* [[TMP35]], align 4 310; VF-FOUR-CHECK-NEXT: [[TMP36:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 24 311; VF-FOUR-CHECK-NEXT: [[TMP37:%.*]] = bitcast float* [[TMP36]] to <4 x float>* 312; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD7:%.*]] = load <4 x float>, <4 x float>* [[TMP37]], align 4 313; VF-FOUR-CHECK-NEXT: [[TMP38:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 28 314; VF-FOUR-CHECK-NEXT: [[TMP39:%.*]] = bitcast float* [[TMP38]] to <4 x float>* 315; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD8:%.*]] = load <4 x float>, <4 x float>* [[TMP39]], align 4 316; VF-FOUR-CHECK-NEXT: [[TMP40:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 32 317; VF-FOUR-CHECK-NEXT: [[TMP41:%.*]] = bitcast float* [[TMP40]] to <4 x float>* 318; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD9:%.*]] = load <4 x float>, <4 x float>* [[TMP41]], align 4 319; VF-FOUR-CHECK-NEXT: [[TMP42:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 36 320; VF-FOUR-CHECK-NEXT: [[TMP43:%.*]] = bitcast float* [[TMP42]] to <4 x float>* 321; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD10:%.*]] = load <4 x float>, <4 x float>* [[TMP43]], align 4 322; VF-FOUR-CHECK-NEXT: [[TMP44:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 40 323; VF-FOUR-CHECK-NEXT: [[TMP45:%.*]] = bitcast float* [[TMP44]] to <4 x float>* 324; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD11:%.*]] = load <4 x float>, <4 x float>* [[TMP45]], align 4 325; VF-FOUR-CHECK-NEXT: [[TMP46:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 44 326; VF-FOUR-CHECK-NEXT: [[TMP47:%.*]] = bitcast float* [[TMP46]] to <4 x float>* 327; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD12:%.*]] = load <4 x float>, <4 x float>* [[TMP47]], align 4 328; VF-FOUR-CHECK-NEXT: [[TMP48:%.*]] = getelementptr inbounds float, float* [[CC:%.*]], i64 [[TMP0]] 329; VF-FOUR-CHECK-NEXT: [[TMP49:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP1]] 330; VF-FOUR-CHECK-NEXT: [[TMP50:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP2]] 331; VF-FOUR-CHECK-NEXT: [[TMP51:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP3]] 332; VF-FOUR-CHECK-NEXT: [[TMP52:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP4]] 333; VF-FOUR-CHECK-NEXT: [[TMP53:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP5]] 334; VF-FOUR-CHECK-NEXT: [[TMP54:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP6]] 335; VF-FOUR-CHECK-NEXT: [[TMP55:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP7]] 336; VF-FOUR-CHECK-NEXT: [[TMP56:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP8]] 337; VF-FOUR-CHECK-NEXT: [[TMP57:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP9]] 338; VF-FOUR-CHECK-NEXT: [[TMP58:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP10]] 339; VF-FOUR-CHECK-NEXT: [[TMP59:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP11]] 340; VF-FOUR-CHECK-NEXT: [[TMP60:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 0 341; VF-FOUR-CHECK-NEXT: [[TMP61:%.*]] = bitcast float* [[TMP60]] to <4 x float>* 342; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD13:%.*]] = load <4 x float>, <4 x float>* [[TMP61]], align 4 343; VF-FOUR-CHECK-NEXT: [[TMP62:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 4 344; VF-FOUR-CHECK-NEXT: [[TMP63:%.*]] = bitcast float* [[TMP62]] to <4 x float>* 345; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD14:%.*]] = load <4 x float>, <4 x float>* [[TMP63]], align 4 346; VF-FOUR-CHECK-NEXT: [[TMP64:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 8 347; VF-FOUR-CHECK-NEXT: [[TMP65:%.*]] = bitcast float* [[TMP64]] to <4 x float>* 348; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD15:%.*]] = load <4 x float>, <4 x float>* [[TMP65]], align 4 349; VF-FOUR-CHECK-NEXT: [[TMP66:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 12 350; VF-FOUR-CHECK-NEXT: [[TMP67:%.*]] = bitcast float* [[TMP66]] to <4 x float>* 351; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD16:%.*]] = load <4 x float>, <4 x float>* [[TMP67]], align 4 352; VF-FOUR-CHECK-NEXT: [[TMP68:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 16 353; VF-FOUR-CHECK-NEXT: [[TMP69:%.*]] = bitcast float* [[TMP68]] to <4 x float>* 354; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD17:%.*]] = load <4 x float>, <4 x float>* [[TMP69]], align 4 355; VF-FOUR-CHECK-NEXT: [[TMP70:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 20 356; VF-FOUR-CHECK-NEXT: [[TMP71:%.*]] = bitcast float* [[TMP70]] to <4 x float>* 357; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD18:%.*]] = load <4 x float>, <4 x float>* [[TMP71]], align 4 358; VF-FOUR-CHECK-NEXT: [[TMP72:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 24 359; VF-FOUR-CHECK-NEXT: [[TMP73:%.*]] = bitcast float* [[TMP72]] to <4 x float>* 360; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD19:%.*]] = load <4 x float>, <4 x float>* [[TMP73]], align 4 361; VF-FOUR-CHECK-NEXT: [[TMP74:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 28 362; VF-FOUR-CHECK-NEXT: [[TMP75:%.*]] = bitcast float* [[TMP74]] to <4 x float>* 363; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD20:%.*]] = load <4 x float>, <4 x float>* [[TMP75]], align 4 364; VF-FOUR-CHECK-NEXT: [[TMP76:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 32 365; VF-FOUR-CHECK-NEXT: [[TMP77:%.*]] = bitcast float* [[TMP76]] to <4 x float>* 366; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD21:%.*]] = load <4 x float>, <4 x float>* [[TMP77]], align 4 367; VF-FOUR-CHECK-NEXT: [[TMP78:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 36 368; VF-FOUR-CHECK-NEXT: [[TMP79:%.*]] = bitcast float* [[TMP78]] to <4 x float>* 369; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD22:%.*]] = load <4 x float>, <4 x float>* [[TMP79]], align 4 370; VF-FOUR-CHECK-NEXT: [[TMP80:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 40 371; VF-FOUR-CHECK-NEXT: [[TMP81:%.*]] = bitcast float* [[TMP80]] to <4 x float>* 372; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD23:%.*]] = load <4 x float>, <4 x float>* [[TMP81]], align 4 373; VF-FOUR-CHECK-NEXT: [[TMP82:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 44 374; VF-FOUR-CHECK-NEXT: [[TMP83:%.*]] = bitcast float* [[TMP82]] to <4 x float>* 375; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD24:%.*]] = load <4 x float>, <4 x float>* [[TMP83]], align 4 376; VF-FOUR-CHECK-NEXT: [[TMP84:%.*]] = fadd fast <4 x float> [[WIDE_LOAD]], [[WIDE_LOAD13]] 377; VF-FOUR-CHECK-NEXT: [[TMP85:%.*]] = fadd fast <4 x float> [[WIDE_LOAD2]], [[WIDE_LOAD14]] 378; VF-FOUR-CHECK-NEXT: [[TMP86:%.*]] = fadd fast <4 x float> [[WIDE_LOAD3]], [[WIDE_LOAD15]] 379; VF-FOUR-CHECK-NEXT: [[TMP87:%.*]] = fadd fast <4 x float> [[WIDE_LOAD4]], [[WIDE_LOAD16]] 380; VF-FOUR-CHECK-NEXT: [[TMP88:%.*]] = fadd fast <4 x float> [[WIDE_LOAD5]], [[WIDE_LOAD17]] 381; VF-FOUR-CHECK-NEXT: [[TMP89:%.*]] = fadd fast <4 x float> [[WIDE_LOAD6]], [[WIDE_LOAD18]] 382; VF-FOUR-CHECK-NEXT: [[TMP90:%.