1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py 2; This test verifies that the loop vectorizer will NOT vectorize loops that 3; will produce a tail loop with the optimize for size or the minimize size 4; attributes. This is a target-dependent version of the test. 5; RUN: opt < %s -loop-vectorize -force-vector-width=64 -S -mtriple=x86_64-unknown-linux -mcpu=skx | FileCheck %s 6; RUN: opt < %s -loop-vectorize -S -mtriple=x86_64-unknown-linux -mcpu=skx | FileCheck %s --check-prefix AUTOVF 7 8target datalayout = "E-m:e-p:32:32-i64:32-f64:32:64-a:0:32-n32-S128" 9 10@tab = common global [32 x i8] zeroinitializer, align 1 11 12define i32 @foo_optsize() #0 { 13; CHECK-LABEL: @foo_optsize( 14; CHECK-NEXT: entry: 15; CHECK-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 16; CHECK: vector.ph: 17; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 18; CHECK: vector.body: 19; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 20; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <64 x i32> undef, i32 [[INDEX]], i32 0 21; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <64 x i32> [[BROADCAST_SPLATINSERT]], <64 x i32> undef, <64 x i32> zeroinitializer 22; CHECK-NEXT: [[INDUCTION:%.*]] = add <64 x i32> [[BROADCAST_SPLAT]], <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63> 23; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[INDEX]], 0 24; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds [32 x i8], [32 x i8]* @tab, i32 0, i32 [[TMP0]] 25; CHECK-NEXT: [[TMP2:%.*]] = icmp ule <64 x i32> [[INDUCTION]], <i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202> 26; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i8, i8* [[TMP1]], i32 0 27; CHECK-NEXT: [[TMP4:%.*]] = bitcast i8* [[TMP3]] to <64 x i8>* 28; CHECK-NEXT: [[WIDE_MASKED_LOAD:%.*]] = call <64 x i8> @llvm.masked.load.v64i8.p0v64i8(<64 x i8>* [[TMP4]], i32 1, <64 x i1> [[TMP2]], <64 x i8> undef) 29; CHECK-NEXT: [[TMP5:%.*]] = icmp eq <64 x i8> [[WIDE_MASKED_LOAD]], zeroinitializer 30; CHECK-NEXT: [[TMP6:%.*]] = extractelement <64 x i1> [[TMP5]], i32 0 31; CHECK-NEXT: [[TMP7:%.*]] = select <64 x i1> [[TMP5]], <64 x i8> <i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2>, <64 x i8> <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1> 32; CHECK-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP3]] to <64 x i8>* 33; CHECK-NEXT: call void @llvm.masked.store.v64i8.p0v64i8(<64 x i8> [[TMP7]], <64 x i8>* [[TMP8]], i32 1, <64 x i1> [[TMP2]]) 34; CHECK-NEXT: [[INDEX_NEXT]] = add i32 [[INDEX]], 64 35; CHECK-NEXT: [[TMP9:%.*]] = icmp eq i32 [[INDEX_NEXT]], 256 36; CHECK-NEXT: br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop !0 37; CHECK: middle.block: 38; CHECK-NEXT: br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]] 39; CHECK: scalar.ph: 40; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ 256, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 41; CHECK-NEXT: br label [[FOR_BODY:%.*]] 42; CHECK: for.body: 43; CHECK-NEXT: [[I_08:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ] 44; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [32 x i8], [32 x i8]* @tab, i32 0, i32 [[I_08]] 45; CHECK-NEXT: [[TMP10:%.*]] = load i8, i8* [[ARRAYIDX]], align 1 46; CHECK-NEXT: [[CMP1:%.*]] = icmp eq i8 [[TMP10]], 0 47; CHECK-NEXT: [[DOT:%.*]] = select i1 [[CMP1]], i8 2, i8 1 48; CHECK-NEXT: store i8 [[DOT]], i8* [[ARRAYIDX]], align 1 49; CHECK-NEXT: [[INC]] = add nsw i32 [[I_08]], 1 50; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[I_08]], 202 51; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop !2 52; CHECK: for.end: 53; CHECK-NEXT: ret i32 0 54; 55; AUTOVF-LABEL: @foo_optsize( 56; AUTOVF-NEXT: entry: 57; AUTOVF-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 58; AUTOVF: vector.ph: 59; AUTOVF-NEXT: br label [[VECTOR_BODY:%.