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: [[TMP0:%.*]] = add i32 [[INDEX]], 0 21; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <64 x i32> poison, i32 [[INDEX]], i32 0 22; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <64 x i32> [[BROADCAST_SPLATINSERT]], <64 x i32> poison, <64 x i32> zeroinitializer 23; 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> 24; CHECK-NEXT: [[TMP1:%.*]] = 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> 25; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds [32 x i8], [32 x i8]* @tab, i32 0, i32 [[TMP0]] 26; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i8, i8* [[TMP2]], 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> [[TMP1]], <64 x i8> poison) 29; CHECK-NEXT: [[TMP5:%.*]] = icmp eq <64 x i8> [[WIDE_MASKED_LOAD]], zeroinitializer 30; CHECK-NEXT: [[TMP6:%.*]] = 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> 31; CHECK-NEXT: [[TMP7:%.*]] = bitcast i8* [[TMP3]] to <64 x i8>* 32; CHECK-NEXT: call void @llvm.masked.store.v64i8.p0v64i8(<64 x i8> [[TMP6]], <64 x i8>* [[TMP7]], i32 1, <64 x i1> [[TMP1]]) 33; CHECK-NEXT: [[INDEX_NEXT]] = add i32 [[INDEX]], 64 34; CHECK-NEXT: [[TMP8:%.*]] = icmp eq i32 [[INDEX_NEXT]], 256 35; CHECK-NEXT: br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] 36; CHECK: middle.block: 37; CHECK-NEXT: br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]] 38; CHECK: scalar.ph: 39; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ 256, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 40; CHECK-NEXT: br label [[FOR_BODY:%.*]] 41; CHECK: for.body: 42; CHECK-NEXT: [[I_08:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ] 43; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [32 x i8], [32 x i8]* @tab, i32 0, i32 [[I_08]] 44; CHECK-NEXT: [[TMP9:%.*]] = load i8, i8* [[ARRAYIDX]], align 1 45; CHECK-NEXT: [[CMP1:%.*]] = icmp eq i8 [[TMP9]], 0 46; CHECK-NEXT: [[DOT:%.*]] = select i1 [[CMP1]], i8 2, i8 1 47; CHECK-NEXT: store i8 [[DOT]], i8* [[ARRAYIDX]], align 1 48; CHECK-NEXT: [[INC]] = add nsw i32 [[I_08]], 1 49; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[I_08]], 202 50; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], [[LOOP2:!llvm.loop !.*]] 51; CHECK: for.end: 52; CHECK-NEXT: ret i32 0 53; 54; AUTOVF-LABEL: @foo_optsize( 55; AUTOVF-NEXT: entry: 56; AUTOVF-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 57; AUTOVF: vector.ph: 58; AUTOVF-NEXT: br label [[VECTOR_BODY:%.*]] 59; AUTOVF: vector.body: 60; AUTOVF-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 61; AUTOVF-NEXT: [[TMP0:%.*]] = add i32 [[INDEX]], 0 62; AUTOVF-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <32 x i32> poison, i32 [[INDEX]], i32 0 63; AUTOVF-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <32 x i32> [[BROADCAST_SPLATINSERT]], <32 x i32> poison, <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: [[TMP1:%.*]] = 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> 66; AUTOVF-NEXT: [[TMP2:%.*]] = getelementptr inbounds [32 x i8], [32 x i8]* @tab, i32 0, i32 [[TMP0]] 67; AUTOVF-NEXT: [[TMP3:%.*]] = getelementptr inbounds i8, i8* [[TMP2]], i32 0 68; AUTOVF-NEXT: [[TMP4:%.*]] = bitcast i8* [[TMP3]] to <32 x i8>* 69; AUTOVF-NEXT: [[WIDE_MASKED_LOAD:%.