1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py 2; RUN: opt -S -mtriple=x86_64-pc_linux -loop-vectorize -instcombine < %s | FileCheck %s --check-prefix=SSE 3; RUN: opt -S -mtriple=x86_64-pc_linux -loop-vectorize -instcombine -mcpu=sandybridge < %s | FileCheck %s --check-prefix=AVX1 4; RUN: opt -S -mtriple=x86_64-pc_linux -loop-vectorize -instcombine -mcpu=haswell < %s | FileCheck %s --check-prefix=AVX2 5; RUN: opt -S -mtriple=x86_64-pc_linux -loop-vectorize -instcombine -mcpu=slm < %s | FileCheck %s --check-prefix=SSE 6; RUN: opt -S -mtriple=x86_64-pc_linux -loop-vectorize -instcombine -mcpu=atom < %s | FileCheck %s --check-prefix=ATOM 7 8define void @foo(i32* noalias nocapture %a, i32* noalias nocapture readonly %b) { 9; SSE-LABEL: @foo( 10; SSE-NEXT: entry: 11; SSE-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 12; SSE: vector.ph: 13; SSE-NEXT: br label [[VECTOR_BODY:%.*]] 14; SSE: vector.body: 15; SSE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 16; SSE-NEXT: [[TMP0:%.*]] = shl nsw i64 [[INDEX]], 1 17; SSE-NEXT: [[TMP1:%.*]] = shl i64 [[INDEX]], 1 18; SSE-NEXT: [[TMP2:%.*]] = or i64 [[TMP1]], 8 19; SSE-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], i64 [[TMP0]] 20; SSE-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[TMP2]] 21; SSE-NEXT: [[TMP5:%.*]] = bitcast i32* [[TMP3]] to <8 x i32>* 22; SSE-NEXT: [[TMP6:%.*]] = bitcast i32* [[TMP4]] to <8 x i32>* 23; SSE-NEXT: [[WIDE_VEC:%.*]] = load <8 x i32>, <8 x i32>* [[TMP5]], align 4 24; SSE-NEXT: [[WIDE_VEC1:%.*]] = load <8 x i32>, <8 x i32>* [[TMP6]], align 4 25; SSE-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <8 x i32> [[WIDE_VEC]], <8 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6> 26; SSE-NEXT: [[STRIDED_VEC2:%.*]] = shufflevector <8 x i32> [[WIDE_VEC1]], <8 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6> 27; SSE-NEXT: [[STRIDED_VEC3:%.*]] = shufflevector <8 x i32> [[WIDE_VEC]], <8 x i32> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7> 28; SSE-NEXT: [[STRIDED_VEC4:%.*]] = shufflevector <8 x i32> [[WIDE_VEC1]], <8 x i32> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7> 29; SSE-NEXT: [[TMP7:%.*]] = add nsw <4 x i32> [[STRIDED_VEC3]], [[STRIDED_VEC]] 30; SSE-NEXT: [[TMP8:%.*]] = add nsw <4 x i32> [[STRIDED_VEC4]], [[STRIDED_VEC2]] 31; SSE-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[INDEX]] 32; SSE-NEXT: [[TMP10:%.*]] = bitcast i32* [[TMP9]] to <4 x i32>* 33; SSE-NEXT: store <4 x i32> [[TMP7]], <4 x i32>* [[TMP10]], align 4 34; SSE-NEXT: [[TMP11:%.*]] = getelementptr inbounds i32, i32* [[TMP9]], i64 4 35; SSE-NEXT: [[TMP12:%.*]] = bitcast i32* [[TMP11]] to <4 x i32>* 36; SSE-NEXT: store <4 x i32> [[TMP8]], <4 x i32>* [[TMP12]], align 4 37; SSE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8 38; SSE-NEXT: [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024 39; SSE-NEXT: br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] 40; SSE: middle.block: 41; SSE-NEXT: br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]] 42; SSE: scalar.ph: 43; SSE-NEXT: br label [[FOR_BODY:%.*]] 44; SSE: for.cond.cleanup: 45; SSE-NEXT: ret void 46; SSE: for.body: 47; SSE-NEXT: br i1 poison, label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP2:![0-9]+]] 48; 49; AVX1-LABEL: @foo( 50; AVX1-NEXT: entry: 51; AVX1-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 52; AVX1: vector.ph: 53; AVX1-NEXT: br label [[VECTOR_BODY:%.*]] 54; AVX1: vector.body: 55; AVX1-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 56; AVX1-NEXT: [[TMP0:%.*]] = shl nsw i64 [[INDEX]], 1 57; AVX1-NEXT: [[TMP1:%.