*]] = fadd fast <4 x float> [[WIDE_LOAD7]], [[WIDE_LOAD19]] 383; VF-FOUR-CHECK-NEXT: [[TMP91:%.*]] = fadd fast <4 x float> [[WIDE_LOAD8]], [[WIDE_LOAD20]] 384; VF-FOUR-CHECK-NEXT: [[TMP92:%.*]] = fadd fast <4 x float> [[WIDE_LOAD9]], [[WIDE_LOAD21]] 385; VF-FOUR-CHECK-NEXT: [[TMP93:%.*]] = fadd fast <4 x float> [[WIDE_LOAD10]], [[WIDE_LOAD22]] 386; VF-FOUR-CHECK-NEXT: [[TMP94:%.*]] = fadd fast <4 x float> [[WIDE_LOAD11]], [[WIDE_LOAD23]] 387; VF-FOUR-CHECK-NEXT: [[TMP95:%.*]] = fadd fast <4 x float> [[WIDE_LOAD12]], [[WIDE_LOAD24]] 388; VF-FOUR-CHECK-NEXT: [[TMP96:%.*]] = getelementptr inbounds float, float* [[AA:%.*]], i64 [[TMP0]] 389; VF-FOUR-CHECK-NEXT: [[TMP97:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP1]] 390; VF-FOUR-CHECK-NEXT: [[TMP98:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP2]] 391; VF-FOUR-CHECK-NEXT: [[TMP99:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP3]] 392; VF-FOUR-CHECK-NEXT: [[TMP100:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP4]] 393; VF-FOUR-CHECK-NEXT: [[TMP101:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP5]] 394; VF-FOUR-CHECK-NEXT: [[TMP102:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP6]] 395; VF-FOUR-CHECK-NEXT: [[TMP103:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP7]] 396; VF-FOUR-CHECK-NEXT: [[TMP104:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP8]] 397; VF-FOUR-CHECK-NEXT: [[TMP105:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP9]] 398; VF-FOUR-CHECK-NEXT: [[TMP106:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP10]] 399; VF-FOUR-CHECK-NEXT: [[TMP107:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP11]] 400; VF-FOUR-CHECK-NEXT: [[TMP108:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 0 401; VF-FOUR-CHECK-NEXT: [[TMP109:%.*]] = bitcast float* [[TMP108]] to <4 x float>* 402; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP84]], <4 x float>* [[TMP109]], align 4 403; VF-FOUR-CHECK-NEXT: [[TMP110:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 4 404; VF-FOUR-CHECK-NEXT: [[TMP111:%.*]] = bitcast float* [[TMP110]] to <4 x float>* 405; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP85]], <4 x float>* [[TMP111]], align 4 406; VF-FOUR-CHECK-NEXT: [[TMP112:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 8 407; VF-FOUR-CHECK-NEXT: [[TMP113:%.*]] = bitcast float* [[TMP112]] to <4 x float>* 408; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP86]], <4 x float>* [[TMP113]], align 4 409; VF-FOUR-CHECK-NEXT: [[TMP114:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 12 410; VF-FOUR-CHECK-NEXT: [[TMP115:%.*]] = bitcast float* [[TMP114]] to <4 x float>* 411; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP87]], <4 x float>* [[TMP115]], align 4 412; VF-FOUR-CHECK-NEXT: [[TMP116:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 16 413; VF-FOUR-CHECK-NEXT: [[TMP117:%.*]] = bitcast float* [[TMP116]] to <4 x float>* 414; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP88]], <4 x float>* [[TMP117]], align 4 415; VF-FOUR-CHECK-NEXT: [[TMP118:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 20 416; VF-FOUR-CHECK-NEXT: [[TMP119:%.*]] = bitcast float* [[TMP118]] to <4 x float>* 417; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP89]], <4 x float>* [[TMP119]], align 4 418; VF-FOUR-CHECK-NEXT: [[TMP120:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 24 419; VF-FOUR-CHECK-NEXT: [[TMP121:%.*]] = bitcast float* [[TMP120]] to <4 x float>* 420; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP90]], <4 x float>* [[TMP121]], align 4 421; VF-FOUR-CHECK-NEXT: [[TMP122:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 28 422; VF-FOUR-CHECK-NEXT: [[TMP123:%.*]] = bitcast float* [[TMP122]] to <4 x float>* 423; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP91]], <4 x float>* [[TMP123]], align 4 424; VF-FOUR-CHECK-NEXT: [[TMP124:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 32 425; VF-FOUR-CHECK-NEXT: [[TMP125:%.*]] = bitcast float* [[TMP124]] to <4 x float>* 426; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP92]], <4 x float>* [[TMP125]], align 4 427; VF-FOUR-CHECK-NEXT: [[TMP126:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 36 428; VF-FOUR-CHECK-NEXT: [[TMP127:%.*]] = bitcast float* [[TMP126]] to <4 x float>* 429; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP93]], <4 x float>* [[TMP127]], align 4 430; VF-FOUR-CHECK-NEXT: [[TMP128:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 40 431; VF-FOUR-CHECK-NEXT: [[TMP129:%.*]] = bitcast float* [[TMP128]] to <4 x float>* 432; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP94]], <4 x float>* [[TMP129]], align 4 433; VF-FOUR-CHECK-NEXT: [[TMP130:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 44 434; VF-FOUR-CHECK-NEXT: [[TMP131:%.*]] = bitcast float* [[TMP130]] to <4 x float>* 435; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP95]], <4 x float>* [[TMP131]], align 4 436; VF-FOUR-CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 48 437; VF-FOUR-CHECK-NEXT: [[TMP132:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] 438; VF-FOUR-CHECK-NEXT: br i1 [[TMP132]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] 439; VF-FOUR-CHECK: middle.block: 440; VF-FOUR-CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC]] 441; VF-FOUR-CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]] 442; VF-FOUR-CHECK: vec.epilog.iter.check: 443; VF-FOUR-CHECK-NEXT: [[N_VEC_REMAINING:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_VEC]] 444; VF-FOUR-CHECK-NEXT: [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 4 445; VF-FOUR-CHECK-NEXT: br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]] 446; VF-FOUR-CHECK: vec.epilog.ph: 447; VF-FOUR-CHECK-NEXT: [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ] 448; VF-FOUR-CHECK-NEXT: [[N_MOD_VF25:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 4 449; VF-FOUR-CHECK-NEXT: [[N_VEC26:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF25]] 450; VF-FOUR-CHECK-NEXT: br label [[VEC_EPILOG_VECTOR_BODY:%.*]] 451; VF-FOUR-CHECK: vec.epilog.vector.body: 452; VF-FOUR-CHECK-NEXT: [[OFFSET_IDX:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT31:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ] 453; VF-FOUR-CHECK-NEXT: [[TMP133:%.*]] = add i64 [[OFFSET_IDX]], 0 454; VF-FOUR-CHECK-NEXT: [[TMP134:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP133]] 455; VF-FOUR-CHECK-NEXT: [[TMP135:%.*]] = getelementptr inbounds float, float* [[TMP134]], i32 0 456; VF-FOUR-CHECK-NEXT: [[TMP136:%.*]] = bitcast float* [[TMP135]] to <4 x float>* 457; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD29:%.*]] = load <4 x float>, <4 x float>* [[TMP136]], align 4 458; VF-FOUR-CHECK-NEXT: [[TMP137:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP133]] 459; VF-FOUR-CHECK-NEXT: [[TMP138:%.*]] = getelementptr inbounds float, float* [[TMP137]], i32 0 460; VF-FOUR-CHECK-NEXT: [[TMP139:%.*]] = bitcast float* [[TMP138]] to <4 x float>* 461; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD30:%.