*]] = call <32 x i8> @llvm.masked.load.v32i8.p0v32i8(<32 x i8>* [[TMP4]], i32 1, <32 x i1> [[TMP1]], <32 x i8> poison) 70; AUTOVF-NEXT: [[TMP5:%.*]] = icmp eq <32 x i8> [[WIDE_MASKED_LOAD]], zeroinitializer 71; AUTOVF-NEXT: [[TMP6:%.*]] = 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> 72; AUTOVF-NEXT: [[TMP7:%.*]] = bitcast i8* [[TMP3]] to <32 x i8>* 73; AUTOVF-NEXT: call void @llvm.masked.store.v32i8.p0v32i8(<32 x i8> [[TMP6]], <32 x i8>* [[TMP7]], i32 1, <32 x i1> [[TMP1]]) 74; AUTOVF-NEXT: [[INDEX_NEXT]] = add i32 [[INDEX]], 32 75; AUTOVF-NEXT: [[TMP8:%.*]] = icmp eq i32 [[INDEX_NEXT]], 224 76; AUTOVF-NEXT: br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] 77; AUTOVF: middle.block: 78; AUTOVF-NEXT: br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]] 79; AUTOVF: scalar.ph: 80; AUTOVF-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ 224, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 81; AUTOVF-NEXT: br label [[FOR_BODY:%.*]] 82; AUTOVF: for.body: 83; AUTOVF-NEXT: [[I_08:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ] 84; AUTOVF-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [32 x i8], [32 x i8]* @tab, i32 0, i32 [[I_08]] 85; AUTOVF-NEXT: [[TMP9:%.*]] = load i8, i8* [[ARRAYIDX]], align 1 86; AUTOVF-NEXT: [[CMP1:%.*]] = icmp eq i8 [[TMP9]], 0 87; AUTOVF-NEXT: [[DOT:%.*]] = select i1 [[CMP1]], i8 2, i8 1 88; AUTOVF-NEXT: store i8 [[DOT]], i8* [[ARRAYIDX]], align 1 89; AUTOVF-NEXT: [[INC]] = add nsw i32 [[I_08]], 1 90; AUTOVF-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[I_08]], 202 91; AUTOVF-NEXT: br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], [[LOOP2:!llvm.loop !.*]] 92; AUTOVF: for.end: 93; AUTOVF-NEXT: ret i32 0 94; 95 96entry: 97 br label %for.body 98 99for.body: ; preds = %for.body, %entry 100 %i.08 = phi i32 [ 0, %entry ], [ %inc, %for.body ] 101 %arrayidx = getelementptr inbounds [32 x i8], [32 x i8]* @tab, i32 0, i32 %i.08 102 %0 = load i8, i8* %arrayidx, align 1 103 %cmp1 = icmp eq i8 %0, 0 104 %. = select i1 %cmp1, i8 2, i8 1 105 store i8 %., i8* %arrayidx, align 1 106 %inc = add nsw i32 %i.08, 1 107 %exitcond = icmp eq i32 %i.08, 202 108 br i1 %exitcond, label %for.end, label %for.body 109 110for.end: ; preds = %for.body 111 ret i32 0 112} 113 114attributes #0 = { optsize } 115 116define i32 @foo_minsize() #1 { 117; CHECK-LABEL: @foo_minsize( 118; CHECK-NEXT: entry: 119; CHECK-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 120; CHECK: vector.ph: 121; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 122; CHECK: vector.body: 123; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 124; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[INDEX]], 0 125; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <64 x i32> poison, i32 [[INDEX]], i32 0 126; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <64 x i32> [[BROADCAST_SPLATINSERT]], <64 x i32> poison, <64 x i32> zeroinitializer 127; 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> 128; CHECK-NEXT: [[TMP1:%.*]] = 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> 129; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds [32 x i8], [32 x i8]* @tab, i32 0, i32 [[TMP0]] 130; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i8, i8* [[TMP2]], i32 0 131; CHECK-NEXT: [[TMP4:%.