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: [[AA1:%.*]] = bitcast float* [[AA:%.*]] to i8* 13; VF-TWO-CHECK-NEXT: [[BB3:%.*]] = bitcast float* [[BB:%.*]] to i8* 14; VF-TWO-CHECK-NEXT: [[CC6:%.*]] = bitcast float* [[CC:%.*]] to i8* 15; VF-TWO-CHECK-NEXT: [[CMP1:%.*]] = icmp sgt i32 [[N:%.*]], 0 16; VF-TWO-CHECK-NEXT: br i1 [[CMP1]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_END:%.*]] 17; VF-TWO-CHECK: iter.check: 18; VF-TWO-CHECK-NEXT: [[WIDE_TRIP_COUNT:%.*]] = zext i32 [[N]] to i64 19; VF-TWO-CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 2 20; VF-TWO-CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]] 21; VF-TWO-CHECK: vector.memcheck: 22; VF-TWO-CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr float, float* [[AA]], i64 [[WIDE_TRIP_COUNT]] 23; VF-TWO-CHECK-NEXT: [[SCEVGEP2:%.*]] = bitcast float* [[SCEVGEP]] to i8* 24; VF-TWO-CHECK-NEXT: [[SCEVGEP4:%.*]] = getelementptr float, float* [[BB]], i64 [[WIDE_TRIP_COUNT]] 25; VF-TWO-CHECK-NEXT: [[SCEVGEP45:%.*]] = bitcast float* [[SCEVGEP4]] to i8* 26; VF-TWO-CHECK-NEXT: [[SCEVGEP7:%.*]] = getelementptr float, float* [[CC]], i64 [[WIDE_TRIP_COUNT]] 27; VF-TWO-CHECK-NEXT: [[SCEVGEP78:%.*]] = bitcast float* [[SCEVGEP7]] to i8* 28; VF-TWO-CHECK-NEXT: [[BOUND0:%.*]] = icmp ult i8* [[AA1]], [[SCEVGEP45]] 29; VF-TWO-CHECK-NEXT: [[BOUND1:%.*]] = icmp ult i8* [[BB3]], [[SCEVGEP2]] 30; VF-TWO-CHECK-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]] 31; VF-TWO-CHECK-NEXT: [[BOUND09:%.*]] = icmp ult i8* [[AA1]], [[SCEVGEP78]] 32; VF-TWO-CHECK-NEXT: [[BOUND110:%.*]] = icmp ult i8* [[CC6]], [[SCEVGEP2]] 33; VF-TWO-CHECK-NEXT: [[FOUND_CONFLICT11:%.*]] = and i1 [[BOUND09]], [[BOUND110]] 34; VF-TWO-CHECK-NEXT: [[CONFLICT_RDX:%.*]] = or i1 [[FOUND_CONFLICT]], [[FOUND_CONFLICT11]] 35; VF-TWO-CHECK-NEXT: [[MEMCHECK_CONFLICT:%.*]] = and i1 [[CONFLICT_RDX]], true 36; VF-TWO-CHECK-NEXT: br i1 [[MEMCHECK_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]] 37; VF-TWO-CHECK: vector.main.loop.iter.check: 38; VF-TWO-CHECK-NEXT: [[MIN_ITERS_CHECK12:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 48 39; VF-TWO-CHECK-NEXT: br i1 [[MIN_ITERS_CHECK12]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]] 40; VF-TWO-CHECK: vector.ph: 41; VF-TWO-CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 48 42; VF-TWO-CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF]] 43; VF-TWO-CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 44; VF-TWO-CHECK: vector.body: 45; VF-TWO-CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 46; VF-TWO-CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0 47; VF-TWO-CHECK-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 4 48; VF-TWO-CHECK-NEXT: [[TMP2:%.*]] = add i64 [[INDEX]], 8 49; VF-TWO-CHECK-NEXT: [[TMP3:%.*]] = add i64 [[INDEX]], 12 50; VF-TWO-CHECK-NEXT: [[TMP4:%.*]] = add i64 [[INDEX]], 16 51; VF-TWO-CHECK-NEXT: [[TMP5:%.*]] = add i64 [[INDEX]], 20 52; VF-TWO-CHECK-NEXT: [[TMP6:%.*]] = add i64 [[INDEX]], 24 53; VF-TWO-CHECK-NEXT: [[TMP7:%.*]] = add i64 [[INDEX]], 28 54; VF-TWO-CHECK-NEXT: [[TMP8:%.*]] = add i64 [[INDEX]], 32 55; VF-TWO-CHECK-NEXT: [[TMP9:%.*]] = add i64 [[INDEX]], 36 56; VF-TWO-CHECK-NEXT: [[TMP10:%.*]] = add i64 [[INDEX]], 40 57; VF-TWO-CHECK-NEXT: [[TMP11:%.*]] = add i64 [[INDEX]], 44 58; VF-TWO-CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP0]] 59; VF-TWO-CHECK-NEXT: [[TMP13:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP1]] 60; VF-TWO-CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP2]] 61; VF-TWO-CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP3]] 62; VF-TWO-CHECK-NEXT: [[TMP16:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP4]] 63; VF-TWO-CHECK-NEXT: [[TMP17:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP5]] 64; VF-TWO-CHECK-NEXT: [[TMP18:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP6]] 65; VF-TWO-CHECK-NEXT: [[TMP19:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP7]] 66; VF-TWO-CHECK-NEXT: [[TMP20:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP8]] 67; VF-TWO-CHECK-NEXT: [[TMP21:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP9]] 68; VF-TWO-CHECK-NEXT: [[TMP22:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP10]] 69; VF-TWO-CHECK-NEXT: [[TMP23:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP11]] 70; VF-TWO-CHECK-NEXT: [[TMP24:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 0 71; VF-TWO-CHECK-NEXT: [[TMP25:%.*]] = bitcast float* [[TMP24]] to <4 x float>* 72; VF-TWO-CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, <4 x float>* [[TMP25]], align 4 73; VF-TWO-CHECK-NEXT: [[TMP26:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 4 74; VF-TWO-CHECK-NEXT: [[TMP27:%.*]] = bitcast float* [[TMP26]] to <4 x float>* 75; VF-TWO-CHECK-NEXT: [[WIDE_LOAD13:%.*]] = load <4 x float>, <4 x float>* [[TMP27]], align 4 76; VF-TWO-CHECK-NEXT: [[TMP28:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 8 77; VF-TWO-CHECK-NEXT: [[TMP29:%.*]] = bitcast float* [[TMP28]] to <4 x float>* 78; VF-TWO-CHECK-NEXT: [[WIDE_LOAD14:%.*]] = load <4 x float>, <4 x float>* [[TMP29]], align 4 79; VF-TWO-CHECK-NEXT: [[TMP30:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 12 80; VF-TWO-CHECK-NEXT: [[TMP31:%.*]] = bitcast float* [[TMP30]] to <4 x float>* 81; VF-TWO-CHECK-NEXT: [[WIDE_LOAD15:%.*]] = load <4 x float>, <4 x float>* [[TMP31]], align 4 82; VF-TWO-CHECK-NEXT: [[TMP32:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 16 83; VF-TWO-CHECK-NEXT: [[TMP33:%.*]] = bitcast float* [[TMP32]] to <4 x float>* 84; VF-TWO-CHECK-NEXT: [[WIDE_LOAD16:%.*]] = load <4 x float>, <4 x float>* [[TMP33]], align 4 85; VF-TWO-CHECK-NEXT: [[TMP34:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 20 86; VF-TWO-CHECK-NEXT: [[TMP35:%.