*]] = load <4 x float>, <4 x float>* [[TMP139]], align 4 462; VF-FOUR-CHECK-NEXT: [[TMP140:%.*]] = fadd fast <4 x float> [[WIDE_LOAD29]], [[WIDE_LOAD30]] 463; VF-FOUR-CHECK-NEXT: [[TMP141:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP133]] 464; VF-FOUR-CHECK-NEXT: [[TMP142:%.*]] = getelementptr inbounds float, float* [[TMP141]], i32 0 465; VF-FOUR-CHECK-NEXT: [[TMP143:%.*]] = bitcast float* [[TMP142]] to <4 x float>* 466; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP140]], <4 x float>* [[TMP143]], align 4 467; VF-FOUR-CHECK-NEXT: [[INDEX_NEXT31]] = add nuw i64 [[OFFSET_IDX]], 4 468; VF-FOUR-CHECK-NEXT: [[TMP144:%.*]] = icmp eq i64 [[INDEX_NEXT31]], [[N_VEC26]] 469; VF-FOUR-CHECK-NEXT: br i1 [[TMP144]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP2:![0-9]+]] 470; VF-FOUR-CHECK: vec.epilog.middle.block: 471; VF-FOUR-CHECK-NEXT: [[CMP_N27:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC26]] 472; VF-FOUR-CHECK-NEXT: br i1 [[CMP_N27]], label [[FOR_END_LOOPEXIT]], label [[VEC_EPILOG_SCALAR_PH]] 473; VF-FOUR-CHECK: vec.epilog.scalar.ph: 474; VF-FOUR-CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC26]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK]] ] 475; VF-FOUR-CHECK-NEXT: br label [[FOR_BODY:%.*]] 476; VF-FOUR-CHECK: for.body: 477; VF-FOUR-CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ] 478; VF-FOUR-CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[INDVARS_IV]] 479; VF-FOUR-CHECK-NEXT: [[TMP145:%.*]] = load float, float* [[ARRAYIDX]], align 4 480; VF-FOUR-CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[INDVARS_IV]] 481; VF-FOUR-CHECK-NEXT: [[TMP146:%.*]] = load float, float* [[ARRAYIDX2]], align 4 482; VF-FOUR-CHECK-NEXT: [[ADD:%.*]] = fadd fast float [[TMP145]], [[TMP146]] 483; VF-FOUR-CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[INDVARS_IV]] 484; VF-FOUR-CHECK-NEXT: store float [[ADD]], float* [[ARRAYIDX4]], align 4 485; VF-FOUR-CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1 486; VF-FOUR-CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i64 [[INDVARS_IV_NEXT]], [[WIDE_TRIP_COUNT]] 487; VF-FOUR-CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_BODY]], label [[FOR_END_LOOPEXIT]], !llvm.loop [[LOOP4:![0-9]+]] 488; VF-FOUR-CHECK: for.end.loopexit: 489; VF-FOUR-CHECK-NEXT: br label [[FOR_END]] 490; VF-FOUR-CHECK: for.end: 491; VF-FOUR-CHECK-NEXT: ret void 492; 493 494 495entry: 496 %cmp1 = icmp sgt i32 %N, 0 497 br i1 %cmp1, label %for.body.preheader, label %for.end 498 499for.body.preheader: ; preds = %entry 500 %wide.trip.count = zext i32 %N to i64 501 br label %for.body 502 503for.body: ; preds = %for.body.preheader, %for.body 504 %indvars.iv = phi i64 [ 0, %for.body.preheader ], [ %indvars.iv.next, %for.body ] 505 %arrayidx = getelementptr inbounds float, float* %bb, i64 %indvars.iv 506 %0 = load float, float* %arrayidx, align 4 507 %arrayidx2 = getelementptr inbounds float, float* %cc, i64 %indvars.iv 508 %1 = load float, float* %arrayidx2, align 4 509 %add = fadd fast float %0, %1 510 %arrayidx4 = getelementptr inbounds float, float* %aa, i64 %indvars.iv 511 store float %add, float* %arrayidx4, align 4 512 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 513 %exitcond = icmp ne i64 %indvars.iv.next, %wide.trip.count 514 br i1 %exitcond, label %for.body, label %for.end.loopexit 515 516for.end.loopexit: ; preds = %for.body 517 br label %for.end 518 519for.end: ; preds = %for.end.loopexit, %entry 520 ret void 521} 522 523define dso_local signext i32 @f2(float* noalias %A, float* noalias %B, i32 signext %n) #0 { 524; VF-TWO-CHECK-LABEL: @f2( 525; VF-TWO-CHECK-NEXT: entry: 526; VF-TWO-CHECK-NEXT: [[CMP1:%.*]] = icmp sgt i32 [[N:%.*]], 1 527; VF-TWO-CHECK-NEXT: br i1 [[CMP1]], label [[ITER_CHECK:%.*]], label [[FOR_END:%.*]] 528; VF-TWO-CHECK: iter.check: 529; VF-TWO-CHECK-NEXT: [[TMP0:%.*]] = add i32 [[N]], -1 530; VF-TWO-CHECK-NEXT: [[WIDE_TRIP_COUNT:%.*]] = zext i32 [[TMP0]] to i64 531; VF-TWO-CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 2 532; VF-TWO-CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]] 533; VF-TWO-CHECK: vector.scevcheck: 534; VF-TWO-CHECK-NEXT: [[TMP1:%.*]] = add nsw i64 [[WIDE_TRIP_COUNT]], -1 535; VF-TWO-CHECK-NEXT: [[TMP2:%.*]] = trunc i64 [[TMP1]] to i32 536; VF-TWO-CHECK-NEXT: [[MUL:%.*]] = call { i32, i1 } @llvm.umul.with.overflow.i32(i32 1, i32 [[TMP2]]) 537; VF-TWO-CHECK-NEXT: [[MUL_RESULT:%.*]] = extractvalue { i32, i1 } [[MUL]], 0 538; VF-TWO-CHECK-NEXT: [[MUL_OVERFLOW:%.*]] = extractvalue { i32, i1 } [[MUL]], 1 539; VF-TWO-CHECK-NEXT: [[TMP3:%.*]] = sub i32 [[TMP0]], [[MUL_RESULT]] 540; VF-TWO-CHECK-NEXT: [[TMP4:%.*]] = icmp sgt i32 [[TMP3]], [[TMP0]] 541; VF-TWO-CHECK-NEXT: [[TMP5:%.*]] = or i1 [[TMP4]], [[MUL_OVERFLOW]] 542; VF-TWO-CHECK-NEXT: [[TMP6:%.*]] = icmp ugt i64 [[TMP1]], 4294967295 543; VF-TWO-CHECK-NEXT: [[TMP7:%.*]] = or i1 [[TMP5]], [[TMP6]] 544; VF-TWO-CHECK-NEXT: br i1 [[TMP7]], label [[VEC_EPILOG_SCALAR_PH]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]] 545; VF-TWO-CHECK: vector.main.loop.iter.check: 546; VF-TWO-CHECK-NEXT: [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 32 547; VF-TWO-CHECK-NEXT: br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]] 548; VF-TWO-CHECK: vector.ph: 549; VF-TWO-CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 32 550; VF-TWO-CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF]] 551; VF-TWO-CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 552; VF-TWO-CHECK: vector.body: 553; VF-TWO-CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 554; VF-TWO-CHECK-NEXT: [[OFFSET_IDX:%.*]] = trunc i64 [[INDEX]] to i32 555; VF-TWO-CHECK-NEXT: [[TMP8:%.*]] = add i32 [[OFFSET_IDX]], 0 556; VF-TWO-CHECK-NEXT: [[TMP9:%.*]] = add i32 [[OFFSET_IDX]], 4 557; VF-TWO-CHECK-NEXT: [[TMP10:%.*]] = add i32 [[OFFSET_IDX]], 8 558; VF-TWO-CHECK-NEXT: [[TMP11:%.*]] = add i32 [[OFFSET_IDX]], 12 559; VF-TWO-CHECK-NEXT: [[TMP12:%.*]] = add i32 [[OFFSET_IDX]], 16 560; VF-TWO-CHECK-NEXT: [[TMP13:%.*]] = add i32 [[OFFSET_IDX]], 20 561; VF-TWO-CHECK-NEXT: [[TMP14:%.*]] = add i32 [[OFFSET_IDX]], 24 562; VF-TWO-CHECK-NEXT: [[TMP15:%.*]] = add i32 [[OFFSET_IDX]], 28 563; VF-TWO-CHECK-NEXT: [[TMP16:%.*]] = add i64 [[INDEX]], 0 564; VF-TWO-CHECK-NEXT: [[TMP17:%.*]] = add i64 [[INDEX]], 4 565; VF-TWO-CHECK-NEXT: [[TMP18:%.*]] = add i64 [[INDEX]], 8 566; VF-TWO-CHECK-NEXT: [[TMP19:%.*]] = add i64 [[INDEX]], 12 567; VF-TWO-CHECK-NEXT: [[TMP20:%.*]] = add i64 [[INDEX]], 16 568; VF-TWO-CHECK-NEXT: [[TMP21:%.*]] = add i64 [[INDEX]], 20 569; VF-TWO-CHECK-NEXT: [[TMP22:%.*]] = add i64 [[INDEX]], 24 570; VF-TWO-CHECK-NEXT: [[TMP23:%.*]] = add i64 [[INDEX]], 28 571; VF-TWO-CHECK-NEXT: [[TMP24:%.*]] = xor i32 [[TMP8]], -1 572; VF-TWO-CHECK-NEXT: [[TMP25:%.*]] = xor i32 [[TMP9]], -1 573; VF-TWO-CHECK-NEXT: [[TMP26:%.*]] = xor i32 [[TMP10]], -1 574; VF-TWO-CHECK-NEXT: [[TMP27:%.*]] = xor i32 [[TMP11]], -1 575; VF-TWO-CHECK-NEXT: [[TMP28:%.