*]] = bitcast i8* [[TMP3]] to <64 x i8>* 132; CHECK-NEXT: [[WIDE_MASKED_LOAD:%.*]] = call <64 x i8> @llvm.masked.load.v64i8.p0v64i8(<64 x i8>* [[TMP4]], i32 1, <64 x i1> [[TMP1]], <64 x i8> poison) 133; CHECK-NEXT: [[TMP5:%.*]] = icmp eq <64 x i8> [[WIDE_MASKED_LOAD]], zeroinitializer 134; CHECK-NEXT: [[TMP6:%.*]] = 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> 135; CHECK-NEXT: [[TMP7:%.*]] = bitcast i8* [[TMP3]] to <64 x i8>* 136; CHECK-NEXT: call void @llvm.masked.store.v64i8.p0v64i8(<64 x i8> [[TMP6]], <64 x i8>* [[TMP7]], i32 1, <64 x i1> [[TMP1]]) 137; CHECK-NEXT: [[INDEX_NEXT]] = add i32 [[INDEX]], 64 138; CHECK-NEXT: [[TMP8:%.*]] = icmp eq i32 [[INDEX_NEXT]], 256 139; CHECK-NEXT: br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], [[LOOP4:!llvm.loop !.*]] 140; CHECK: middle.block: 141; CHECK-NEXT: br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]] 142; CHECK: scalar.ph: 143; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ 256, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 144; CHECK-NEXT: br label [[FOR_BODY:%.*]] 145; CHECK: for.body: 146; CHECK-NEXT: [[I_08:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ] 147; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [32 x i8], [32 x i8]* @tab, i32 0, i32 [[I_08]] 148; CHECK-NEXT: [[TMP9:%.*]] = load i8, i8* [[ARRAYIDX]], align 1 149; CHECK-NEXT: [[CMP1:%.*]] = icmp eq i8 [[TMP9]], 0 150; CHECK-NEXT: [[DOT:%.*]] = select i1 [[CMP1]], i8 2, i8 1 151; CHECK-NEXT: store i8 [[DOT]], i8* [[ARRAYIDX]], align 1 152; CHECK-NEXT: [[INC]] = add nsw i32 [[I_08]], 1 153; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[I_08]], 202 154; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], [[LOOP5:!llvm.loop !.*]] 155; CHECK: for.end: 156; CHECK-NEXT: ret i32 0 157; 158; AUTOVF-LABEL: @foo_minsize( 159; AUTOVF-NEXT: entry: 160; AUTOVF-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 161; AUTOVF: vector.ph: 162; AUTOVF-NEXT: br label [[VECTOR_BODY:%.*]] 163; AUTOVF: vector.body: 164; AUTOVF-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 165; AUTOVF-NEXT: [[TMP0:%.*]] = add i32 [[INDEX]], 0 166; AUTOVF-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <32 x i32> poison, i32 [[INDEX]], i32 0 167; AUTOVF-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <32 x i32> [[BROADCAST_SPLATINSERT]], <32 x i32> poison, <32 x i32> zeroinitializer 168; 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> 169; AUTOVF-NEXT: [[TMP1:%.*]] = 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> 170; AUTOVF-NEXT: [[TMP2:%.*]] = getelementptr inbounds [32 x i8], [32 x i8]* @tab, i32 0, i32 [[TMP0]] 171; AUTOVF-NEXT: [[TMP3:%.*]] = getelementptr inbounds i8, i8* [[TMP2]], i32 0 172; AUTOVF-NEXT: [[TMP4:%.*]] = bitcast i8* [[TMP3]] to <32 x i8>* 173; AUTOVF-NEXT: [[WIDE_MASKED_LOAD:%.*]] = call <32 x i8> @llvm.masked.load.v32i8.p0v32i8(<32 x i8>* [[TMP4]], i32 1, <32 x i1> [[TMP1]], <32 x i8> poison) 174; AUTOVF-NEXT: [[TMP5:%.*]] = icmp eq <32 x i8> [[WIDE_MASKED_LOAD]], zeroinitializer 175; AUTOVF-NEXT: [[TMP6:%.*]] = 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> 176; AUTOVF-NEXT: [[TMP7:%.