*]] = bitcast float* [[TMP34]] to <4 x float>* 87; VF-TWO-CHECK-NEXT: [[WIDE_LOAD17:%.*]] = load <4 x float>, <4 x float>* [[TMP35]], align 4 88; VF-TWO-CHECK-NEXT: [[TMP36:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 24 89; VF-TWO-CHECK-NEXT: [[TMP37:%.*]] = bitcast float* [[TMP36]] to <4 x float>* 90; VF-TWO-CHECK-NEXT: [[WIDE_LOAD18:%.*]] = load <4 x float>, <4 x float>* [[TMP37]], align 4 91; VF-TWO-CHECK-NEXT: [[TMP38:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 28 92; VF-TWO-CHECK-NEXT: [[TMP39:%.*]] = bitcast float* [[TMP38]] to <4 x float>* 93; VF-TWO-CHECK-NEXT: [[WIDE_LOAD19:%.*]] = load <4 x float>, <4 x float>* [[TMP39]], align 4 94; VF-TWO-CHECK-NEXT: [[TMP40:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 32 95; VF-TWO-CHECK-NEXT: [[TMP41:%.*]] = bitcast float* [[TMP40]] to <4 x float>* 96; VF-TWO-CHECK-NEXT: [[WIDE_LOAD20:%.*]] = load <4 x float>, <4 x float>* [[TMP41]], align 4 97; VF-TWO-CHECK-NEXT: [[TMP42:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 36 98; VF-TWO-CHECK-NEXT: [[TMP43:%.*]] = bitcast float* [[TMP42]] to <4 x float>* 99; VF-TWO-CHECK-NEXT: [[WIDE_LOAD21:%.*]] = load <4 x float>, <4 x float>* [[TMP43]], align 4 100; VF-TWO-CHECK-NEXT: [[TMP44:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 40 101; VF-TWO-CHECK-NEXT: [[TMP45:%.*]] = bitcast float* [[TMP44]] to <4 x float>* 102; VF-TWO-CHECK-NEXT: [[WIDE_LOAD22:%.*]] = load <4 x float>, <4 x float>* [[TMP45]], align 4 103; VF-TWO-CHECK-NEXT: [[TMP46:%.*]] = getelementptr inbounds float, float* [[TMP12]], i32 44 104; VF-TWO-CHECK-NEXT: [[TMP47:%.*]] = bitcast float* [[TMP46]] to <4 x float>* 105; VF-TWO-CHECK-NEXT: [[WIDE_LOAD23:%.*]] = load <4 x float>, <4 x float>* [[TMP47]], align 4 106; VF-TWO-CHECK-NEXT: [[TMP48:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP0]] 107; VF-TWO-CHECK-NEXT: [[TMP49:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP1]] 108; VF-TWO-CHECK-NEXT: [[TMP50:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP2]] 109; VF-TWO-CHECK-NEXT: [[TMP51:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP3]] 110; VF-TWO-CHECK-NEXT: [[TMP52:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP4]] 111; VF-TWO-CHECK-NEXT: [[TMP53:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP5]] 112; VF-TWO-CHECK-NEXT: [[TMP54:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP6]] 113; VF-TWO-CHECK-NEXT: [[TMP55:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP7]] 114; VF-TWO-CHECK-NEXT: [[TMP56:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP8]] 115; VF-TWO-CHECK-NEXT: [[TMP57:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP9]] 116; VF-TWO-CHECK-NEXT: [[TMP58:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP10]] 117; VF-TWO-CHECK-NEXT: [[TMP59:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP11]] 118; VF-TWO-CHECK-NEXT: [[TMP60:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 0 119; VF-TWO-CHECK-NEXT: [[TMP61:%.*]] = bitcast float* [[TMP60]] to <4 x float>* 120; VF-TWO-CHECK-NEXT: [[WIDE_LOAD24:%.*]] = load <4 x float>, <4 x float>* [[TMP61]], align 4 121; VF-TWO-CHECK-NEXT: [[TMP62:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 4 122; VF-TWO-CHECK-NEXT: [[TMP63:%.*]] = bitcast float* [[TMP62]] to <4 x float>* 123; VF-TWO-CHECK-NEXT: [[WIDE_LOAD25:%.*]] = load <4 x float>, <4 x float>* [[TMP63]], align 4 124; VF-TWO-CHECK-NEXT: [[TMP64:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 8 125; VF-TWO-CHECK-NEXT: [[TMP65:%.*]] = bitcast float* [[TMP64]] to <4 x float>* 126; VF-TWO-CHECK-NEXT: [[WIDE_LOAD26:%.*]] = load <4 x float>, <4 x float>* [[TMP65]], align 4 127; VF-TWO-CHECK-NEXT: [[TMP66:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 12 128; VF-TWO-CHECK-NEXT: [[TMP67:%.*]] = bitcast float* [[TMP66]] to <4 x float>* 129; VF-TWO-CHECK-NEXT: [[WIDE_LOAD27:%.*]] = load <4 x float>, <4 x float>* [[TMP67]], align 4 130; VF-TWO-CHECK-NEXT: [[TMP68:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 16 131; VF-TWO-CHECK-NEXT: [[TMP69:%.*]] = bitcast float* [[TMP68]] to <4 x float>* 132; VF-TWO-CHECK-NEXT: [[WIDE_LOAD28:%.*]] = load <4 x float>, <4 x float>* [[TMP69]], align 4 133; VF-TWO-CHECK-NEXT: [[TMP70:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 20 134; VF-TWO-CHECK-NEXT: [[TMP71:%.*]] = bitcast float* [[TMP70]] to <4 x float>* 135; VF-TWO-CHECK-NEXT: [[WIDE_LOAD29:%.*]] = load <4 x float>, <4 x float>* [[TMP71]], align 4 136; VF-TWO-CHECK-NEXT: [[TMP72:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 24 137; VF-TWO-CHECK-NEXT: [[TMP73:%.*]] = bitcast float* [[TMP72]] to <4 x float>* 138; VF-TWO-CHECK-NEXT: [[WIDE_LOAD30:%.*]] = load <4 x float>, <4 x float>* [[TMP73]], align 4 139; VF-TWO-CHECK-NEXT: [[TMP74:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 28 140; VF-TWO-CHECK-NEXT: [[TMP75:%.*]] = bitcast float* [[TMP74]] to <4 x float>* 141; VF-TWO-CHECK-NEXT: [[WIDE_LOAD31:%.*]] = load <4 x float>, <4 x float>* [[TMP75]], align 4 142; VF-TWO-CHECK-NEXT: [[TMP76:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 32 143; VF-TWO-CHECK-NEXT: [[TMP77:%.*]] = bitcast float* [[TMP76]] to <4 x float>* 144; VF-TWO-CHECK-NEXT: [[WIDE_LOAD32:%.*]] = load <4 x float>, <4 x float>* [[TMP77]], align 4 145; VF-TWO-CHECK-NEXT: [[TMP78:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 36 146; VF-TWO-CHECK-NEXT: [[TMP79:%.*]] = bitcast float* [[TMP78]] to <4 x float>* 147; VF-TWO-CHECK-NEXT: [[WIDE_LOAD33:%.*]] = load <4 x float>, <4 x float>* [[TMP79]], align 4 148; VF-TWO-CHECK-NEXT: [[TMP80:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 40 149; VF-TWO-CHECK-NEXT: [[TMP81:%.*]] = bitcast float* [[TMP80]] to <4 x float>* 150; VF-TWO-CHECK-NEXT: [[WIDE_LOAD34:%.*]] = load <4 x float>, <4 x float>* [[TMP81]], align 4 151; VF-TWO-CHECK-NEXT: [[TMP82:%.