*]] 60; AUTOVF: vector.body: 61; AUTOVF-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 62; AUTOVF-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <32 x i32> undef, i32 [[INDEX]], i32 0 63; AUTOVF-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <32 x i32> [[BROADCAST_SPLATINSERT]], <32 x i32> undef, <32 x i32> zeroinitializer 64; AUTOVF-NEXT: [[INDUCTION:%.*]] = add <32 x i32> [[BROADCAST_SPLAT]], <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31> 65; AUTOVF-NEXT: [[TMP0:%.*]] = add i32 [[INDEX]], 0 66; AUTOVF-NEXT: [[TMP1:%.*]] = getelementptr inbounds [32 x i8], [32 x i8]* @tab, i32 0, i32 [[TMP0]] 67; AUTOVF-NEXT: [[TMP2:%.*]] = icmp ule <32 x i32> [[INDUCTION]], <i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202> 68; AUTOVF-NEXT: [[TMP3:%.*]] = getelementptr inbounds i8, i8* [[TMP1]], i32 0 69; AUTOVF-NEXT: [[TMP4:%.*]] = bitcast i8* [[TMP3]] to <32 x i8>* 70; AUTOVF-NEXT: [[WIDE_MASKED_LOAD:%.*]] = call <32 x i8> @llvm.masked.load.v32i8.p0v32i8(<32 x i8>* [[TMP4]], i32 1, <32 x i1> [[TMP2]], <32 x i8> undef) 71; AUTOVF-NEXT: [[TMP5:%.*]] = icmp eq <32 x i8> [[WIDE_MASKED_LOAD]], zeroinitializer 72; AUTOVF-NEXT: [[TMP6:%.*]] = extractelement <32 x i1> [[TMP5]], i32 0 73; AUTOVF-NEXT: [[TMP7:%.*]] = select <32 x i1> [[TMP5]], <32 x i8> <i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2>, <32 x i8> <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1> 74; AUTOVF-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP3]] to <32 x i8>* 75; AUTOVF-NEXT: call void @llvm.masked.store.v32i8.p0v32i8(<32 x i8> [[TMP7]], <32 x i8>* [[TMP8]], i32 1, <32 x i1> [[TMP2]]) 76; AUTOVF-NEXT: [[INDEX_NEXT]] = add i32 [[INDEX]], 32 77; AUTOVF-NEXT: [[TMP9:%.*]] = icmp eq i32 [[INDEX_NEXT]], 224 78; AUTOVF-NEXT: br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop !0 79; AUTOVF: middle.block: 80; AUTOVF-NEXT: br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]] 81; AUTOVF: scalar.ph: 82; AUTOVF-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ 224, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 83; AUTOVF-NEXT: br label [[FOR_BODY:%.*]] 84; AUTOVF: for.body: 85; AUTOVF-NEXT: [[I_08:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ] 86; AUTOVF-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [32 x i8], [32 x i8]* @tab, i32 0, i32 [[I_08]] 87; AUTOVF-NEXT: [[TMP10:%.*]] = load i8, i8* [[ARRAYIDX]], align 1 88; AUTOVF-NEXT: [[CMP1:%.*]] = icmp eq i8 [[TMP10]], 0 89; AUTOVF-NEXT: [[DOT:%.*]] = select i1 [[CMP1]], i8 2, i8 1 90; AUTOVF-NEXT: store i8 [[DOT]], i8* [[ARRAYIDX]], align 1 91; AUTOVF-NEXT: [[INC]] = add nsw i32 [[I_08]], 1 92; AUTOVF-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[I_08]], 202 93; AUTOVF-NEXT: br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop !2 94; AUTOVF: for.end: 95; AUTOVF-NEXT: ret i32 0 96; 97 98entry: 99 br label %for.body 100 101for.body: ; preds = %for.body, %entry 102 %i.08 = phi i32 [ 0, %entry ], [ %inc, %for.body ] 103 %arrayidx = getelementptr inbounds [32 x i8], [32 x i8]* @tab, i32 0, i32 %i.08 104 %0 = load i8, i8* %arrayidx, align 1 105 %cmp1 = icmp eq i8 %0, 0 106 %. = select i1 %cmp1, i8 2, i8 1 107 store i8 %., i8* %arrayidx, align 1 108 %inc = add nsw i32 %i.08, 1 109 %exitcond = icmp eq i32 %i.08, 202 110 br i1 %exitcond, label %for.end, label %for.body 111 112for.end: ; preds = %for.body 113 ret i32 0 114} 115 116attributes #0 = { optsize } 117 118define i32 @foo_minsize() #1 { 119; CHECK-LABEL: @foo_minsize( 120; CHECK-NEXT: entry: 121; CHECK-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 122; CHECK: vector.ph: 123; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 124; CHECK: vector.body: 125; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 126; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <64 x i32> undef, i32 [[INDEX]], i32 0 127; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <64 x i32> [[BROADCAST_SPLATINSERT]], <64 x i32> undef, <64 x i32> zeroinitializer 128; CHECK-NEXT: [[INDUCTION:%.