*]] = bitcast i8* [[TMP3]] to <32 x i8>* 177; AUTOVF-NEXT: call void @llvm.masked.store.v32i8.p0v32i8(<32 x i8> [[TMP6]], <32 x i8>* [[TMP7]], i32 1, <32 x i1> [[TMP1]]) 178; AUTOVF-NEXT: [[INDEX_NEXT]] = add i32 [[INDEX]], 32 179; AUTOVF-NEXT: [[TMP8:%.*]] = icmp eq i32 [[INDEX_NEXT]], 224 180; AUTOVF-NEXT: br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]] 181; AUTOVF: middle.block: 182; AUTOVF-NEXT: br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]] 183; AUTOVF: scalar.ph: 184; AUTOVF-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ 224, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 185; AUTOVF-NEXT: br label [[FOR_BODY:%.*]] 186; AUTOVF: for.body: 187; AUTOVF-NEXT: [[I_08:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ] 188; AUTOVF-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [32 x i8], [32 x i8]* @tab, i32 0, i32 [[I_08]] 189; AUTOVF-NEXT: [[TMP9:%.*]] = load i8, i8* [[ARRAYIDX]], align 1 190; AUTOVF-NEXT: [[CMP1:%.*]] = icmp eq i8 [[TMP9]], 0 191; AUTOVF-NEXT: [[DOT:%.*]] = select i1 [[CMP1]], i8 2, i8 1 192; AUTOVF-NEXT: store i8 [[DOT]], i8* [[ARRAYIDX]], align 1 193; AUTOVF-NEXT: [[INC]] = add nsw i32 [[I_08]], 1 194; AUTOVF-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[I_08]], 202 195; AUTOVF-NEXT: br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], [[LOOP5:!llvm.loop !.*]] 196; AUTOVF: for.end: 197; AUTOVF-NEXT: ret i32 0 198; 199 200entry: 201 br label %for.body 202 203for.body: ; preds = %for.body, %entry 204 %i.08 = phi i32 [ 0, %entry ], [ %inc, %for.body ] 205 %arrayidx = getelementptr inbounds [32 x i8], [32 x i8]* @tab, i32 0, i32 %i.08 206 %0 = load i8, i8* %arrayidx, align 1 207 %cmp1 = icmp eq i8 %0, 0 208 %. = select i1 %cmp1, i8 2, i8 1 209 store i8 %., i8* %arrayidx, align 1 210 %inc = add nsw i32 %i.08, 1 211 %exitcond = icmp eq i32 %i.08, 202 212 br i1 %exitcond, label %for.end, label %for.body 213 214for.end: ; preds = %for.body 215 ret i32 0 216} 217 218attributes #1 = { minsize } 219 220 221; We can vectorize this one by refraining from versioning for stride==1. 222define void @scev4stride1(i32* noalias nocapture %a, i32* noalias nocapture readonly %b, i32 %k) #2 { 223; CHECK-LABEL: @scev4stride1( 224; CHECK-NEXT: for.body.preheader: 225; CHECK-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 226; CHECK: vector.ph: 227; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <64 x i32> poison, i32 [[K:%.*]], i32 0 228; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <64 x i32> [[BROADCAST_SPLATINSERT]], <64 x i32> poison, <64 x i32> zeroinitializer 229; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 230; CHECK: vector.body: 231; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 232; CHECK-NEXT: [[VEC_IND:%.*]] = phi <64 x i32> [ <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>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ] 233; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[INDEX]], 0 234; CHECK-NEXT: [[TMP64:%.*]] = mul nsw <64 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]] 235; CHECK-NEXT: [[TMP65:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], <64 x i32> [[TMP64]] 236; CHECK-NEXT: [[WIDE_MASKED_GATHER:%.