*]] = getelementptr inbounds float, float* [[TMP48]], i32 44 152; VF-TWO-CHECK-NEXT: [[TMP83:%.*]] = bitcast float* [[TMP82]] to <4 x float>* 153; VF-TWO-CHECK-NEXT: [[WIDE_LOAD35:%.*]] = load <4 x float>, <4 x float>* [[TMP83]], align 4 154; VF-TWO-CHECK-NEXT: [[TMP84:%.*]] = fadd fast <4 x float> [[WIDE_LOAD]], [[WIDE_LOAD24]] 155; VF-TWO-CHECK-NEXT: [[TMP85:%.*]] = fadd fast <4 x float> [[WIDE_LOAD13]], [[WIDE_LOAD25]] 156; VF-TWO-CHECK-NEXT: [[TMP86:%.*]] = fadd fast <4 x float> [[WIDE_LOAD14]], [[WIDE_LOAD26]] 157; VF-TWO-CHECK-NEXT: [[TMP87:%.*]] = fadd fast <4 x float> [[WIDE_LOAD15]], [[WIDE_LOAD27]] 158; VF-TWO-CHECK-NEXT: [[TMP88:%.*]] = fadd fast <4 x float> [[WIDE_LOAD16]], [[WIDE_LOAD28]] 159; VF-TWO-CHECK-NEXT: [[TMP89:%.*]] = fadd fast <4 x float> [[WIDE_LOAD17]], [[WIDE_LOAD29]] 160; VF-TWO-CHECK-NEXT: [[TMP90:%.*]] = fadd fast <4 x float> [[WIDE_LOAD18]], [[WIDE_LOAD30]] 161; VF-TWO-CHECK-NEXT: [[TMP91:%.*]] = fadd fast <4 x float> [[WIDE_LOAD19]], [[WIDE_LOAD31]] 162; VF-TWO-CHECK-NEXT: [[TMP92:%.*]] = fadd fast <4 x float> [[WIDE_LOAD20]], [[WIDE_LOAD32]] 163; VF-TWO-CHECK-NEXT: [[TMP93:%.*]] = fadd fast <4 x float> [[WIDE_LOAD21]], [[WIDE_LOAD33]] 164; VF-TWO-CHECK-NEXT: [[TMP94:%.*]] = fadd fast <4 x float> [[WIDE_LOAD22]], [[WIDE_LOAD34]] 165; VF-TWO-CHECK-NEXT: [[TMP95:%.*]] = fadd fast <4 x float> [[WIDE_LOAD23]], [[WIDE_LOAD35]] 166; VF-TWO-CHECK-NEXT: [[TMP96:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP0]] 167; VF-TWO-CHECK-NEXT: [[TMP97:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP1]] 168; VF-TWO-CHECK-NEXT: [[TMP98:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP2]] 169; VF-TWO-CHECK-NEXT: [[TMP99:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP3]] 170; VF-TWO-CHECK-NEXT: [[TMP100:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP4]] 171; VF-TWO-CHECK-NEXT: [[TMP101:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP5]] 172; VF-TWO-CHECK-NEXT: [[TMP102:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP6]] 173; VF-TWO-CHECK-NEXT: [[TMP103:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP7]] 174; VF-TWO-CHECK-NEXT: [[TMP104:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP8]] 175; VF-TWO-CHECK-NEXT: [[TMP105:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP9]] 176; VF-TWO-CHECK-NEXT: [[TMP106:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP10]] 177; VF-TWO-CHECK-NEXT: [[TMP107:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP11]] 178; VF-TWO-CHECK-NEXT: [[TMP108:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 0 179; VF-TWO-CHECK-NEXT: [[TMP109:%.*]] = bitcast float* [[TMP108]] to <4 x float>* 180; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP84]], <4 x float>* [[TMP109]], align 4 181; VF-TWO-CHECK-NEXT: [[TMP110:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 4 182; VF-TWO-CHECK-NEXT: [[TMP111:%.*]] = bitcast float* [[TMP110]] to <4 x float>* 183; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP85]], <4 x float>* [[TMP111]], align 4 184; VF-TWO-CHECK-NEXT: [[TMP112:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 8 185; VF-TWO-CHECK-NEXT: [[TMP113:%.*]] = bitcast float* [[TMP112]] to <4 x float>* 186; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP86]], <4 x float>* [[TMP113]], align 4 187; VF-TWO-CHECK-NEXT: [[TMP114:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 12 188; VF-TWO-CHECK-NEXT: [[TMP115:%.*]] = bitcast float* [[TMP114]] to <4 x float>* 189; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP87]], <4 x float>* [[TMP115]], align 4 190; VF-TWO-CHECK-NEXT: [[TMP116:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 16 191; VF-TWO-CHECK-NEXT: [[TMP117:%.*]] = bitcast float* [[TMP116]] to <4 x float>* 192; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP88]], <4 x float>* [[TMP117]], align 4 193; VF-TWO-CHECK-NEXT: [[TMP118:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 20 194; VF-TWO-CHECK-NEXT: [[TMP119:%.*]] = bitcast float* [[TMP118]] to <4 x float>* 195; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP89]], <4 x float>* [[TMP119]], align 4 196; VF-TWO-CHECK-NEXT: [[TMP120:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 24 197; VF-TWO-CHECK-NEXT: [[TMP121:%.*]] = bitcast float* [[TMP120]] to <4 x float>* 198; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP90]], <4 x float>* [[TMP121]], align 4 199; VF-TWO-CHECK-NEXT: [[TMP122:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 28 200; VF-TWO-CHECK-NEXT: [[TMP123:%.*]] = bitcast float* [[TMP122]] to <4 x float>* 201; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP91]], <4 x float>* [[TMP123]], align 4 202; VF-TWO-CHECK-NEXT: [[TMP124:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 32 203; VF-TWO-CHECK-NEXT: [[TMP125:%.*]] = bitcast float* [[TMP124]] to <4 x float>* 204; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP92]], <4 x float>* [[TMP125]], align 4 205; VF-TWO-CHECK-NEXT: [[TMP126:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 36 206; VF-TWO-CHECK-NEXT: [[TMP127:%.*]] = bitcast float* [[TMP126]] to <4 x float>* 207; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP93]], <4 x float>* [[TMP127]], align 4 208; VF-TWO-CHECK-NEXT: [[TMP128:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 40 209; VF-TWO-CHECK-NEXT: [[TMP129:%.*]] = bitcast float* [[TMP128]] to <4 x float>* 210; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP94]], <4 x float>* [[TMP129]], align 4 211; VF-TWO-CHECK-NEXT: [[TMP130:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 44 212; VF-TWO-CHECK-NEXT: [[TMP131:%.