*]] = xor i32 [[TMP12]], -1 576; VF-TWO-CHECK-NEXT: [[TMP29:%.*]] = xor i32 [[TMP13]], -1 577; VF-TWO-CHECK-NEXT: [[TMP30:%.*]] = xor i32 [[TMP14]], -1 578; VF-TWO-CHECK-NEXT: [[TMP31:%.*]] = xor i32 [[TMP15]], -1 579; VF-TWO-CHECK-NEXT: [[TMP32:%.*]] = add i32 [[TMP24]], [[N]] 580; VF-TWO-CHECK-NEXT: [[TMP33:%.*]] = add i32 [[TMP25]], [[N]] 581; VF-TWO-CHECK-NEXT: [[TMP34:%.*]] = add i32 [[TMP26]], [[N]] 582; VF-TWO-CHECK-NEXT: [[TMP35:%.*]] = add i32 [[TMP27]], [[N]] 583; VF-TWO-CHECK-NEXT: [[TMP36:%.*]] = add i32 [[TMP28]], [[N]] 584; VF-TWO-CHECK-NEXT: [[TMP37:%.*]] = add i32 [[TMP29]], [[N]] 585; VF-TWO-CHECK-NEXT: [[TMP38:%.*]] = add i32 [[TMP30]], [[N]] 586; VF-TWO-CHECK-NEXT: [[TMP39:%.*]] = add i32 [[TMP31]], [[N]] 587; VF-TWO-CHECK-NEXT: [[TMP40:%.*]] = sext i32 [[TMP32]] to i64 588; VF-TWO-CHECK-NEXT: [[TMP41:%.*]] = sext i32 [[TMP33]] to i64 589; VF-TWO-CHECK-NEXT: [[TMP42:%.*]] = sext i32 [[TMP34]] to i64 590; VF-TWO-CHECK-NEXT: [[TMP43:%.*]] = sext i32 [[TMP35]] to i64 591; VF-TWO-CHECK-NEXT: [[TMP44:%.*]] = sext i32 [[TMP36]] to i64 592; VF-TWO-CHECK-NEXT: [[TMP45:%.*]] = sext i32 [[TMP37]] to i64 593; VF-TWO-CHECK-NEXT: [[TMP46:%.*]] = sext i32 [[TMP38]] to i64 594; VF-TWO-CHECK-NEXT: [[TMP47:%.*]] = sext i32 [[TMP39]] to i64 595; VF-TWO-CHECK-NEXT: [[TMP48:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 [[TMP40]] 596; VF-TWO-CHECK-NEXT: [[TMP49:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP41]] 597; VF-TWO-CHECK-NEXT: [[TMP50:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP42]] 598; VF-TWO-CHECK-NEXT: [[TMP51:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP43]] 599; VF-TWO-CHECK-NEXT: [[TMP52:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP44]] 600; VF-TWO-CHECK-NEXT: [[TMP53:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP45]] 601; VF-TWO-CHECK-NEXT: [[TMP54:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP46]] 602; VF-TWO-CHECK-NEXT: [[TMP55:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP47]] 603; VF-TWO-CHECK-NEXT: [[TMP56:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 0 604; VF-TWO-CHECK-NEXT: [[TMP57:%.*]] = getelementptr inbounds float, float* [[TMP56]], i32 -3 605; VF-TWO-CHECK-NEXT: [[TMP58:%.*]] = bitcast float* [[TMP57]] to <4 x float>* 606; VF-TWO-CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, <4 x float>* [[TMP58]], align 4 607; 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> 608; VF-TWO-CHECK-NEXT: [[TMP59:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 -4 609; VF-TWO-CHECK-NEXT: [[TMP60:%.*]] = getelementptr inbounds float, float* [[TMP59]], i32 -3 610; VF-TWO-CHECK-NEXT: [[TMP61:%.*]] = bitcast float* [[TMP60]] to <4 x float>* 611; VF-TWO-CHECK-NEXT: [[WIDE_LOAD2:%.*]] = load <4 x float>, <4 x float>* [[TMP61]], align 4 612; 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> 613; VF-TWO-CHECK-NEXT: [[TMP62:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 -8 614; VF-TWO-CHECK-NEXT: [[TMP63:%.*]] = getelementptr inbounds float, float* [[TMP62]], i32 -3 615; VF-TWO-CHECK-NEXT: [[TMP64:%.*]] = bitcast float* [[TMP63]] to <4 x float>* 616; VF-TWO-CHECK-NEXT: [[WIDE_LOAD4:%.*]] = load <4 x float>, <4 x float>* [[TMP64]], align 4 617; 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> 618; VF-TWO-CHECK-NEXT: [[TMP65:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 -12 619; VF-TWO-CHECK-NEXT: [[TMP66:%.*]] = getelementptr inbounds float, float* [[TMP65]], i32 -3 620; VF-TWO-CHECK-NEXT: [[TMP67:%.*]] = bitcast float* [[TMP66]] to <4 x float>* 621; VF-TWO-CHECK-NEXT: [[WIDE_LOAD6:%.*]] = load <4 x float>, <4 x float>* [[TMP67]], align 4 622; 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> 623; VF-TWO-CHECK-NEXT: [[TMP68:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 -16 624; VF-TWO-CHECK-NEXT: [[TMP69:%.*]] = getelementptr inbounds float, float* [[TMP68]], i32 -3 625; VF-TWO-CHECK-NEXT: [[TMP70:%.*]] = bitcast float* [[TMP69]] to <4 x float>* 626; VF-TWO-CHECK-NEXT: [[WIDE_LOAD8:%.*]] = load <4 x float>, <4 x float>* [[TMP70]], align 4 627; 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> 628; VF-TWO-CHECK-NEXT: [[TMP71:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 -20 629; VF-TWO-CHECK-NEXT: [[TMP72:%.*]] = getelementptr inbounds float, float* [[TMP71]], i32 -3 630; VF-TWO-CHECK-NEXT: [[TMP73:%.*]] = bitcast float* [[TMP72]] to <4 x float>* 631; VF-TWO-CHECK-NEXT: [[WIDE_LOAD10:%.*]] = load <4 x float>, <4 x float>* [[TMP73]], align 4 632; 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> 633; VF-TWO-CHECK-NEXT: [[TMP74:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 -24 634; VF-TWO-CHECK-NEXT: [[TMP75:%.*]] = getelementptr inbounds float, float* [[TMP74]], i32 -3 635; VF-TWO-CHECK-NEXT: [[TMP76:%.*]] = bitcast float* [[TMP75]] to <4 x float>* 636; VF-TWO-CHECK-NEXT: [[WIDE_LOAD12:%.*]] = load <4 x float>, <4 x float>* [[TMP76]], align 4 637; 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> 638; VF-TWO-CHECK-NEXT: [[TMP77:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 -28 639; VF-TWO-CHECK-NEXT: [[TMP78:%.*]] = getelementptr inbounds float, float* [[TMP77]], i32 -3 640; VF-TWO-CHECK-NEXT: [[TMP79:%.*]] = bitcast float* [[TMP78]] to <4 x float>* 641; VF-TWO-CHECK-NEXT: [[WIDE_LOAD14:%.*]] = load <4 x float>, <4 x float>* [[TMP79]], align 4 642; 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> 643; VF-TWO-CHECK-NEXT: [[TMP80:%.*]] = fadd fast <4 x float> [[REVERSE]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 644; VF-TWO-CHECK-NEXT: [[TMP81:%.*]] = fadd fast <4 x float> [[REVERSE3]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 645; VF-TWO-CHECK-NEXT: [[TMP82:%.*]] = fadd fast <4 x float> [[REVERSE5]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 646; VF-TWO-CHECK-NEXT: [[TMP83:%.*]] = fadd fast <4 x float> [[REVERSE7]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 647; VF-TWO-CHECK-NEXT: [[TMP84:%.*]] = fadd fast <4 x float> [[REVERSE9]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 648; VF-TWO-CHECK-NEXT: [[TMP85:%.*]] = fadd fast <4 x float> [[REVERSE11]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 649; VF-TWO-CHECK-NEXT: [[TMP86:%.*]] = fadd fast <4 x float> [[REVERSE13]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 650; VF-TWO-CHECK-NEXT: [[TMP87:%.*]] = fadd fast <4 x float> [[REVERSE15]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 651; VF-TWO-CHECK-NEXT: [[TMP88:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[TMP16]] 652; VF-TWO-CHECK-NEXT: [[TMP89:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP17]] 653; VF-TWO-CHECK-NEXT: [[TMP90:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP18]] 654; VF-TWO-CHECK-NEXT: [[TMP91:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP19]] 655; VF-TWO-CHECK-NEXT: [[TMP92:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP20]] 656; VF-TWO-CHECK-NEXT: [[TMP93:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP21]] 657; VF-TWO-CHECK-NEXT: [[TMP94:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP22]] 658; VF-TWO-CHECK-NEXT: [[TMP95:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP23]] 659; VF-TWO-CHECK-NEXT: [[TMP96:%.