*]] = call <64 x i32> @llvm.masked.gather.v64i32.v64p0i32(<64 x i32*> [[TMP65]], i32 4, <64 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <64 x i32> undef) 237; CHECK-NEXT: [[TMP66:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i32 [[TMP0]] 238; CHECK-NEXT: [[TMP67:%.*]] = getelementptr inbounds i32, i32* [[TMP66]], i32 0 239; CHECK-NEXT: [[TMP68:%.*]] = bitcast i32* [[TMP67]] to <64 x i32>* 240; CHECK-NEXT: store <64 x i32> [[WIDE_MASKED_GATHER]], <64 x i32>* [[TMP68]], align 4 241; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 64 242; CHECK-NEXT: [[VEC_IND_NEXT]] = add <64 x i32> [[VEC_IND]], <i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64, i32 64> 243; CHECK-NEXT: [[TMP69:%.*]] = icmp eq i32 [[INDEX_NEXT]], 256 244; CHECK-NEXT: br i1 [[TMP69]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], [[LOOP6:!llvm.loop !.*]] 245; CHECK: middle.block: 246; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 256, 256 247; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]] 248; CHECK: scalar.ph: 249; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ 256, [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ] 250; CHECK-NEXT: br label [[FOR_BODY:%.*]] 251; CHECK: for.body: 252; CHECK-NEXT: [[I_07:%.*]] = phi i32 [ [[INC:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ] 253; CHECK-NEXT: [[MUL:%.*]] = mul nsw i32 [[I_07]], [[K]] 254; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[B]], i32 [[MUL]] 255; CHECK-NEXT: [[TMP70:%.*]] = load i32, i32* [[ARRAYIDX]], align 4 256; CHECK-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 [[I_07]] 257; CHECK-NEXT: store i32 [[TMP70]], i32* [[ARRAYIDX1]], align 4 258; CHECK-NEXT: [[INC]] = add nuw nsw i32 [[I_07]], 1 259; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INC]], 256 260; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], [[LOOP7:!llvm.loop !.*]] 261; CHECK: for.end.loopexit: 262; CHECK-NEXT: ret void 263; 264; AUTOVF-LABEL: @scev4stride1( 265; AUTOVF-NEXT: for.body.preheader: 266; AUTOVF-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 267; AUTOVF: vector.ph: 268; AUTOVF-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <8 x i32> poison, i32 [[K:%.*]], i32 0 269; AUTOVF-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <8 x i32> [[BROADCAST_SPLATINSERT]], <8 x i32> poison, <8 x i32> zeroinitializer 270; AUTOVF-NEXT: br label [[VECTOR_BODY:%.*]] 271; AUTOVF: vector.body: 272; AUTOVF-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 273; AUTOVF-NEXT: [[VEC_IND:%.*]] = phi <8 x i32> [ <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ] 274; AUTOVF-NEXT: [[TMP0:%.*]] = add i32 [[INDEX]], 0 275; AUTOVF-NEXT: [[TMP8:%.*]] = mul nsw <8 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]] 276; AUTOVF-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], <8 x i32> [[TMP8]] 277; AUTOVF-NEXT: [[WIDE_MASKED_GATHER:%.*]] = call <8 x i32> @llvm.masked.gather.v8i32.v8p0i32(<8 x i32*> [[TMP9]], i32 4, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x i32> undef) 278; AUTOVF-NEXT: [[TMP10:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i32 [[TMP0]] 279; AUTOVF-NEXT: [[TMP11:%.*]] = getelementptr inbounds i32, i32* [[TMP10]], i32 0 280; AUTOVF-NEXT: [[TMP12:%.