*]] = shl i64 [[INDEX]], 1 58; AVX1-NEXT: [[TMP2:%.*]] = or i64 [[TMP1]], 8 59; AVX1-NEXT: [[TMP3:%.*]] = shl i64 [[INDEX]], 1 60; AVX1-NEXT: [[TMP4:%.*]] = or i64 [[TMP3]], 16 61; AVX1-NEXT: [[TMP5:%.*]] = shl i64 [[INDEX]], 1 62; AVX1-NEXT: [[TMP6:%.*]] = or i64 [[TMP5]], 24 63; AVX1-NEXT: [[TMP7:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], i64 [[TMP0]] 64; AVX1-NEXT: [[TMP8:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[TMP2]] 65; AVX1-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[TMP4]] 66; AVX1-NEXT: [[TMP10:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[TMP6]] 67; AVX1-NEXT: [[TMP11:%.*]] = bitcast i32* [[TMP7]] to <8 x i32>* 68; AVX1-NEXT: [[TMP12:%.*]] = bitcast i32* [[TMP8]] to <8 x i32>* 69; AVX1-NEXT: [[TMP13:%.*]] = bitcast i32* [[TMP9]] to <8 x i32>* 70; AVX1-NEXT: [[TMP14:%.*]] = bitcast i32* [[TMP10]] to <8 x i32>* 71; AVX1-NEXT: [[WIDE_VEC:%.*]] = load <8 x i32>, <8 x i32>* [[TMP11]], align 4 72; AVX1-NEXT: [[WIDE_VEC1:%.*]] = load <8 x i32>, <8 x i32>* [[TMP12]], align 4 73; AVX1-NEXT: [[WIDE_VEC2:%.*]] = load <8 x i32>, <8 x i32>* [[TMP13]], align 4 74; AVX1-NEXT: [[WIDE_VEC3:%.*]] = load <8 x i32>, <8 x i32>* [[TMP14]], align 4 75; AVX1-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <8 x i32> [[WIDE_VEC]], <8 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6> 76; AVX1-NEXT: [[STRIDED_VEC4:%.*]] = shufflevector <8 x i32> [[WIDE_VEC1]], <8 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6> 77; AVX1-NEXT: [[STRIDED_VEC5:%.*]] = shufflevector <8 x i32> [[WIDE_VEC2]], <8 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6> 78; AVX1-NEXT: [[STRIDED_VEC6:%.*]] = shufflevector <8 x i32> [[WIDE_VEC3]], <8 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6> 79; AVX1-NEXT: [[STRIDED_VEC7:%.*]] = shufflevector <8 x i32> [[WIDE_VEC]], <8 x i32> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7> 80; AVX1-NEXT: [[STRIDED_VEC8:%.*]] = shufflevector <8 x i32> [[WIDE_VEC1]], <8 x i32> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7> 81; AVX1-NEXT: [[STRIDED_VEC9:%.*]] = shufflevector <8 x i32> [[WIDE_VEC2]], <8 x i32> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7> 82; AVX1-NEXT: [[STRIDED_VEC10:%.*]] = shufflevector <8 x i32> [[WIDE_VEC3]], <8 x i32> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7> 83; AVX1-NEXT: [[TMP15:%.*]] = add nsw <4 x i32> [[STRIDED_VEC7]], [[STRIDED_VEC]] 84; AVX1-NEXT: [[TMP16:%.*]] = add nsw <4 x i32> [[STRIDED_VEC8]], [[STRIDED_VEC4]] 85; AVX1-NEXT: [[TMP17:%.*]] = add nsw <4 x i32> [[STRIDED_VEC9]], [[STRIDED_VEC5]] 86; AVX1-NEXT: [[TMP18:%.*]] = add nsw <4 x i32> [[STRIDED_VEC10]], [[STRIDED_VEC6]] 87; AVX1-NEXT: [[TMP19:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[INDEX]] 88; AVX1-NEXT: [[TMP20:%.*]] = bitcast i32* [[TMP19]] to <4 x i32>* 89; AVX1-NEXT: store <4 x i32> [[TMP15]], <4 x i32>* [[TMP20]], align 4 90; AVX1-NEXT: [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[TMP19]], i64 4 91; AVX1-NEXT: [[TMP22:%.*]] = bitcast i32* [[TMP21]] to <4 x i32>* 92; AVX1-NEXT: store <4 x i32> [[TMP16]], <4 x i32>* [[TMP22]], align 4 93; AVX1-NEXT: [[TMP23:%.*]] = getelementptr inbounds i32, i32* [[TMP19]], i64 8 94; AVX1-NEXT: [[TMP24:%.*]] = bitcast i32* [[TMP23]] to <4 x i32>* 95; AVX1-NEXT: store <4 x i32> [[TMP17]], <4 x i32>* [[TMP24]], align 4 96; AVX1-NEXT: [[TMP25:%.*]] = getelementptr inbounds i32, i32* [[TMP19]], i64 12 97; AVX1-NEXT: [[TMP26:%.