*]] = bitcast float* [[TMP130]] to <4 x float>* 213; VF-TWO-CHECK-NEXT: store <4 x float> [[TMP95]], <4 x float>* [[TMP131]], align 4 214; VF-TWO-CHECK-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], 48 215; VF-TWO-CHECK-NEXT: [[TMP132:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] 216; VF-TWO-CHECK-NEXT: br i1 [[TMP132]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOPID_MV:!.*]] 217; VF-TWO-CHECK: middle.block: 218; VF-TWO-CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC]] 219; VF-TWO-CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]] 220; VF-TWO-CHECK: vec.epilog.iter.check: 221; VF-TWO-CHECK-NEXT: [[N_VEC_REMAINING:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_VEC]] 222; VF-TWO-CHECK-NEXT: [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 2 223; VF-TWO-CHECK-NEXT: br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[SCALAR_PH]], label [[VEC_EPILOG_PH]] 224; VF-TWO-CHECK: vec.epilog.ph: 225; VF-TWO-CHECK-NEXT: [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ] 226; VF-TWO-CHECK-NEXT: [[N_MOD_VF36:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 2 227; VF-TWO-CHECK-NEXT: [[N_VEC37:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF36]] 228; VF-TWO-CHECK-NEXT: br label [[VEC_EPILOG_BODY:%.*]] 229; VF-TWO-CHECK: vec.epilog.vector.body: 230; VF-TWO-CHECK-NEXT: [[INDEX38:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT39:%.*]], [[VEC_EPILOG_BODY]] ] 231; VF-TWO-CHECK-NEXT: [[TMP133:%.*]] = add i64 [[INDEX38]], 0 232; VF-TWO-CHECK-NEXT: [[TMP134:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[TMP133]] 233; VF-TWO-CHECK-NEXT: [[TMP135:%.*]] = getelementptr inbounds float, float* [[TMP134]], i32 0 234; VF-TWO-CHECK-NEXT: [[TMP136:%.*]] = bitcast float* [[TMP135]] to <2 x float>* 235; VF-TWO-CHECK-NEXT: [[WIDE_LOAD41:%.*]] = load <2 x float>, <2 x float>* [[TMP136]], align 4 236; VF-TWO-CHECK-NEXT: [[TMP137:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[TMP133]] 237; VF-TWO-CHECK-NEXT: [[TMP138:%.*]] = getelementptr inbounds float, float* [[TMP137]], i32 0 238; VF-TWO-CHECK-NEXT: [[TMP139:%.*]] = bitcast float* [[TMP138]] to <2 x float>* 239; VF-TWO-CHECK-NEXT: [[WIDE_LOAD42:%.*]] = load <2 x float>, <2 x float>* [[TMP139]], align 4 240; VF-TWO-CHECK-NEXT: [[TMP140:%.*]] = fadd fast <2 x float> [[WIDE_LOAD41]], [[WIDE_LOAD42]] 241; VF-TWO-CHECK-NEXT: [[TMP141:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[TMP133]] 242; VF-TWO-CHECK-NEXT: [[TMP142:%.*]] = getelementptr inbounds float, float* [[TMP141]], i32 0 243; VF-TWO-CHECK-NEXT: [[TMP143:%.*]] = bitcast float* [[TMP142]] to <2 x float>* 244; VF-TWO-CHECK-NEXT: store <2 x float> [[TMP140]], <2 x float>* [[TMP143]], align 4 245; VF-TWO-CHECK-NEXT: [[INDEX_NEXT39]] = add i64 [[INDEX38]], 2 246; VF-TWO-CHECK-NEXT: [[TMP144:%.*]] = icmp eq i64 [[INDEX_NEXT39]], [[N_VEC37]] 247; VF-TWO-CHECK-NEXT: br i1 [[TMP144]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_BODY]], !llvm.loop [[LOOPID_EV:!.*]] 248; VF-TWO-CHECK: vec.epilog.middle.block: 249; VF-TWO-CHECK-NEXT: [[CMP_N40:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC37]] 250; VF-TWO-CHECK-NEXT: br i1 [[CMP_N40]], label [[FOR_END_LOOPEXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]] 251; VF-TWO-CHECK: scalar.ph: 252; VF-TWO-CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC37]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MEMCHECK]] ], [ 0, [[FOR_BODY_PREHEADER]] ] 253; VF-TWO-CHECK-NEXT: br label [[FOR_BODY:%.*]] 254; VF-TWO-CHECK: for.body: 255; VF-TWO-CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ] 256; VF-TWO-CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[BB]], i64 [[INDVARS_IV]] 257; VF-TWO-CHECK-NEXT: [[TMP145:%.*]] = load float, float* [[ARRAYIDX]], align 4 258; VF-TWO-CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[CC]], i64 [[INDVARS_IV]] 259; VF-TWO-CHECK-NEXT: [[TMP146:%.*]] = load float, float* [[ARRAYIDX2]], align 4 260; VF-TWO-CHECK-NEXT: [[ADD:%.*]] = fadd fast float [[TMP145]], [[TMP146]] 261; VF-TWO-CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[AA]], i64 [[INDVARS_IV]] 262; VF-TWO-CHECK-NEXT: store float [[ADD]], float* [[ARRAYIDX4]], align 4 263; VF-TWO-CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1 264; VF-TWO-CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i64 [[INDVARS_IV_NEXT]], [[WIDE_TRIP_COUNT]] 265; VF-TWO-CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_BODY]], label [[FOR_END_LOOPEXIT_LOOPEXIT]], !llvm.loop [[LOOPID_MS:!.*]] 266; VF-TWO-CHECK: for.end.loopexit.loopexit: 267; VF-TWO-CHECK-NEXT: br label [[FOR_END_LOOPEXIT]] 268; VF-TWO-CHECK: for.end.loopexit: 269; VF-TWO-CHECK-NEXT: br label [[FOR_END]] 270; VF-TWO-CHECK: for.end: 271; VF-TWO-CHECK-NEXT: ret void 272; 273; VF-TWO-CHECK-DAG: [[LOOPID_MV]] = distinct !{[[LOOPID_MV]], [[LOOPID_DISABLE_VECT:!.*]]} 274; VF-TWO-CHECK-DAG: [[LOOPID_EV]] = distinct !{[[LOOPID_EV]], [[LOOPID_DISABLE_UNROLL:!.*]], [[LOOPID_DISABLE_VECT:!.*]]} 275; VF-TWO-CHECK-DAG: [[LOOPID_DISABLE_VECT]] = [[DISABLE_VECT_STR:!{!"llvm.loop.isvectorized".*}.*]] 276; VF-TWO-CHECK-DAG: [[LOOPID_DISABLE_UNROLL]] = [[DISABLE_UNROLL_STR:!{!"llvm.loop.unroll.runtime.disable"}.*]] 277 278 279entry: 280 %cmp1 = icmp sgt i32 %N, 0 281 br i1 %cmp1, label %for.body.preheader, label %for.end 282 283for.body.preheader: ; preds = %entry 284 %wide.trip.count = zext i32 %N to i64 285 br label %for.body 286 287for.body: ; preds = %for.body.preheader, %for.body 288 %indvars.iv = phi i64 [ 0, %for.body.preheader ], [ %indvars.iv.next, %for.body ] 289 %arrayidx = getelementptr inbounds float, float* %bb, i64 %indvars.iv 290 %0 = load float, float* %arrayidx, align 4 291 %arrayidx2 = getelementptr inbounds float, float* %cc, i64 %indvars.iv 292 %1 = load float, float* %arrayidx2, align 4 293 %add = fadd fast float %0, %1 294 %arrayidx4 = getelementptr inbounds float, float* %aa, i64 %indvars.iv 295 store float %add, float* %arrayidx4, align 4 296 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 297 %exitcond = icmp ne i64 %indvars.iv.next, %wide.trip.count 298 br i1 %exitcond, label %for.body, label %for.end.loopexit 299 300for.end.loopexit: ; preds = %for.body 301 br label %for.end 302 303for.end: ; preds = %for.end.loopexit, %entry 304 ret void 305} 306 307define dso_local signext i32 @f2(float* noalias %A, float* noalias %B, i32 signext %n) #0 { 308; VF-FOUR-CHECK-LABEL: @f2( 309; VF-FOUR-CHECK-NEXT: entry: 310; VF-FOUR-CHECK-NEXT: [[A1:%.