*]] = getelementptr inbounds float, float* [[TMP88]], i32 0 660; VF-TWO-CHECK-NEXT: [[TMP97:%.*]] = bitcast float* [[TMP96]] to <4 x float>* 661; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP80]], <4 x float>* [[TMP97]], align 4 662; VF-TWO-CHECK-NEXT: [[TMP98:%.*]] = getelementptr inbounds float, float* [[TMP88]], i32 4 663; VF-TWO-CHECK-NEXT: [[TMP99:%.*]] = bitcast float* [[TMP98]] to <4 x float>* 664; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP81]], <4 x float>* [[TMP99]], align 4 665; VF-TWO-CHECK-NEXT: [[TMP100:%.*]] = getelementptr inbounds float, float* [[TMP88]], i32 8 666; VF-TWO-CHECK-NEXT: [[TMP101:%.*]] = bitcast float* [[TMP100]] to <4 x float>* 667; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP82]], <4 x float>* [[TMP101]], align 4 668; VF-TWO-CHECK-NEXT: [[TMP102:%.*]] = getelementptr inbounds float, float* [[TMP88]], i32 12 669; VF-TWO-CHECK-NEXT: [[TMP103:%.*]] = bitcast float* [[TMP102]] to <4 x float>* 670; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP83]], <4 x float>* [[TMP103]], align 4 671; VF-TWO-CHECK-NEXT: [[TMP104:%.*]] = getelementptr inbounds float, float* [[TMP88]], i32 16 672; VF-TWO-CHECK-NEXT: [[TMP105:%.*]] = bitcast float* [[TMP104]] to <4 x float>* 673; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP84]], <4 x float>* [[TMP105]], align 4 674; VF-TWO-CHECK-NEXT: [[TMP106:%.*]] = getelementptr inbounds float, float* [[TMP88]], i32 20 675; VF-TWO-CHECK-NEXT: [[TMP107:%.*]] = bitcast float* [[TMP106]] to <4 x float>* 676; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP85]], <4 x float>* [[TMP107]], align 4 677; VF-TWO-CHECK-NEXT: [[TMP108:%.*]] = getelementptr inbounds float, float* [[TMP88]], i32 24 678; VF-TWO-CHECK-NEXT: [[TMP109:%.*]] = bitcast float* [[TMP108]] to <4 x float>* 679; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP86]], <4 x float>* [[TMP109]], align 4 680; VF-TWO-CHECK-NEXT: [[TMP110:%.*]] = getelementptr inbounds float, float* [[TMP88]], i32 28 681; VF-TWO-CHECK-NEXT: [[TMP111:%.*]] = bitcast float* [[TMP110]] to <4 x float>* 682; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP87]], <4 x float>* [[TMP111]], align 4 683; VF-TWO-CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 32 684; VF-TWO-CHECK-NEXT: [[TMP112:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] 685; VF-TWO-CHECK-NEXT: br i1 [[TMP112]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]] 686; VF-TWO-CHECK: middle.block: 687; VF-TWO-CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC]] 688; VF-TWO-CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]] 689; VF-TWO-CHECK: vec.epilog.iter.check: 690; VF-TWO-CHECK-NEXT: [[IND_END19:%.*]] = trunc i64 [[N_VEC]] to i32 691; VF-TWO-CHECK-NEXT: [[N_VEC_REMAINING:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_VEC]] 692; VF-TWO-CHECK-NEXT: [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 2 693; VF-TWO-CHECK-NEXT: br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]] 694; VF-TWO-CHECK: vec.epilog.ph: 695; VF-TWO-CHECK-NEXT: [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ] 696; VF-TWO-CHECK-NEXT: [[N_MOD_VF16:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 2 697; VF-TWO-CHECK-NEXT: [[N_VEC17:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF16]] 698; VF-TWO-CHECK-NEXT: [[IND_END:%.*]] = trunc i64 [[N_VEC17]] to i32 699; VF-TWO-CHECK-NEXT: br label [[VEC_EPILOG_VECTOR_BODY:%.*]] 700; VF-TWO-CHECK: vec.epilog.vector.body: 701; VF-TWO-CHECK-NEXT: [[OFFSET_IDX23:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT26:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ] 702; VF-TWO-CHECK-NEXT: [[OFFSET_IDX22:%.*]] = trunc i64 [[OFFSET_IDX23]] to i32 703; VF-TWO-CHECK-NEXT: [[TMP113:%.*]] = add i32 [[OFFSET_IDX22]], 0 704; VF-TWO-CHECK-NEXT: [[TMP114:%.*]] = add i64 [[OFFSET_IDX23]], 0 705; VF-TWO-CHECK-NEXT: [[TMP115:%.*]] = xor i32 [[TMP113]], -1 706; VF-TWO-CHECK-NEXT: [[TMP116:%.*]] = add i32 [[TMP115]], [[N]] 707; VF-TWO-CHECK-NEXT: [[TMP117:%.*]] = sext i32 [[TMP116]] to i64 708; VF-TWO-CHECK-NEXT: [[TMP118:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP117]] 709; VF-TWO-CHECK-NEXT: [[TMP119:%.*]] = getelementptr inbounds float, float* [[TMP118]], i32 0 710; VF-TWO-CHECK-NEXT: [[TMP120:%.*]] = getelementptr inbounds float, float* [[TMP119]], i32 -1 711; VF-TWO-CHECK-NEXT: [[TMP121:%.*]] = bitcast float* [[TMP120]] to <2 x float>* 712; VF-TWO-CHECK-NEXT: [[WIDE_LOAD24:%.*]] = load <2 x float>, <2 x float>* [[TMP121]], align 4 713; VF-TWO-CHECK-NEXT: [[REVERSE25:%.*]] = shufflevector <2 x float> [[WIDE_LOAD24]], <2 x float> poison, <2 x i32> <i32 1, i32 0> 714; VF-TWO-CHECK-NEXT: [[TMP122:%.*]] = fadd fast <2 x float> [[REVERSE25]], <float 1.000000e+00, float 1.000000e+00> 715; VF-TWO-CHECK-NEXT: [[TMP123:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP114]] 716; VF-TWO-CHECK-NEXT: [[TMP124:%.*]] = getelementptr inbounds float, float* [[TMP123]], i32 0 717; VF-TWO-CHECK-NEXT: [[TMP125:%.*]] = bitcast float* [[TMP124]] to <2 x float>* 718; VF-TWO-CHECK-NEXT: store <2 x float> [[TMP122]], <2 x float>* [[TMP125]], align 4 719; VF-TWO-CHECK-NEXT: [[INDEX_NEXT26]] = add nuw i64 [[OFFSET_IDX23]], 2 720; VF-TWO-CHECK-NEXT: [[TMP126:%.*]] = icmp eq i64 [[INDEX_NEXT26]], [[N_VEC17]] 721; VF-TWO-CHECK-NEXT: br i1 [[TMP126]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]] 722; VF-TWO-CHECK: vec.epilog.middle.block: 723; VF-TWO-CHECK-NEXT: [[CMP_N20:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC17]] 724; VF-TWO-CHECK-NEXT: br i1 [[CMP_N20]], label [[FOR_END_LOOPEXIT]], label [[VEC_EPILOG_SCALAR_PH]] 725; VF-TWO-CHECK: vec.epilog.scalar.ph: 726; 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]] ] 727; VF-TWO-CHECK-NEXT: [[BC_RESUME_VAL18:%.*]] = phi i32 [ [[IND_END]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[IND_END19]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ITER_CHECK]] ] 728; VF-TWO-CHECK-NEXT: br label [[FOR_BODY:%.*]] 729; VF-TWO-CHECK: for.body: 730; VF-TWO-CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ] 731; VF-TWO-CHECK-NEXT: [[I_014:%.*]] = phi i32 [ [[BC_RESUME_VAL18]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ] 732; VF-TWO-CHECK-NEXT: [[TMP127:%.*]] = xor i32 [[I_014]], -1 733; VF-TWO-CHECK-NEXT: [[SUB2:%.*]] = add i32 [[TMP127]], [[N]] 734; VF-TWO-CHECK-NEXT: [[IDXPROM:%.*]] = sext i32 [[SUB2]] to i64 735; VF-TWO-CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[IDXPROM]] 736; VF-TWO-CHECK-NEXT: [[TMP128:%.*]] = load float, float* [[ARRAYIDX]], align 4 737; VF-TWO-CHECK-NEXT: [[CONV3:%.