*]] = bitcast i32* [[TMP25]] to <4 x i32>* 98; AVX1-NEXT: store <4 x i32> [[TMP18]], <4 x i32>* [[TMP26]], align 4 99; AVX1-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16 100; AVX1-NEXT: [[TMP27:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024 101; AVX1-NEXT: br i1 [[TMP27]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] 102; AVX1: middle.block: 103; AVX1-NEXT: br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]] 104; AVX1: scalar.ph: 105; AVX1-NEXT: br label [[FOR_BODY:%.*]] 106; AVX1: for.cond.cleanup: 107; AVX1-NEXT: ret void 108; AVX1: for.body: 109; AVX1-NEXT: br i1 poison, label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP2:![0-9]+]] 110; 111; AVX2-LABEL: @foo( 112; AVX2-NEXT: entry: 113; AVX2-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 114; AVX2: vector.ph: 115; AVX2-NEXT: br label [[VECTOR_BODY:%.*]] 116; AVX2: vector.body: 117; AVX2-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 118; AVX2-NEXT: [[TMP0:%.*]] = shl nsw i64 [[INDEX]], 1 119; AVX2-NEXT: [[TMP1:%.*]] = shl i64 [[INDEX]], 1 120; AVX2-NEXT: [[TMP2:%.*]] = or i64 [[TMP1]], 16 121; AVX2-NEXT: [[TMP3:%.*]] = shl i64 [[INDEX]], 1 122; AVX2-NEXT: [[TMP4:%.*]] = or i64 [[TMP3]], 32 123; AVX2-NEXT: [[TMP5:%.*]] = shl i64 [[INDEX]], 1 124; AVX2-NEXT: [[TMP6:%.*]] = or i64 [[TMP5]], 48 125; AVX2-NEXT: [[TMP7:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], i64 [[TMP0]] 126; AVX2-NEXT: [[TMP8:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[TMP2]] 127; AVX2-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[TMP4]] 128; AVX2-NEXT: [[TMP10:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[TMP6]] 129; AVX2-NEXT: [[TMP11:%.*]] = bitcast i32* [[TMP7]] to <16 x i32>* 130; AVX2-NEXT: [[TMP12:%.*]] = bitcast i32* [[TMP8]] to <16 x i32>* 131; AVX2-NEXT: [[TMP13:%.*]] = bitcast i32* [[TMP9]] to <16 x i32>* 132; AVX2-NEXT: [[TMP14:%.*]] = bitcast i32* [[TMP10]] to <16 x i32>* 133; AVX2-NEXT: [[WIDE_VEC:%.*]] = load <16 x i32>, <16 x i32>* [[TMP11]], align 4 134; AVX2-NEXT: [[WIDE_VEC1:%.*]] = load <16 x i32>, <16 x i32>* [[TMP12]], align 4 135; AVX2-NEXT: [[WIDE_VEC2:%.*]] = load <16 x i32>, <16 x i32>* [[TMP13]], align 4 136; AVX2-NEXT: [[WIDE_VEC3:%.*]] = load <16 x i32>, <16 x i32>* [[TMP14]], align 4 137; AVX2-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <16 x i32> [[WIDE_VEC]], <16 x i32> poison, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14> 138; AVX2-NEXT: [[STRIDED_VEC4:%.*]] = shufflevector <16 x i32> [[WIDE_VEC1]], <16 x i32> poison, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14> 139; AVX2-NEXT: [[STRIDED_VEC5:%.*]] = shufflevector <16 x i32> [[WIDE_VEC2]], <16 x i32> poison, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14> 140; AVX2-NEXT: [[STRIDED_VEC6:%.*]] = shufflevector <16 x i32> [[WIDE_VEC3]], <16 x i32> poison, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14> 141; AVX2-NEXT: [[STRIDED_VEC7:%.*]] = shufflevector <16 x i32> [[WIDE_VEC]], <16 x i32> poison, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15> 142; AVX2-NEXT: [[STRIDED_VEC8:%.*]] = shufflevector <16 x i32> [[WIDE_VEC1]], <16 x i32> poison, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15> 143; AVX2-NEXT: [[STRIDED_VEC9:%.*]] = shufflevector <16 x i32> [[WIDE_VEC2]], <16 x i32> poison, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15> 144; AVX2-NEXT: [[STRIDED_VEC10:%.*]] = shufflevector <16 x i32> [[WIDE_VEC3]], <16 x i32> poison, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15> 145; AVX2-NEXT: [[TMP15:%.*]] = add nsw <8 x i32> [[STRIDED_VEC7]], [[STRIDED_VEC]] 146; AVX2-NEXT: [[TMP16:%.*]] = add nsw <8 x i32> [[STRIDED_VEC8]], [[STRIDED_VEC4]] 147; AVX2-NEXT: [[TMP17:%.