*]] = bitcast float* [[A:%.*]] to i8* 311; VF-FOUR-CHECK-NEXT: [[CMP1:%.*]] = icmp sgt i32 [[N:%.*]], 1 312; VF-FOUR-CHECK-NEXT: br i1 [[CMP1]], label [[ITER_CHECK:%.*]], label [[FOR_END:%.*]] 313; VF-FOUR-CHECK: iter.check: 314; VF-FOUR-CHECK-NEXT: [[TMP0:%.*]] = add i32 [[N]], -1 315; VF-FOUR-CHECK-NEXT: [[WIDE_TRIP_COUNT:%.*]] = zext i32 [[TMP0]] to i64 316; VF-FOUR-CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 4 317; VF-FOUR-CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]] 318; VF-FOUR-CHECK: vector.scevcheck: 319; VF-FOUR-CHECK-NEXT: [[TMP1:%.*]] = add nsw i64 [[WIDE_TRIP_COUNT]], -1 320; VF-FOUR-CHECK-NEXT: [[TMP2:%.*]] = trunc i64 [[TMP1]] to i32 321; VF-FOUR-CHECK-NEXT: [[MUL:%.*]] = call { i32, i1 } @llvm.umul.with.overflow.i32(i32 1, i32 [[TMP2]]) 322; VF-FOUR-CHECK-NEXT: [[MUL_RESULT:%.*]] = extractvalue { i32, i1 } [[MUL]], 0 323; VF-FOUR-CHECK-NEXT: [[MUL_OVERFLOW:%.*]] = extractvalue { i32, i1 } [[MUL]], 1 324; VF-FOUR-CHECK-NEXT: [[TMP3:%.*]] = add i32 [[TMP0]], [[MUL_RESULT]] 325; VF-FOUR-CHECK-NEXT: [[TMP4:%.*]] = sub i32 [[TMP0]], [[MUL_RESULT]] 326; VF-FOUR-CHECK-NEXT: [[TMP5:%.*]] = icmp sgt i32 [[TMP4]], [[TMP0]] 327; VF-FOUR-CHECK-NEXT: [[TMP6:%.*]] = icmp slt i32 [[TMP3]], [[TMP0]] 328; VF-FOUR-CHECK-NEXT: [[TMP7:%.*]] = select i1 true, i1 [[TMP5]], i1 [[TMP6]] 329; VF-FOUR-CHECK-NEXT: [[TMP8:%.*]] = icmp ugt i64 [[TMP1]], 4294967295 330; VF-FOUR-CHECK-NEXT: [[TMP9:%.*]] = or i1 [[TMP7]], [[TMP8]] 331; VF-FOUR-CHECK-NEXT: [[TMP10:%.*]] = or i1 [[TMP9]], [[MUL_OVERFLOW]] 332; VF-FOUR-CHECK-NEXT: [[TMP11:%.*]] = or i1 false, [[TMP10]] 333; VF-FOUR-CHECK-NEXT: br i1 [[TMP11]], label [[VEC_EPILOG_SCALAR_PH]], label [[VECTOR_MEM_CHECK:%.*]] 334; VF-FOUR-CHECK: vector.memcheck: 335; VF-FOUR-CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr float, float* [[A]], i64 [[WIDE_TRIP_COUNT]] 336; VF-FOUR-CHECK-NEXT: [[SCEVGEP2:%.*]] = bitcast float* [[SCEVGEP]] to i8* 337; VF-FOUR-CHECK-NEXT: [[TMP12:%.*]] = sext i32 [[TMP0]] to i64 338; VF-FOUR-CHECK-NEXT: [[TMP13:%.*]] = add i64 [[TMP12]], 1 339; VF-FOUR-CHECK-NEXT: [[TMP14:%.*]] = sub i64 [[TMP13]], [[WIDE_TRIP_COUNT]] 340; VF-FOUR-CHECK-NEXT: [[SCEVGEP3:%.*]] = getelementptr float, float* [[B:%.*]], i64 [[TMP14]] 341; VF-FOUR-CHECK-NEXT: [[SCEVGEP34:%.*]] = bitcast float* [[SCEVGEP3]] to i8* 342; VF-FOUR-CHECK-NEXT: [[TMP15:%.*]] = add nsw i64 [[TMP12]], 1 343; VF-FOUR-CHECK-NEXT: [[SCEVGEP5:%.*]] = getelementptr float, float* [[B]], i64 [[TMP15]] 344; VF-FOUR-CHECK-NEXT: [[SCEVGEP56:%.*]] = bitcast float* [[SCEVGEP5]] to i8* 345; VF-FOUR-CHECK-NEXT: [[BOUND0:%.*]] = icmp ult i8* [[A1]], [[SCEVGEP56]] 346; VF-FOUR-CHECK-NEXT: [[BOUND1:%.*]] = icmp ult i8* [[SCEVGEP34]], [[SCEVGEP2]] 347; VF-FOUR-CHECK-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]] 348; VF-FOUR-CHECK-NEXT: [[MEMCHECK_CONFLICT:%.*]] = and i1 [[FOUND_CONFLICT]], true 349; VF-FOUR-CHECK-NEXT: br i1 [[MEMCHECK_CONFLICT]], label [[VEC_EPILOG_SCALAR_PH]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]] 350; VF-FOUR-CHECK: vector.main.loop.iter.check: 351; VF-FOUR-CHECK-NEXT: [[MIN_ITERS_CHECK7:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 32 352; VF-FOUR-CHECK-NEXT: br i1 [[MIN_ITERS_CHECK7]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]] 353; VF-FOUR-CHECK: vector.ph: 354; VF-FOUR-CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 32 355; VF-FOUR-CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF]] 356; VF-FOUR-CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 357; VF-FOUR-CHECK: vector.body: 358; VF-FOUR-CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 359; VF-FOUR-CHECK-NEXT: [[TMP16:%.*]] = add i64 [[INDEX]], 0 360; VF-FOUR-CHECK-NEXT: [[TMP17:%.*]] = add i64 [[INDEX]], 4 361; VF-FOUR-CHECK-NEXT: [[TMP18:%.*]] = add i64 [[INDEX]], 8 362; VF-FOUR-CHECK-NEXT: [[TMP19:%.*]] = add i64 [[INDEX]], 12 363; VF-FOUR-CHECK-NEXT: [[TMP20:%.*]] = add i64 [[INDEX]], 16 364; VF-FOUR-CHECK-NEXT: [[TMP21:%.*]] = add i64 [[INDEX]], 20 365; VF-FOUR-CHECK-NEXT: [[TMP22:%.*]] = add i64 [[INDEX]], 24 366; VF-FOUR-CHECK-NEXT: [[TMP23:%.*]] = add i64 [[INDEX]], 28 367; VF-FOUR-CHECK-NEXT: [[OFFSET_IDX:%.*]] = trunc i64 [[INDEX]] to i32 368; VF-FOUR-CHECK-NEXT: [[TMP24:%.*]] = add i32 [[OFFSET_IDX]], 0 369; VF-FOUR-CHECK-NEXT: [[TMP25:%.*]] = add i32 [[OFFSET_IDX]], 4 370; VF-FOUR-CHECK-NEXT: [[TMP26:%.*]] = add i32 [[OFFSET_IDX]], 8 371; VF-FOUR-CHECK-NEXT: [[TMP27:%.*]] = add i32 [[OFFSET_IDX]], 12 372; VF-FOUR-CHECK-NEXT: [[TMP28:%.*]] = add i32 [[OFFSET_IDX]], 16 373; VF-FOUR-CHECK-NEXT: [[TMP29:%.*]] = add i32 [[OFFSET_IDX]], 20 374; VF-FOUR-CHECK-NEXT: [[TMP30:%.*]] = add i32 [[OFFSET_IDX]], 24 375; VF-FOUR-CHECK-NEXT: [[TMP31:%.*]] = add i32 [[OFFSET_IDX]], 28 376; VF-FOUR-CHECK-NEXT: [[TMP32:%.*]] = xor i32 [[TMP24]], -1 377; VF-FOUR-CHECK-NEXT: [[TMP33:%.*]] = xor i32 [[TMP25]], -1 378; VF-FOUR-CHECK-NEXT: [[TMP34:%.*]] = xor i32 [[TMP26]], -1 379; VF-FOUR-CHECK-NEXT: [[TMP35:%.*]] = xor i32 [[TMP27]], -1 380; VF-FOUR-CHECK-NEXT: [[TMP36:%.*]] = xor i32 [[TMP28]], -1 381; VF-FOUR-CHECK-NEXT: [[TMP37:%.*]] = xor i32 [[TMP29]], -1 382; VF-FOUR-CHECK-NEXT: [[TMP38:%.*]] = xor i32 [[TMP30]], -1 383; VF-FOUR-CHECK-NEXT: [[TMP39:%.*]] = xor i32 [[TMP31]], -1 384; VF-FOUR-CHECK-NEXT: [[TMP40:%.