*]] = fadd fast float [[TMP128]], 1.000000e+00 738; VF-TWO-CHECK-NEXT: [[ARRAYIDX5:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDVARS_IV]] 739; VF-TWO-CHECK-NEXT: store float [[CONV3]], float* [[ARRAYIDX5]], align 4 740; VF-TWO-CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1 741; VF-TWO-CHECK-NEXT: [[INC]] = add nuw nsw i32 [[I_014]], 1 742; VF-TWO-CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i64 [[INDVARS_IV_NEXT]], [[WIDE_TRIP_COUNT]] 743; VF-TWO-CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_BODY]], label [[FOR_END_LOOPEXIT]], !llvm.loop [[LOOPID_MS_CM:![0-9]+]] 744; VF-TWO-CHECK: for.end.loopexit: 745; VF-TWO-CHECK-NEXT: br label [[FOR_END]] 746; VF-TWO-CHECK: for.end: 747; VF-TWO-CHECK-NEXT: ret i32 0 748; 749; VF-FOUR-CHECK-LABEL: @f2( 750; VF-FOUR-CHECK-NEXT: entry: 751; VF-FOUR-CHECK-NEXT: [[CMP1:%.*]] = icmp sgt i32 [[N:%.*]], 1 752; VF-FOUR-CHECK-NEXT: br i1 [[CMP1]], label [[ITER_CHECK:%.*]], label [[FOR_END:%.*]] 753; VF-FOUR-CHECK: iter.check: 754; VF-FOUR-CHECK-NEXT: [[TMP0:%.*]] = add i32 [[N]], -1 755; VF-FOUR-CHECK-NEXT: [[WIDE_TRIP_COUNT:%.*]] = zext i32 [[TMP0]] to i64 756; VF-FOUR-CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 4 757; VF-FOUR-CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]] 758; VF-FOUR-CHECK: vector.scevcheck: 759; VF-FOUR-CHECK-NEXT: [[TMP1:%.*]] = add nsw i64 [[WIDE_TRIP_COUNT]], -1 760; VF-FOUR-CHECK-NEXT: [[TMP2:%.*]] = trunc i64 [[TMP1]] to i32 761; VF-FOUR-CHECK-NEXT: [[MUL:%.*]] = call { i32, i1 } @llvm.umul.with.overflow.i32(i32 1, i32 [[TMP2]]) 762; VF-FOUR-CHECK-NEXT: [[MUL_RESULT:%.*]] = extractvalue { i32, i1 } [[MUL]], 0 763; VF-FOUR-CHECK-NEXT: [[MUL_OVERFLOW:%.*]] = extractvalue { i32, i1 } [[MUL]], 1 764; VF-FOUR-CHECK-NEXT: [[TMP3:%.*]] = sub i32 [[TMP0]], [[MUL_RESULT]] 765; VF-FOUR-CHECK-NEXT: [[TMP4:%.*]] = icmp sgt i32 [[TMP3]], [[TMP0]] 766; VF-FOUR-CHECK-NEXT: [[TMP5:%.*]] = or i1 [[TMP4]], [[MUL_OVERFLOW]] 767; VF-FOUR-CHECK-NEXT: [[TMP6:%.*]] = icmp ugt i64 [[TMP1]], 4294967295 768; VF-FOUR-CHECK-NEXT: [[TMP7:%.*]] = or i1 [[TMP5]], [[TMP6]] 769; VF-FOUR-CHECK-NEXT: br i1 [[TMP7]], label [[VEC_EPILOG_SCALAR_PH]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]] 770; VF-FOUR-CHECK: vector.main.loop.iter.check: 771; VF-FOUR-CHECK-NEXT: [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 32 772; VF-FOUR-CHECK-NEXT: br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]] 773; VF-FOUR-CHECK: vector.ph: 774; VF-FOUR-CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 32 775; VF-FOUR-CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF]] 776; VF-FOUR-CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 777; VF-FOUR-CHECK: vector.body: 778; VF-FOUR-CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 779; VF-FOUR-CHECK-NEXT: [[OFFSET_IDX:%.*]] = trunc i64 [[INDEX]] to i32 780; VF-FOUR-CHECK-NEXT: [[TMP8:%.*]] = add i32 [[OFFSET_IDX]], 0 781; VF-FOUR-CHECK-NEXT: [[TMP9:%.*]] = add i32 [[OFFSET_IDX]], 4 782; VF-FOUR-CHECK-NEXT: [[TMP10:%.*]] = add i32 [[OFFSET_IDX]], 8 783; VF-FOUR-CHECK-NEXT: [[TMP11:%.*]] = add i32 [[OFFSET_IDX]], 12 784; VF-FOUR-CHECK-NEXT: [[TMP12:%.*]] = add i32 [[OFFSET_IDX]], 16 785; VF-FOUR-CHECK-NEXT: [[TMP13:%.*]] = add i32 [[OFFSET_IDX]], 20 786; VF-FOUR-CHECK-NEXT: [[TMP14:%.*]] = add i32 [[OFFSET_IDX]], 24 787; VF-FOUR-CHECK-NEXT: [[TMP15:%.*]] = add i32 [[OFFSET_IDX]], 28 788; VF-FOUR-CHECK-NEXT: [[TMP16:%.*]] = add i64 [[INDEX]], 0 789; VF-FOUR-CHECK-NEXT: [[TMP17:%.*]] = add i64 [[INDEX]], 4 790; VF-FOUR-CHECK-NEXT: [[TMP18:%.*]] = add i64 [[INDEX]], 8 791; VF-FOUR-CHECK-NEXT: [[TMP19:%.*]] = add i64 [[INDEX]], 12 792; VF-FOUR-CHECK-NEXT: [[TMP20:%.*]] = add i64 [[INDEX]], 16 793; VF-FOUR-CHECK-NEXT: [[TMP21:%.*]] = add i64 [[INDEX]], 20 794; VF-FOUR-CHECK-NEXT: [[TMP22:%.*]] = add i64 [[INDEX]], 24 795; VF-FOUR-CHECK-NEXT: [[TMP23:%.*]] = add i64 [[INDEX]], 28 796; VF-FOUR-CHECK-NEXT: [[TMP24:%.*]] = xor i32 [[TMP8]], -1 797; VF-FOUR-CHECK-NEXT: [[TMP25:%.*]] = xor i32 [[TMP9]], -1 798; VF-FOUR-CHECK-NEXT: [[TMP26:%.*]] = xor i32 [[TMP10]], -1 799; VF-FOUR-CHECK-NEXT: [[TMP27:%.*]] = xor i32 [[TMP11]], -1 800; VF-FOUR-CHECK-NEXT: [[TMP28:%.*]] = xor i32 [[TMP12]], -1 801; VF-FOUR-CHECK-NEXT: [[TMP29:%.*]] = xor i32 [[TMP13]], -1 802; VF-FOUR-CHECK-NEXT: [[TMP30:%.*]] = xor i32 [[TMP14]], -1 803; VF-FOUR-CHECK-NEXT: [[TMP31:%.*]] = xor i32 [[TMP15]], -1 804; VF-FOUR-CHECK-NEXT: [[TMP32:%.*]] = add i32 [[TMP24]], [[N]] 805; VF-FOUR-CHECK-NEXT: [[TMP33:%.*]] = add i32 [[TMP25]], [[N]] 806; VF-FOUR-CHECK-NEXT: [[TMP34:%.*]] = add i32 [[TMP26]], [[N]] 807; VF-FOUR-CHECK-NEXT: [[TMP35:%.*]] = add i32 [[TMP27]], [[N]] 808; VF-FOUR-CHECK-NEXT: [[TMP36:%.*]] = add i32 [[TMP28]], [[N]] 809; VF-FOUR-CHECK-NEXT: [[TMP37:%.*]] = add i32 [[TMP29]], [[N]] 810; VF-FOUR-CHECK-NEXT: [[TMP38:%.*]] = add i32 [[TMP30]], [[N]] 811; VF-FOUR-CHECK-NEXT: [[TMP39:%.*]] = add i32 [[TMP31]], [[N]] 812; VF-FOUR-CHECK-NEXT: [[TMP40:%.*]] = sext i32 [[TMP32]] to i64 813; VF-FOUR-CHECK-NEXT: [[TMP41:%.*]] = sext i32 [[TMP33]] to i64 814; VF-FOUR-CHECK-NEXT: [[TMP42:%.*]] = sext i32 [[TMP34]] to i64 815; VF-FOUR-CHECK-NEXT: [[TMP43:%.*]] = sext i32 [[TMP35]] to i64 816; VF-FOUR-CHECK-NEXT: [[TMP44:%.*]] = sext i32 [[TMP36]] to i64 817; VF-FOUR-CHECK-NEXT: [[TMP45:%.*]] = sext i32 [[TMP37]] to i64 818; VF-FOUR-CHECK-NEXT: [[TMP46:%.*]] = sext i32 [[TMP38]] to i64 819; VF-FOUR-CHECK-NEXT: [[TMP47:%.*]] = sext i32 [[TMP39]] to i64 820; VF-FOUR-CHECK-NEXT: [[TMP48:%.*]] = getelementptr inbounds float, float* [[B:%.*]], i64 [[TMP40]] 821; VF-FOUR-CHECK-NEXT: [[TMP49:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP41]] 822; VF-FOUR-CHECK-NEXT: [[TMP50:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP42]] 823; VF-FOUR-CHECK-NEXT: [[TMP51:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP43]] 824; VF-FOUR-CHECK-NEXT: [[TMP52:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP44]] 825; VF-FOUR-CHECK-NEXT: [[TMP53:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP45]] 826; VF-FOUR-CHECK-NEXT: [[TMP54:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP46]] 827; VF-FOUR-CHECK-NEXT: [[TMP55:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP47]] 828; VF-FOUR-CHECK-NEXT: [[TMP56:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 0 829; VF-FOUR-CHECK-NEXT: [[TMP57:%.*]] = getelementptr inbounds float, float* [[TMP56]], i32 -3 830; VF-FOUR-CHECK-NEXT: [[TMP58:%.*]] = bitcast float* [[TMP57]] to <4 x float>* 831; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, <4 x float>* [[TMP58]], align 4 832; 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> 833; VF-FOUR-CHECK-NEXT: [[TMP59:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 -4 834; VF-FOUR-CHECK-NEXT: [[TMP60:%.*]] = getelementptr inbounds float, float* [[TMP59]], i32 -3 835; VF-FOUR-CHECK-NEXT: [[TMP61:%.*]] = bitcast float* [[TMP60]] to <4 x float>* 836; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD2:%.*]] = load <4 x float>, <4 x float>* [[TMP61]], align 4 837; 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> 838; VF-FOUR-CHECK-NEXT: [[TMP62:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 -8 839; VF-FOUR-CHECK-NEXT: [[TMP63:%.