*]] = add nsw <8 x i32> [[STRIDED_VEC9]], [[STRIDED_VEC5]] 148; AVX2-NEXT: [[TMP18:%.*]] = add nsw <8 x i32> [[STRIDED_VEC10]], [[STRIDED_VEC6]] 149; AVX2-NEXT: [[TMP19:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[INDEX]] 150; AVX2-NEXT: [[TMP20:%.*]] = bitcast i32* [[TMP19]] to <8 x i32>* 151; AVX2-NEXT: store <8 x i32> [[TMP15]], <8 x i32>* [[TMP20]], align 4 152; AVX2-NEXT: [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[TMP19]], i64 8 153; AVX2-NEXT: [[TMP22:%.*]] = bitcast i32* [[TMP21]] to <8 x i32>* 154; AVX2-NEXT: store <8 x i32> [[TMP16]], <8 x i32>* [[TMP22]], align 4 155; AVX2-NEXT: [[TMP23:%.*]] = getelementptr inbounds i32, i32* [[TMP19]], i64 16 156; AVX2-NEXT: [[TMP24:%.*]] = bitcast i32* [[TMP23]] to <8 x i32>* 157; AVX2-NEXT: store <8 x i32> [[TMP17]], <8 x i32>* [[TMP24]], align 4 158; AVX2-NEXT: [[TMP25:%.*]] = getelementptr inbounds i32, i32* [[TMP19]], i64 24 159; AVX2-NEXT: [[TMP26:%.*]] = bitcast i32* [[TMP25]] to <8 x i32>* 160; AVX2-NEXT: store <8 x i32> [[TMP18]], <8 x i32>* [[TMP26]], align 4 161; AVX2-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 32 162; AVX2-NEXT: [[TMP27:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024 163; AVX2-NEXT: br i1 [[TMP27]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] 164; AVX2: middle.block: 165; AVX2-NEXT: br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]] 166; AVX2: scalar.ph: 167; AVX2-NEXT: br label [[FOR_BODY:%.*]] 168; AVX2: for.cond.cleanup: 169; AVX2-NEXT: ret void 170; AVX2: for.body: 171; AVX2-NEXT: br i1 poison, label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP2:![0-9]+]] 172; 173; ATOM-LABEL: @foo( 174; ATOM-NEXT: entry: 175; ATOM-NEXT: br label [[FOR_BODY:%.*]] 176; ATOM: for.cond.cleanup: 177; ATOM-NEXT: ret void 178; ATOM: for.body: 179; ATOM-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ] 180; ATOM-NEXT: [[TMP0:%.*]] = shl nuw nsw i64 [[INDVARS_IV]], 1 181; ATOM-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], i64 [[TMP0]] 182; ATOM-NEXT: [[TMP1:%.*]] = load i32, i32* [[ARRAYIDX]], align 4 183; ATOM-NEXT: [[TMP2:%.*]] = or i64 [[TMP0]], 1 184; ATOM-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[TMP2]] 185; ATOM-NEXT: [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX3]], align 4 186; ATOM-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP3]], [[TMP1]] 187; ATOM-NEXT: [[ARRAYIDX6:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[INDVARS_IV]] 188; ATOM-NEXT: store i32 [[ADD4]], i32* [[ARRAYIDX6]], align 4 189; ATOM-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1 190; ATOM-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 1024 191; ATOM-NEXT: br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP:%.*]], label [[FOR_BODY]] 192; 193entry: 194 br label %for.body 195 196for.cond.cleanup: ; preds = %for.body 197 ret void 198 199for.body: ; preds = %for.body, %entry 200 %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ] 201 %0 = shl nsw i64 %indvars.iv, 1 202 %arrayidx = getelementptr inbounds i32, i32* %b, i64 %0 203 %1 = load i32, i32* %arrayidx, align 4 204 %2 = or i64 %0, 1 205 %arrayidx3 = getelementptr inbounds i32, i32* %b, i64 %2 206 %3 = load i32, i32* %arrayidx3, align 4 207 %add4 = add nsw i32 %3, %1 208 %arrayidx6 = getelementptr inbounds i32, i32* %a, i64 %indvars.iv 209 store i32 %add4, i32* %arrayidx6, align 4 210 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 211 %exitcond = icmp eq i64 %indvars.iv.next, 1024 212 br i1 %exitcond, label %for.cond.cleanup, label %for.body 213} 214