*]] = add <64 x i32> [[BROADCAST_SPLAT]], <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63> 129; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[INDEX]], 0 130; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds [32 x i8], [32 x i8]* @tab, i32 0, i32 [[TMP0]] 131; CHECK-NEXT: [[TMP2:%.*]] = icmp ule <64 x i32> [[INDUCTION]], <i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202> 132; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i8, i8* [[TMP1]], i32 0 133; CHECK-NEXT: [[TMP4:%.*]] = bitcast i8* [[TMP3]] to <64 x i8>* 134; CHECK-NEXT: [[WIDE_MASKED_LOAD:%.*]] = call <64 x i8> @llvm.masked.load.v64i8.p0v64i8(<64 x i8>* [[TMP4]], i32 1, <64 x i1> [[TMP2]], <64 x i8> undef) 135; CHECK-NEXT: [[TMP5:%.*]] = icmp eq <64 x i8> [[WIDE_MASKED_LOAD]], zeroinitializer 136; CHECK-NEXT: [[TMP6:%.*]] = extractelement <64 x i1> [[TMP5]], i32 0 137; CHECK-NEXT: [[TMP7:%.*]] = select <64 x i1> [[TMP5]], <64 x i8> <i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2>, <64 x i8> <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1> 138; CHECK-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP3]] to <64 x i8>* 139; CHECK-NEXT: call void @llvm.masked.store.v64i8.p0v64i8(<64 x i8> [[TMP7]], <64 x i8>* [[TMP8]], i32 1, <64 x i1> [[TMP2]]) 140; CHECK-NEXT: [[INDEX_NEXT]] = add i32 [[INDEX]], 64 141; CHECK-NEXT: [[TMP9:%.*]] = icmp eq i32 [[INDEX_NEXT]], 256 142; CHECK-NEXT: br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop !4 143; CHECK: middle.block: 144; CHECK-NEXT: br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]] 145; CHECK: scalar.ph: 146; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ 256, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 147; CHECK-NEXT: br label [[FOR_BODY:%.*]] 148; CHECK: for.body: 149; CHECK-NEXT: [[I_08:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ] 150; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [32 x i8], [32 x i8]* @tab, i32 0, i32 [[I_08]] 151; CHECK-NEXT: [[TMP10:%.*]] = load i8, i8* [[ARRAYIDX]], align 1 152; CHECK-NEXT: [[CMP1:%.*]] = icmp eq i8 [[TMP10]], 0 153; CHECK-NEXT: [[DOT:%.*]] = select i1 [[CMP1]], i8 2, i8 1 154; CHECK-NEXT: store i8 [[DOT]], i8* [[ARRAYIDX]], align 1 155; CHECK-NEXT: [[INC]] = add nsw i32 [[I_08]], 1 156; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[I_08]], 202 157; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop !5 158; CHECK: for.end: 159; CHECK-NEXT: ret i32 0 160; 161; AUTOVF-LABEL: @foo_minsize( 162; AUTOVF-NEXT: entry: 163; AUTOVF-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 164; AUTOVF: vector.ph: 165; AUTOVF-NEXT: br label [[VECTOR_BODY:%.*]] 166; AUTOVF: vector.body: 167; AUTOVF-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 168; AUTOVF-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <32 x i32> undef, i32 [[INDEX]], i32 0 169; AUTOVF-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <32 x i32> [[BROADCAST_SPLATINSERT]], <32 x i32> undef, <32 x i32> zeroinitializer 170; AUTOVF-NEXT: [[INDUCTION:%.*]] = add <32 x i32> [[BROADCAST_SPLAT]], <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31> 171; AUTOVF-NEXT: [[TMP0:%.