*]] = getelementptr inbounds float, float* [[TMP62]], i32 -3 840; VF-FOUR-CHECK-NEXT: [[TMP64:%.*]] = bitcast float* [[TMP63]] to <4 x float>* 841; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD4:%.*]] = load <4 x float>, <4 x float>* [[TMP64]], align 4 842; 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> 843; VF-FOUR-CHECK-NEXT: [[TMP65:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 -12 844; VF-FOUR-CHECK-NEXT: [[TMP66:%.*]] = getelementptr inbounds float, float* [[TMP65]], i32 -3 845; VF-FOUR-CHECK-NEXT: [[TMP67:%.*]] = bitcast float* [[TMP66]] to <4 x float>* 846; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD6:%.*]] = load <4 x float>, <4 x float>* [[TMP67]], align 4 847; 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> 848; VF-FOUR-CHECK-NEXT: [[TMP68:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 -16 849; VF-FOUR-CHECK-NEXT: [[TMP69:%.*]] = getelementptr inbounds float, float* [[TMP68]], i32 -3 850; VF-FOUR-CHECK-NEXT: [[TMP70:%.*]] = bitcast float* [[TMP69]] to <4 x float>* 851; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD8:%.*]] = load <4 x float>, <4 x float>* [[TMP70]], align 4 852; 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> 853; VF-FOUR-CHECK-NEXT: [[TMP71:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 -20 854; VF-FOUR-CHECK-NEXT: [[TMP72:%.*]] = getelementptr inbounds float, float* [[TMP71]], i32 -3 855; VF-FOUR-CHECK-NEXT: [[TMP73:%.*]] = bitcast float* [[TMP72]] to <4 x float>* 856; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD10:%.*]] = load <4 x float>, <4 x float>* [[TMP73]], align 4 857; 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> 858; VF-FOUR-CHECK-NEXT: [[TMP74:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 -24 859; VF-FOUR-CHECK-NEXT: [[TMP75:%.*]] = getelementptr inbounds float, float* [[TMP74]], i32 -3 860; VF-FOUR-CHECK-NEXT: [[TMP76:%.*]] = bitcast float* [[TMP75]] to <4 x float>* 861; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD12:%.*]] = load <4 x float>, <4 x float>* [[TMP76]], align 4 862; 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> 863; VF-FOUR-CHECK-NEXT: [[TMP77:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 -28 864; VF-FOUR-CHECK-NEXT: [[TMP78:%.*]] = getelementptr inbounds float, float* [[TMP77]], i32 -3 865; VF-FOUR-CHECK-NEXT: [[TMP79:%.*]] = bitcast float* [[TMP78]] to <4 x float>* 866; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD14:%.*]] = load <4 x float>, <4 x float>* [[TMP79]], align 4 867; 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> 868; VF-FOUR-CHECK-NEXT: [[TMP80:%.*]] = fadd fast <4 x float> [[REVERSE]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 869; VF-FOUR-CHECK-NEXT: [[TMP81:%.*]] = fadd fast <4 x float> [[REVERSE3]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 870; VF-FOUR-CHECK-NEXT: [[TMP82:%.*]] = fadd fast <4 x float> [[REVERSE5]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 871; VF-FOUR-CHECK-NEXT: [[TMP83:%.*]] = fadd fast <4 x float> [[REVERSE7]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 872; VF-FOUR-CHECK-NEXT: [[TMP84:%.*]] = fadd fast <4 x float> [[REVERSE9]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 873; VF-FOUR-CHECK-NEXT: [[TMP85:%.*]] = fadd fast <4 x float> [[REVERSE11]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 874; VF-FOUR-CHECK-NEXT: [[TMP86:%.*]] = fadd fast <4 x float> [[REVERSE13]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 875; VF-FOUR-CHECK-NEXT: [[TMP87:%.*]] = fadd fast <4 x float> [[REVERSE15]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 876; VF-FOUR-CHECK-NEXT: [[TMP88:%.*]] = getelementptr inbounds float, float* [[A:%.*]], i64 [[TMP16]] 877; VF-FOUR-CHECK-NEXT: [[TMP89:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP17]] 878; VF-FOUR-CHECK-NEXT: [[TMP90:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP18]] 879; VF-FOUR-CHECK-NEXT: [[TMP91:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP19]] 880; VF-FOUR-CHECK-NEXT: [[TMP92:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP20]] 881; VF-FOUR-CHECK-NEXT: [[TMP93:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP21]] 882; VF-FOUR-CHECK-NEXT: [[TMP94:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP22]] 883; VF-FOUR-CHECK-NEXT: [[TMP95:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP23]] 884; VF-FOUR-CHECK-NEXT: [[TMP96:%.*]] = getelementptr inbounds float, float* [[TMP88]], i32 0 885; VF-FOUR-CHECK-NEXT: [[TMP97:%.*]] = bitcast float* [[TMP96]] to <4 x float>* 886; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP80]], <4 x float>* [[TMP97]], align 4 887; VF-FOUR-CHECK-NEXT: [[TMP98:%.*]] = getelementptr inbounds float, float* [[TMP88]], i32 4 888; VF-FOUR-CHECK-NEXT: [[TMP99:%.*]] = bitcast float* [[TMP98]] to <4 x float>* 889; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP81]], <4 x float>* [[TMP99]], align 4 890; VF-FOUR-CHECK-NEXT: [[TMP100:%.*]] = getelementptr inbounds float, float* [[TMP88]], i32 8 891; VF-FOUR-CHECK-NEXT: [[TMP101:%.*]] = bitcast float* [[TMP100]] to <4 x float>* 892; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP82]], <4 x float>* [[TMP101]], align 4 893; VF-FOUR-CHECK-NEXT: [[TMP102:%.*]] = getelementptr inbounds float, float* [[TMP88]], i32 12 894; VF-FOUR-CHECK-NEXT: [[TMP103:%.*]] = bitcast float* [[TMP102]] to <4 x float>* 895; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP83]], <4 x float>* [[TMP103]], align 4 896; VF-FOUR-CHECK-NEXT: [[TMP104:%.*]] = getelementptr inbounds float, float* [[TMP88]], i32 16 897; VF-FOUR-CHECK-NEXT: [[TMP105:%.*]] = bitcast float* [[TMP104]] to <4 x float>* 898; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP84]], <4 x float>* [[TMP105]], align 4 899; VF-FOUR-CHECK-NEXT: [[TMP106:%.*]] = getelementptr inbounds float, float* [[TMP88]], i32 20 900; VF-FOUR-CHECK-NEXT: [[TMP107:%.*]] = bitcast float* [[TMP106]] to <4 x float>* 901; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP85]], <4 x float>* [[TMP107]], align 4 902; VF-FOUR-CHECK-NEXT: [[TMP108:%.*]] = getelementptr inbounds float, float* [[TMP88]], i32 24 903; VF-FOUR-CHECK-NEXT: [[TMP109:%.*]] = bitcast float* [[TMP108]] to <4 x float>* 904; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP86]], <4 x float>* [[TMP109]], align 4 905; VF-FOUR-CHECK-NEXT: [[TMP110:%.*]] = getelementptr inbounds float, float* [[TMP88]], i32 28 906; VF-FOUR-CHECK-NEXT: [[TMP111:%.*]] = bitcast float* [[TMP110]] to <4 x float>* 907; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP87]], <4 x float>* [[TMP111]], align 4 908; VF-FOUR-CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 32 909; VF-FOUR-CHECK-NEXT: [[TMP112:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] 910; VF-FOUR-CHECK-NEXT: br i1 [[TMP112]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOPID_MV_CM:![0-9]+]] 911; VF-FOUR-CHECK: middle.block: 912; VF-FOUR-CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC]] 913; VF-FOUR-CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]] 914; VF-FOUR-CHECK: vec.epilog.iter.check: 915; VF-FOUR-CHECK-NEXT: [[IND_END19:%.