*]] = add i32 [[TMP32]], [[N]] 385; VF-FOUR-CHECK-NEXT: [[TMP41:%.*]] = add i32 [[TMP33]], [[N]] 386; VF-FOUR-CHECK-NEXT: [[TMP42:%.*]] = add i32 [[TMP34]], [[N]] 387; VF-FOUR-CHECK-NEXT: [[TMP43:%.*]] = add i32 [[TMP35]], [[N]] 388; VF-FOUR-CHECK-NEXT: [[TMP44:%.*]] = add i32 [[TMP36]], [[N]] 389; VF-FOUR-CHECK-NEXT: [[TMP45:%.*]] = add i32 [[TMP37]], [[N]] 390; VF-FOUR-CHECK-NEXT: [[TMP46:%.*]] = add i32 [[TMP38]], [[N]] 391; VF-FOUR-CHECK-NEXT: [[TMP47:%.*]] = add i32 [[TMP39]], [[N]] 392; VF-FOUR-CHECK-NEXT: [[TMP48:%.*]] = sext i32 [[TMP40]] to i64 393; VF-FOUR-CHECK-NEXT: [[TMP49:%.*]] = sext i32 [[TMP41]] to i64 394; VF-FOUR-CHECK-NEXT: [[TMP50:%.*]] = sext i32 [[TMP42]] to i64 395; VF-FOUR-CHECK-NEXT: [[TMP51:%.*]] = sext i32 [[TMP43]] to i64 396; VF-FOUR-CHECK-NEXT: [[TMP52:%.*]] = sext i32 [[TMP44]] to i64 397; VF-FOUR-CHECK-NEXT: [[TMP53:%.*]] = sext i32 [[TMP45]] to i64 398; VF-FOUR-CHECK-NEXT: [[TMP54:%.*]] = sext i32 [[TMP46]] to i64 399; VF-FOUR-CHECK-NEXT: [[TMP55:%.*]] = sext i32 [[TMP47]] to i64 400; VF-FOUR-CHECK-NEXT: [[TMP56:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP48]] 401; VF-FOUR-CHECK-NEXT: [[TMP57:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP49]] 402; VF-FOUR-CHECK-NEXT: [[TMP58:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP50]] 403; VF-FOUR-CHECK-NEXT: [[TMP59:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP51]] 404; VF-FOUR-CHECK-NEXT: [[TMP60:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP52]] 405; VF-FOUR-CHECK-NEXT: [[TMP61:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP53]] 406; VF-FOUR-CHECK-NEXT: [[TMP62:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP54]] 407; VF-FOUR-CHECK-NEXT: [[TMP63:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP55]] 408; VF-FOUR-CHECK-NEXT: [[TMP64:%.*]] = getelementptr inbounds float, float* [[TMP56]], i32 0 409; VF-FOUR-CHECK-NEXT: [[TMP65:%.*]] = getelementptr inbounds float, float* [[TMP64]], i32 -3 410; VF-FOUR-CHECK-NEXT: [[TMP66:%.*]] = bitcast float* [[TMP65]] to <4 x float>* 411; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, <4 x float>* [[TMP66]], align 4 412; VF-FOUR-CHECK-NEXT: [[REVERSE:%.*]] = shufflevector <4 x float> [[WIDE_LOAD]], <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 413; VF-FOUR-CHECK-NEXT: [[TMP67:%.*]] = getelementptr inbounds float, float* [[TMP56]], i32 -4 414; VF-FOUR-CHECK-NEXT: [[TMP68:%.*]] = getelementptr inbounds float, float* [[TMP67]], i32 -3 415; VF-FOUR-CHECK-NEXT: [[TMP69:%.*]] = bitcast float* [[TMP68]] to <4 x float>* 416; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD8:%.*]] = load <4 x float>, <4 x float>* [[TMP69]], align 4 417; VF-FOUR-CHECK-NEXT: [[REVERSE9:%.*]] = shufflevector <4 x float> [[WIDE_LOAD8]], <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 418; VF-FOUR-CHECK-NEXT: [[TMP70:%.*]] = getelementptr inbounds float, float* [[TMP56]], i32 -8 419; VF-FOUR-CHECK-NEXT: [[TMP71:%.*]] = getelementptr inbounds float, float* [[TMP70]], i32 -3 420; VF-FOUR-CHECK-NEXT: [[TMP72:%.*]] = bitcast float* [[TMP71]] to <4 x float>* 421; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD10:%.*]] = load <4 x float>, <4 x float>* [[TMP72]], align 4 422; VF-FOUR-CHECK-NEXT: [[REVERSE11:%.*]] = shufflevector <4 x float> [[WIDE_LOAD10]], <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 423; VF-FOUR-CHECK-NEXT: [[TMP73:%.*]] = getelementptr inbounds float, float* [[TMP56]], i32 -12 424; VF-FOUR-CHECK-NEXT: [[TMP74:%.*]] = getelementptr inbounds float, float* [[TMP73]], i32 -3 425; VF-FOUR-CHECK-NEXT: [[TMP75:%.*]] = bitcast float* [[TMP74]] to <4 x float>* 426; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD12:%.*]] = load <4 x float>, <4 x float>* [[TMP75]], align 4 427; VF-FOUR-CHECK-NEXT: [[REVERSE13:%.*]] = shufflevector <4 x float> [[WIDE_LOAD12]], <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 428; VF-FOUR-CHECK-NEXT: [[TMP76:%.*]] = getelementptr inbounds float, float* [[TMP56]], i32 -16 429; VF-FOUR-CHECK-NEXT: [[TMP77:%.*]] = getelementptr inbounds float, float* [[TMP76]], i32 -3 430; VF-FOUR-CHECK-NEXT: [[TMP78:%.*]] = bitcast float* [[TMP77]] to <4 x float>* 431; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD14:%.*]] = load <4 x float>, <4 x float>* [[TMP78]], align 4 432; VF-FOUR-CHECK-NEXT: [[REVERSE15:%.*]] = shufflevector <4 x float> [[WIDE_LOAD14]], <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 433; VF-FOUR-CHECK-NEXT: [[TMP79:%.*]] = getelementptr inbounds float, float* [[TMP56]], i32 -20 434; VF-FOUR-CHECK-NEXT: [[TMP80:%.*]] = getelementptr inbounds float, float* [[TMP79]], i32 -3 435; VF-FOUR-CHECK-NEXT: [[TMP81:%.*]] = bitcast float* [[TMP80]] to <4 x float>* 436; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD16:%.*]] = load <4 x float>, <4 x float>* [[TMP81]], align 4 437; VF-FOUR-CHECK-NEXT: [[REVERSE17:%.*]] = shufflevector <4 x float> [[WIDE_LOAD16]], <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 438; VF-FOUR-CHECK-NEXT: [[TMP82:%.*]] = getelementptr inbounds float, float* [[TMP56]], i32 -24 439; VF-FOUR-CHECK-NEXT: [[TMP83:%.*]] = getelementptr inbounds float, float* [[TMP82]], i32 -3 440; VF-FOUR-CHECK-NEXT: [[TMP84:%.*]] = bitcast float* [[TMP83]] to <4 x float>* 441; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD18:%.*]] = load <4 x float>, <4 x float>* [[TMP84]], align 4 442; VF-FOUR-CHECK-NEXT: [[REVERSE19:%.*]] = shufflevector <4 x float> [[WIDE_LOAD18]], <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 443; VF-FOUR-CHECK-NEXT: [[TMP85:%.*]] = getelementptr inbounds float, float* [[TMP56]], i32 -28 444; VF-FOUR-CHECK-NEXT: [[TMP86:%.*]] = getelementptr inbounds float, float* [[TMP85]], i32 -3 445; VF-FOUR-CHECK-NEXT: [[TMP87:%.