*]] = add i32 [[INDEX]], 0 172; AUTOVF-NEXT: [[TMP1:%.*]] = getelementptr inbounds [32 x i8], [32 x i8]* @tab, i32 0, i32 [[TMP0]] 173; AUTOVF-NEXT: [[TMP2:%.*]] = icmp ule <32 x i32> [[INDUCTION]], <i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202, i32 202> 174; AUTOVF-NEXT: [[TMP3:%.*]] = getelementptr inbounds i8, i8* [[TMP1]], i32 0 175; AUTOVF-NEXT: [[TMP4:%.*]] = bitcast i8* [[TMP3]] to <32 x i8>* 176; AUTOVF-NEXT: [[WIDE_MASKED_LOAD:%.*]] = call <32 x i8> @llvm.masked.load.v32i8.p0v32i8(<32 x i8>* [[TMP4]], i32 1, <32 x i1> [[TMP2]], <32 x i8> undef) 177; AUTOVF-NEXT: [[TMP5:%.*]] = icmp eq <32 x i8> [[WIDE_MASKED_LOAD]], zeroinitializer 178; AUTOVF-NEXT: [[TMP6:%.*]] = extractelement <32 x i1> [[TMP5]], i32 0 179; AUTOVF-NEXT: [[TMP7:%.*]] = select <32 x i1> [[TMP5]], <32 x i8> <i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2>, <32 x i8> <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1> 180; AUTOVF-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP3]] to <32 x i8>* 181; AUTOVF-NEXT: call void @llvm.masked.store.v32i8.p0v32i8(<32 x i8> [[TMP7]], <32 x i8>* [[TMP8]], i32 1, <32 x i1> [[TMP2]]) 182; AUTOVF-NEXT: [[INDEX_NEXT]] = add i32 [[INDEX]], 32 183; AUTOVF-NEXT: [[TMP9:%.*]] = icmp eq i32 [[INDEX_NEXT]], 224 184; AUTOVF-NEXT: br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop !4 185; AUTOVF: middle.block: 186; AUTOVF-NEXT: br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]] 187; AUTOVF: scalar.ph: 188; AUTOVF-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ 224, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 189; AUTOVF-NEXT: br label [[FOR_BODY:%.*]] 190; AUTOVF: for.body: 191; AUTOVF-NEXT: [[I_08:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ] 192; AUTOVF-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [32 x i8], [32 x i8]* @tab, i32 0, i32 [[I_08]] 193; AUTOVF-NEXT: [[TMP10:%.*]] = load i8, i8* [[ARRAYIDX]], align 1 194; AUTOVF-NEXT: [[CMP1:%.*]] = icmp eq i8 [[TMP10]], 0 195; AUTOVF-NEXT: [[DOT:%.*]] = select i1 [[CMP1]], i8 2, i8 1 196; AUTOVF-NEXT: store i8 [[DOT]], i8* [[ARRAYIDX]], align 1 197; AUTOVF-NEXT: [[INC]] = add nsw i32 [[I_08]], 1 198; AUTOVF-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[I_08]], 202 199; AUTOVF-NEXT: br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop !5 200; AUTOVF: for.end: 201; AUTOVF-NEXT: ret i32 0 202; 203 204entry: 205 br label %for.body 206 207for.body: ; preds = %for.body, %entry 208 %i.08 = phi i32 [ 0, %entry ], [ %inc, %for.body ] 209 %arrayidx = getelementptr inbounds [32 x i8], [32 x i8]* @tab, i32 0, i32 %i.08 210 %0 = load i8, i8* %arrayidx, align 1 211 %cmp1 = icmp eq i8 %0, 0 212 %. = select i1 %cmp1, i8 2, i8 1 213 store i8 %., i8* %arrayidx, align 1 214 %inc = add nsw i32 %i.08, 1 215 %exitcond = icmp eq i32 %i.08, 202 216 br i1 %exitcond, label %for.end, label %for.body 217 218for.end: ; preds = %for.body 219 ret i32 0 220} 221 222attributes #1 = { minsize } 223 224 225; We can't vectorize this one because we version for stride==1; even having TC 226; a multiple of VF. 227define void @scev4stride1(i32* noalias nocapture %a, i32* noalias nocapture readonly %b, i32 %k) #2 { 228; CHECK-LABEL: @scev4stride1( 229; CHECK-NEXT: for.body.preheader: 230; CHECK-NEXT: br label [[FOR_BODY:%.*]] 231; CHECK: for.body: 232; CHECK-NEXT: [[I_07:%.*]] = phi i32 [ [[INC:%.*]], [[FOR_BODY]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ] 233; CHECK-NEXT: [[MUL:%.*]] = mul nsw i32 [[I_07]], [[K:%.*]] 234; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], i32 [[MUL]] 235; CHECK-NEXT: [[TMP0:%.