*]] = trunc i64 [[N_VEC]] to i32 916; VF-FOUR-CHECK-NEXT: [[N_VEC_REMAINING:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_VEC]] 917; VF-FOUR-CHECK-NEXT: [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 4 918; VF-FOUR-CHECK-NEXT: br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]] 919; VF-FOUR-CHECK: vec.epilog.ph: 920; VF-FOUR-CHECK-NEXT: [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ] 921; VF-FOUR-CHECK-NEXT: [[N_MOD_VF16:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 4 922; VF-FOUR-CHECK-NEXT: [[N_VEC17:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF16]] 923; VF-FOUR-CHECK-NEXT: [[IND_END:%.*]] = trunc i64 [[N_VEC17]] to i32 924; VF-FOUR-CHECK-NEXT: br label [[VEC_EPILOG_VECTOR_BODY:%.*]] 925; VF-FOUR-CHECK: vec.epilog.vector.body: 926; VF-FOUR-CHECK-NEXT: [[OFFSET_IDX23:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT26:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ] 927; VF-FOUR-CHECK-NEXT: [[OFFSET_IDX22:%.*]] = trunc i64 [[OFFSET_IDX23]] to i32 928; VF-FOUR-CHECK-NEXT: [[TMP113:%.*]] = add i32 [[OFFSET_IDX22]], 0 929; VF-FOUR-CHECK-NEXT: [[TMP114:%.*]] = add i64 [[OFFSET_IDX23]], 0 930; VF-FOUR-CHECK-NEXT: [[TMP115:%.*]] = xor i32 [[TMP113]], -1 931; VF-FOUR-CHECK-NEXT: [[TMP116:%.*]] = add i32 [[TMP115]], [[N]] 932; VF-FOUR-CHECK-NEXT: [[TMP117:%.*]] = sext i32 [[TMP116]] to i64 933; VF-FOUR-CHECK-NEXT: [[TMP118:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP117]] 934; VF-FOUR-CHECK-NEXT: [[TMP119:%.*]] = getelementptr inbounds float, float* [[TMP118]], i32 0 935; VF-FOUR-CHECK-NEXT: [[TMP120:%.*]] = getelementptr inbounds float, float* [[TMP119]], i32 -3 936; VF-FOUR-CHECK-NEXT: [[TMP121:%.*]] = bitcast float* [[TMP120]] to <4 x float>* 937; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD24:%.*]] = load <4 x float>, <4 x float>* [[TMP121]], align 4 938; 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> 939; VF-FOUR-CHECK-NEXT: [[TMP122:%.*]] = fadd fast <4 x float> [[REVERSE25]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 940; VF-FOUR-CHECK-NEXT: [[TMP123:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP114]] 941; VF-FOUR-CHECK-NEXT: [[TMP124:%.*]] = getelementptr inbounds float, float* [[TMP123]], i32 0 942; VF-FOUR-CHECK-NEXT: [[TMP125:%.*]] = bitcast float* [[TMP124]] to <4 x float>* 943; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP122]], <4 x float>* [[TMP125]], align 4 944; VF-FOUR-CHECK-NEXT: [[INDEX_NEXT26]] = add nuw i64 [[OFFSET_IDX23]], 4 945; VF-FOUR-CHECK-NEXT: [[TMP126:%.*]] = icmp eq i64 [[INDEX_NEXT26]], [[N_VEC17]] 946; VF-FOUR-CHECK-NEXT: br i1 [[TMP126]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOPID_EV_CM:![0-9]+]] 947; VF-FOUR-CHECK: vec.epilog.middle.block: 948; VF-FOUR-CHECK-NEXT: [[CMP_N20:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC17]] 949; VF-FOUR-CHECK-NEXT: br i1 [[CMP_N20]], label [[FOR_END_LOOPEXIT]], label [[VEC_EPILOG_SCALAR_PH]] 950; VF-FOUR-CHECK: vec.epilog.scalar.ph: 951; 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]] ] 952; VF-FOUR-CHECK-NEXT: [[BC_RESUME_VAL18:%.*]] = phi i32 [ [[IND_END]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[IND_END19]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ITER_CHECK]] ] 953; VF-FOUR-CHECK-NEXT: br label [[FOR_BODY:%.*]] 954; VF-FOUR-CHECK: for.body: 955; VF-FOUR-CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ] 956; VF-FOUR-CHECK-NEXT: [[I_014:%.*]] = phi i32 [ [[BC_RESUME_VAL18]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ] 957; VF-FOUR-CHECK-NEXT: [[TMP127:%.*]] = xor i32 [[I_014]], -1 958; VF-FOUR-CHECK-NEXT: [[SUB2:%.*]] = add i32 [[TMP127]], [[N]] 959; VF-FOUR-CHECK-NEXT: [[IDXPROM:%.*]] = sext i32 [[SUB2]] to i64 960; VF-FOUR-CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[IDXPROM]] 961; VF-FOUR-CHECK-NEXT: [[TMP128:%.*]] = load float, float* [[ARRAYIDX]], align 4 962; VF-FOUR-CHECK-NEXT: [[CONV3:%.*]] = fadd fast float [[TMP128]], 1.000000e+00 963; VF-FOUR-CHECK-NEXT: [[ARRAYIDX5:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDVARS_IV]] 964; VF-FOUR-CHECK-NEXT: store float [[CONV3]], float* [[ARRAYIDX5]], align 4 965; VF-FOUR-CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1 966; VF-FOUR-CHECK-NEXT: [[INC]] = add nuw nsw i32 [[I_014]], 1 967; VF-FOUR-CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i64 [[INDVARS_IV_NEXT]], [[WIDE_TRIP_COUNT]] 968; VF-FOUR-CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_BODY]], label [[FOR_END_LOOPEXIT]], !llvm.loop [[LOOP7:![0-9]+]] 969; VF-FOUR-CHECK: for.end.loopexit: 970; VF-FOUR-CHECK-NEXT: br label [[FOR_END]] 971; VF-FOUR-CHECK: for.end: 972; VF-FOUR-CHECK-NEXT: ret i32 0 973; 974entry: 975 %cmp1 = icmp sgt i32 %n, 1 976 br i1 %cmp1, label %for.body.preheader, label %for.end 977 978for.body.preheader: ; preds = %entry 979 %0 = add i32 %n, -1 980 %wide.trip.count = zext i32 %0 to i64 981 br label %for.body 982 983for.body: ; preds = %for.body.preheader, %for.body 984 %indvars.iv = phi i64 [ 0, %for.body.preheader ], [ %indvars.iv.next, %for.body ] 985 %i.014 = phi i32 [ 0, %for.body.preheader ], [ %inc, %for.body ] 986 %1 = xor i32 %i.014, -1 987 %sub2 = add i32 %1, %n 988 %idxprom = sext i32 %sub2 to i64 989 %arrayidx = getelementptr inbounds float, float* %B, i64 %idxprom 990 %2 = load float, float* %arrayidx, align 4 991 %conv3 = fadd fast float %2, 1.000000e+00 992 %arrayidx5 = getelementptr inbounds float, float* %A, i64 %indvars.iv 993 store float %conv3, float* %arrayidx5, align 4 994 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 995 %inc = add nuw nsw i32 %i.014, 1 996 %exitcond = icmp ne i64 %indvars.iv.next, %wide.trip.count 997 br i1 %exitcond, label %for.body, label %for.end.loopexit 998 999for.end.loopexit: ; preds = %for.body 1000 br label %for.end 1001 1002for.end: ; preds = %for.end.loopexit, %entry 1003 ret i32 0 1004} 1005 1006; VF-TWO-CHECK-DAG: [[LOOPID_MV]] = distinct !{[[LOOPID_MV]], [[LOOPID_DISABLE_VECT:!.*]]} 1007; VF-TWO-CHECK-DAG: [[LOOPID_EV]] = distinct !{[[LOOPID_EV]], [[LOOPID_DISABLE_VECT]], [[LOOPID_DISABLE_UNROLL:!.*]]} 1008; VF-TWO-CHECK-DAG: [[LOOPID_DISABLE_VECT]] = [[DISABLE_VECT_STR:!{!"llvm.loop.isvectorized".*}.*]] 1009; VF-TWO-CHECK-DAG: [[LOOPID_DISABLE_UNROLL]] = [[DISABLE_UNROLL_STR:!{!"llvm.loop.unroll.runtime.disable"}.*]] 1010; 1011; VF-FOUR-CHECK-DAG: [[LOOPID_MV_CM]] = distinct !{[[LOOPID_MV_CM]], [[LOOPID_DISABLE_VECT_CM:!.*]]} 1012; VF-FOUR-CHECK-DAG: [[LOOPID_EV_CM]] = distinct !{[[LOOPID_EV_CM]], [[LOOPID_DISABLE_VECT_CM]], [[LOOPID_DISABLE_UNROLL_CM:!.*]]} 1013; VF-FOUR-CHECK-DAG: [[LOOPID_DISABLE_VECT_CM]] = [[DISABLE_VECT_STR_CM:!{!"llvm.loop.isvectorized".*}.*]] 1014; VF-FOUR-CHECK-DAG: [[LOOPID_DISABLE_UNROLL_CM]] = [[DISABLE_UNROLL_STR_CM:!{!"llvm.loop.unroll.runtime.disable"}.*]] 1015; 1016attributes #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" } 1017