*]] = bitcast float* [[TMP86]] to <4 x float>* 446; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD20:%.*]] = load <4 x float>, <4 x float>* [[TMP87]], align 4 447; VF-FOUR-CHECK-NEXT: [[REVERSE21:%.*]] = shufflevector <4 x float> [[WIDE_LOAD20]], <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 448; VF-FOUR-CHECK-NEXT: [[TMP88:%.*]] = fadd fast <4 x float> [[REVERSE]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 449; VF-FOUR-CHECK-NEXT: [[TMP89:%.*]] = fadd fast <4 x float> [[REVERSE9]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 450; VF-FOUR-CHECK-NEXT: [[TMP90:%.*]] = fadd fast <4 x float> [[REVERSE11]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 451; VF-FOUR-CHECK-NEXT: [[TMP91:%.*]] = fadd fast <4 x float> [[REVERSE13]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 452; VF-FOUR-CHECK-NEXT: [[TMP92:%.*]] = fadd fast <4 x float> [[REVERSE15]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 453; VF-FOUR-CHECK-NEXT: [[TMP93:%.*]] = fadd fast <4 x float> [[REVERSE17]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 454; VF-FOUR-CHECK-NEXT: [[TMP94:%.*]] = fadd fast <4 x float> [[REVERSE19]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 455; VF-FOUR-CHECK-NEXT: [[TMP95:%.*]] = fadd fast <4 x float> [[REVERSE21]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 456; VF-FOUR-CHECK-NEXT: [[TMP96:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP16]] 457; VF-FOUR-CHECK-NEXT: [[TMP97:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP17]] 458; VF-FOUR-CHECK-NEXT: [[TMP98:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP18]] 459; VF-FOUR-CHECK-NEXT: [[TMP99:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP19]] 460; VF-FOUR-CHECK-NEXT: [[TMP100:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP20]] 461; VF-FOUR-CHECK-NEXT: [[TMP101:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP21]] 462; VF-FOUR-CHECK-NEXT: [[TMP102:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP22]] 463; VF-FOUR-CHECK-NEXT: [[TMP103:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP23]] 464; VF-FOUR-CHECK-NEXT: [[TMP104:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 0 465; VF-FOUR-CHECK-NEXT: [[TMP105:%.*]] = bitcast float* [[TMP104]] to <4 x float>* 466; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP88]], <4 x float>* [[TMP105]], align 4 467; VF-FOUR-CHECK-NEXT: [[TMP106:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 4 468; VF-FOUR-CHECK-NEXT: [[TMP107:%.*]] = bitcast float* [[TMP106]] to <4 x float>* 469; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP89]], <4 x float>* [[TMP107]], align 4 470; VF-FOUR-CHECK-NEXT: [[TMP108:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 8 471; VF-FOUR-CHECK-NEXT: [[TMP109:%.*]] = bitcast float* [[TMP108]] to <4 x float>* 472; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP90]], <4 x float>* [[TMP109]], align 4 473; VF-FOUR-CHECK-NEXT: [[TMP110:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 12 474; VF-FOUR-CHECK-NEXT: [[TMP111:%.*]] = bitcast float* [[TMP110]] to <4 x float>* 475; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP91]], <4 x float>* [[TMP111]], align 4 476; VF-FOUR-CHECK-NEXT: [[TMP112:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 16 477; VF-FOUR-CHECK-NEXT: [[TMP113:%.*]] = bitcast float* [[TMP112]] to <4 x float>* 478; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP92]], <4 x float>* [[TMP113]], align 4 479; VF-FOUR-CHECK-NEXT: [[TMP114:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 20 480; VF-FOUR-CHECK-NEXT: [[TMP115:%.*]] = bitcast float* [[TMP114]] to <4 x float>* 481; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP93]], <4 x float>* [[TMP115]], align 4 482; VF-FOUR-CHECK-NEXT: [[TMP116:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 24 483; VF-FOUR-CHECK-NEXT: [[TMP117:%.*]] = bitcast float* [[TMP116]] to <4 x float>* 484; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP94]], <4 x float>* [[TMP117]], align 4 485; VF-FOUR-CHECK-NEXT: [[TMP118:%.*]] = getelementptr inbounds float, float* [[TMP96]], i32 28 486; VF-FOUR-CHECK-NEXT: [[TMP119:%.*]] = bitcast float* [[TMP118]] to <4 x float>* 487; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP95]], <4 x float>* [[TMP119]], align 4 488; VF-FOUR-CHECK-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], 32 489; VF-FOUR-CHECK-NEXT: [[TMP120:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] 490; VF-FOUR-CHECK-NEXT: br i1 [[TMP120]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOPID_MV_CM:!.*]] 491; VF-FOUR-CHECK: middle.block: 492; VF-FOUR-CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC]] 493; VF-FOUR-CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]] 494; VF-FOUR-CHECK: vec.epilog.iter.check: 495; VF-FOUR-CHECK-NEXT: [[IND_END27:%.*]] = trunc i64 [[N_VEC]] to i32 496; VF-FOUR-CHECK-NEXT: [[N_VEC_REMAINING:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_VEC]] 497; VF-FOUR-CHECK-NEXT: [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 4 498; VF-FOUR-CHECK-NEXT: br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]] 499; VF-FOUR-CHECK: vec.epilog.ph: 500; VF-FOUR-CHECK-NEXT: [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ] 501; VF-FOUR-CHECK-NEXT: [[N_MOD_VF22:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 4 502; VF-FOUR-CHECK-NEXT: [[N_VEC23:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF22]] 503; VF-FOUR-CHECK-NEXT: [[IND_END:%.*]] = trunc i64 [[N_VEC23]] to i32 504; VF-FOUR-CHECK-NEXT: br label [[VEC_EPILOG_VECTOR_BODY:%.*]] 505; VF-FOUR-CHECK: vec.epilog.vector.body: 506; VF-FOUR-CHECK-NEXT: [[INDEX24:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT25:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ] 507; VF-FOUR-CHECK-NEXT: [[TMP121:%.