*]] = load i32, i32* [[ARRAYIDX]], align 4 236; CHECK-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i32 [[I_07]] 237; CHECK-NEXT: store i32 [[TMP0]], i32* [[ARRAYIDX1]], align 4 238; CHECK-NEXT: [[INC]] = add nuw nsw i32 [[I_07]], 1 239; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INC]], 256 240; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT:%.*]], label [[FOR_BODY]] 241; CHECK: for.end.loopexit: 242; CHECK-NEXT: ret void 243; 244; AUTOVF-LABEL: @scev4stride1( 245; AUTOVF-NEXT: for.body.preheader: 246; AUTOVF-NEXT: br label [[FOR_BODY:%.*]] 247; AUTOVF: for.body: 248; AUTOVF-NEXT: [[I_07:%.*]] = phi i32 [ [[INC:%.*]], [[FOR_BODY]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ] 249; AUTOVF-NEXT: [[MUL:%.*]] = mul nsw i32 [[I_07]], [[K:%.*]] 250; AUTOVF-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], i32 [[MUL]] 251; AUTOVF-NEXT: [[TMP0:%.*]] = load i32, i32* [[ARRAYIDX]], align 4 252; AUTOVF-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i32 [[I_07]] 253; AUTOVF-NEXT: store i32 [[TMP0]], i32* [[ARRAYIDX1]], align 4 254; AUTOVF-NEXT: [[INC]] = add nuw nsw i32 [[I_07]], 1 255; AUTOVF-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INC]], 256 256; AUTOVF-NEXT: br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT:%.*]], label [[FOR_BODY]] 257; AUTOVF: for.end.loopexit: 258; AUTOVF-NEXT: ret void 259; 260for.body.preheader: 261 br label %for.body 262 263for.body: ; preds = %for.body.preheader, %for.body 264 %i.07 = phi i32 [ %inc, %for.body ], [ 0, %for.body.preheader ] 265 %mul = mul nsw i32 %i.07, %k 266 %arrayidx = getelementptr inbounds i32, i32* %b, i32 %mul 267 %0 = load i32, i32* %arrayidx, align 4 268 %arrayidx1 = getelementptr inbounds i32, i32* %a, i32 %i.07 269 store i32 %0, i32* %arrayidx1, align 4 270 %inc = add nuw nsw i32 %i.07, 1 271 %exitcond = icmp eq i32 %inc, 256 272 br i1 %exitcond, label %for.end.loopexit, label %for.body 273 274for.end.loopexit: ; preds = %for.body 275 ret void 276} 277 278attributes #2 = { optsize } 279 280 281; PR39497 282; We can't vectorize this one because we version for overflow check and tiny 283; trip count leads to opt-for-size (which otherwise could fold the tail by 284; masking). 285define i32 @main() local_unnamed_addr { 286; CHECK-LABEL: @main( 287; CHECK-NEXT: while.cond: 288; CHECK-NEXT: br label [[FOR_COND:%.*]] 289; CHECK: for.cond: 290; CHECK-NEXT: [[D_0:%.*]] = phi i32 [ 0, [[WHILE_COND:%.*]] ], [ [[ADD:%.*]], [[FOR_COND]] ] 291; CHECK-NEXT: [[CONV:%.*]] = and i32 [[D_0]], 65535 292; CHECK-NEXT: [[CMP:%.*]] = icmp ult i32 [[CONV]], 4 293; CHECK-NEXT: [[ADD]] = add nuw nsw i32 [[CONV]], 1 294; CHECK-NEXT: br i1 [[CMP]], label [[FOR_COND]], label [[WHILE_COND_LOOPEXIT:%.*]] 295; CHECK: while.cond.loopexit: 296; CHECK-NEXT: ret i32 0 297; 298; AUTOVF-LABEL: @main( 299; AUTOVF-NEXT: while.cond: 300; AUTOVF-NEXT: br label [[FOR_COND:%.*]] 301; AUTOVF: for.cond: 302; AUTOVF-NEXT: [[D_0:%.*]] = phi i32 [ 0, [[WHILE_COND:%.*]] ], [ [[ADD:%.*]], [[FOR_COND]] ] 303; AUTOVF-NEXT: [[CONV:%.*]] = and i32 [[D_0]], 65535 304; AUTOVF-NEXT: [[CMP:%.*]] = icmp ult i32 [[CONV]], 4 305; AUTOVF-NEXT: [[ADD]] = add nuw nsw i32 [[CONV]], 1 306; AUTOVF-NEXT: br i1 [[CMP]], label [[FOR_COND]], label [[WHILE_COND_LOOPEXIT:%.*]] 307; AUTOVF: while.cond.loopexit: 308; AUTOVF-NEXT: ret i32 0 309; 310while.cond: 311 br label %for.cond 312 313for.cond: 314 %d.0 = phi i32 [ 0, %while.cond ], [ %add, %for.cond ] 315 %conv = and i32 %d.0, 65535 316 %cmp = icmp ult i32 %conv, 4 317 %add = add nuw nsw i32 %conv, 1 318 br i1 %cmp, label %for.cond, label %while.cond.loopexit 319 320while.cond.loopexit: 321 ret i32 0 322} 323