*]] = add i64 [[INDEX24]], 0 508; VF-FOUR-CHECK-NEXT: [[OFFSET_IDX29:%.*]] = trunc i64 [[INDEX24]] to i32 509; VF-FOUR-CHECK-NEXT: [[TMP122:%.*]] = add i32 [[OFFSET_IDX29]], 0 510; VF-FOUR-CHECK-NEXT: [[TMP123:%.*]] = xor i32 [[TMP122]], -1 511; VF-FOUR-CHECK-NEXT: [[TMP124:%.*]] = add i32 [[TMP123]], [[N]] 512; VF-FOUR-CHECK-NEXT: [[TMP125:%.*]] = sext i32 [[TMP124]] to i64 513; VF-FOUR-CHECK-NEXT: [[TMP126:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[TMP125]] 514; VF-FOUR-CHECK-NEXT: [[TMP127:%.*]] = getelementptr inbounds float, float* [[TMP126]], i32 0 515; VF-FOUR-CHECK-NEXT: [[TMP128:%.*]] = getelementptr inbounds float, float* [[TMP127]], i32 -3 516; VF-FOUR-CHECK-NEXT: [[TMP129:%.*]] = bitcast float* [[TMP128]] to <4 x float>* 517; VF-FOUR-CHECK-NEXT: [[WIDE_LOAD30:%.*]] = load <4 x float>, <4 x float>* [[TMP129]], align 4 518; VF-FOUR-CHECK-NEXT: [[REVERSE31:%.*]] = shufflevector <4 x float> [[WIDE_LOAD30]], <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 519; VF-FOUR-CHECK-NEXT: [[TMP130:%.*]] = fadd fast <4 x float> [[REVERSE31]], <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00> 520; VF-FOUR-CHECK-NEXT: [[TMP131:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[TMP121]] 521; VF-FOUR-CHECK-NEXT: [[TMP132:%.*]] = getelementptr inbounds float, float* [[TMP131]], i32 0 522; VF-FOUR-CHECK-NEXT: [[TMP133:%.*]] = bitcast float* [[TMP132]] to <4 x float>* 523; VF-FOUR-CHECK-NEXT: store <4 x float> [[TMP130]], <4 x float>* [[TMP133]], align 4 524; VF-FOUR-CHECK-NEXT: [[INDEX_NEXT25]] = add i64 [[INDEX24]], 4 525; VF-FOUR-CHECK-NEXT: [[TMP134:%.*]] = icmp eq i64 [[INDEX_NEXT25]], [[N_VEC23]] 526; VF-FOUR-CHECK-NEXT: br i1 [[TMP134]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOPID_EV_CM:!.*]] 527; VF-FOUR-CHECK: vec.epilog.middle.block: 528; VF-FOUR-CHECK-NEXT: [[CMP_N28:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC23]] 529; VF-FOUR-CHECK-NEXT: br i1 [[CMP_N28]], label [[FOR_END_LOOPEXIT_LOOPEXIT:%.*]], label [[VEC_EPILOG_SCALAR_PH]] 530; VF-FOUR-CHECK: vec.epilog.scalar.ph: 531; VF-FOUR-CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC23]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[VECTOR_MEM_CHECK]] ], [ 0, [[ITER_CHECK]] ] 532; VF-FOUR-CHECK-NEXT: [[BC_RESUME_VAL26:%.*]] = phi i32 [ [[IND_END]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[IND_END27]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[VECTOR_MEM_CHECK]] ], [ 0, [[ITER_CHECK]] ] 533; VF-FOUR-CHECK-NEXT: br label [[FOR_BODY:%.*]] 534; VF-FOUR-CHECK: for.body: 535; VF-FOUR-CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ] 536; VF-FOUR-CHECK-NEXT: [[I_014:%.*]] = phi i32 [ [[BC_RESUME_VAL26]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ] 537; VF-FOUR-CHECK-NEXT: [[TMP135:%.*]] = xor i32 [[I_014]], -1 538; VF-FOUR-CHECK-NEXT: [[SUB2:%.*]] = add i32 [[TMP135]], [[N]] 539; VF-FOUR-CHECK-NEXT: [[IDXPROM:%.*]] = sext i32 [[SUB2]] to i64 540; VF-FOUR-CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[B]], i64 [[IDXPROM]] 541; VF-FOUR-CHECK-NEXT: [[TMP136:%.*]] = load float, float* [[ARRAYIDX]], align 4 542; VF-FOUR-CHECK-NEXT: [[CONV3:%.*]] = fadd fast float [[TMP136]], 1.000000e+00 543; VF-FOUR-CHECK-NEXT: [[ARRAYIDX5:%.*]] = getelementptr inbounds float, float* [[A]], i64 [[INDVARS_IV]] 544; VF-FOUR-CHECK-NEXT: store float [[CONV3]], float* [[ARRAYIDX5]], align 4 545; VF-FOUR-CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1 546; VF-FOUR-CHECK-NEXT: [[INC]] = add nuw nsw i32 [[I_014]], 1 547; VF-FOUR-CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i64 [[INDVARS_IV_NEXT]], [[WIDE_TRIP_COUNT]] 548; VF-FOUR-CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_BODY]], label [[FOR_END_LOOPEXIT_LOOPEXIT]], !llvm.loop [[LOOPID_MS_CM:!.*]] 549; VF-FOUR-CHECK: for.end.loopexit.loopexit: 550; VF-FOUR-CHECK-NEXT: br label [[FOR_END_LOOPEXIT]] 551; VF-FOUR-CHECK: for.end.loopexit: 552; VF-FOUR-CHECK-NEXT: br label [[FOR_END]] 553; VF-FOUR-CHECK: for.end: 554; VF-FOUR-CHECK-NEXT: ret i32 0 555 556; VF-FOUR-CHECK-DAG: [[LOOPID_MV_CM]] = distinct !{[[LOOPID_MV_CM]], [[LOOPID_DISABLE_VECT_CM:!.*]]} 557; VF-FOUR-CHECK-DAG: [[LOOPID_EV_CM]] = distinct !{[[LOOPID_EV_CM]], [[LOOPID_DISABLE_UNROLL_CM:!.*]], [[LOOPID_DISABLE_VECT_CM:!.*]]} 558; VF-FOUR-CHECK-DAG: [[LOOPID_DISABLE_VECT_CM]] = [[DISABLE_VECT_STR_CM:!{!"llvm.loop.isvectorized".*}.*]] 559; VF-FOUR-CHECK-DAG: [[LOOPID_DISABLE_UNROLL_CM]] = [[DISABLE_UNROLL_STR_CM:!{!"llvm.loop.unroll.runtime.disable"}.*]] 560 561entry: 562 %cmp1 = icmp sgt i32 %n, 1 563 br i1 %cmp1, label %for.body.preheader, label %for.end 564 565for.body.preheader: ; preds = %entry 566 %0 = add i32 %n, -1 567 %wide.trip.count = zext i32 %0 to i64 568 br label %for.body 569 570for.body: ; preds = %for.body.preheader, %for.body 571 %indvars.iv = phi i64 [ 0, %for.body.preheader ], [ %indvars.iv.next, %for.body ] 572 %i.014 = phi i32 [ 0, %for.body.preheader ], [ %inc, %for.body ] 573 %1 = xor i32 %i.014, -1 574 %sub2 = add i32 %1, %n 575 %idxprom = sext i32 %sub2 to i64 576 %arrayidx = getelementptr inbounds float, float* %B, i64 %idxprom 577 %2 = load float, float* %arrayidx, align 4 578 %conv3 = fadd fast float %2, 1.000000e+00 579 %arrayidx5 = getelementptr inbounds float, float* %A, i64 %indvars.iv 580 store float %conv3, float* %arrayidx5, align 4 581 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 582 %inc = add nuw nsw i32 %i.014, 1 583 %exitcond = icmp ne i64 %indvars.iv.next, %wide.trip.count 584 br i1 %exitcond, label %for.body, label %for.end.loopexit 585 586for.end.loopexit: ; preds = %for.body 587 br label %for.end 588 589for.end: ; preds = %for.end.loopexit, %entry 590 ret i32 0 591} 592 593attributes #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" } 594