*]] = bitcast i32* [[TMP11]] to <8 x i32>* 281; AUTOVF-NEXT: store <8 x i32> [[WIDE_MASKED_GATHER]], <8 x i32>* [[TMP12]], align 4 282; AUTOVF-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8 283; AUTOVF-NEXT: [[VEC_IND_NEXT]] = add <8 x i32> [[VEC_IND]], <i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8> 284; AUTOVF-NEXT: [[TMP13:%.*]] = icmp eq i32 [[INDEX_NEXT]], 256 285; AUTOVF-NEXT: br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], [[LOOP6:!llvm.loop !.*]] 286; AUTOVF: middle.block: 287; AUTOVF-NEXT: [[CMP_N:%.*]] = icmp eq i32 256, 256 288; AUTOVF-NEXT: br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]] 289; AUTOVF: scalar.ph: 290; AUTOVF-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ 256, [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ] 291; AUTOVF-NEXT: br label [[FOR_BODY:%.*]] 292; AUTOVF: for.body: 293; AUTOVF-NEXT: [[I_07:%.*]] = phi i32 [ [[INC:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ] 294; AUTOVF-NEXT: [[MUL:%.*]] = mul nsw i32 [[I_07]], [[K]] 295; AUTOVF-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[B]], i32 [[MUL]] 296; AUTOVF-NEXT: [[TMP14:%.*]] = load i32, i32* [[ARRAYIDX]], align 4 297; AUTOVF-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i32, i32* [[A]], i32 [[I_07]] 298; AUTOVF-NEXT: store i32 [[TMP14]], i32* [[ARRAYIDX1]], align 4 299; AUTOVF-NEXT: [[INC]] = add nuw nsw i32 [[I_07]], 1 300; AUTOVF-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INC]], 256 301; AUTOVF-NEXT: br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], [[LOOP7:!llvm.loop !.*]] 302; AUTOVF: for.end.loopexit: 303; AUTOVF-NEXT: ret void 304; 305for.body.preheader: 306 br label %for.body 307 308for.body: ; preds = %for.body.preheader, %for.body 309 %i.07 = phi i32 [ %inc, %for.body ], [ 0, %for.body.preheader ] 310 %mul = mul nsw i32 %i.07, %k 311 %arrayidx = getelementptr inbounds i32, i32* %b, i32 %mul 312 %0 = load i32, i32* %arrayidx, align 4 313 %arrayidx1 = getelementptr inbounds i32, i32* %a, i32 %i.07 314 store i32 %0, i32* %arrayidx1, align 4 315 %inc = add nuw nsw i32 %i.07, 1 316 %exitcond = icmp eq i32 %inc, 256 317 br i1 %exitcond, label %for.end.loopexit, label %for.body 318 319for.end.loopexit: ; preds = %for.body 320 ret void 321} 322 323attributes #2 = { optsize } 324 325 326; PR39497 327; We can't vectorize this one because we version for overflow check and tiny 328; trip count leads to opt-for-size (which otherwise could fold the tail by 329; masking). 330define i32 @main() local_unnamed_addr { 331; CHECK-LABEL: @main( 332; CHECK-NEXT: while.cond: 333; CHECK-NEXT: br label [[FOR_COND:%.*]] 334; CHECK: for.cond: 335; CHECK-NEXT: [[D_0:%.*]] = phi i32 [ 0, [[WHILE_COND:%.*]] ], [ [[ADD:%.*]], [[FOR_COND]] ] 336; CHECK-NEXT: [[CONV:%.*]] = and i32 [[D_0]], 65535 337; CHECK-NEXT: [[CMP:%.*]] = icmp ult i32 [[CONV]], 4 338; CHECK-NEXT: [[ADD]] = add nuw nsw i32 [[CONV]], 1 339; CHECK-NEXT: br i1 [[CMP]], label [[FOR_COND]], label [[WHILE_COND_LOOPEXIT:%.*]] 340; CHECK: while.cond.loopexit: 341; CHECK-NEXT: ret i32 0 342; 343; AUTOVF-LABEL: @main( 344; AUTOVF-NEXT: while.cond: 345; AUTOVF-NEXT: br label [[FOR_COND:%.*]] 346; AUTOVF: for.cond: 347; AUTOVF-NEXT: [[D_0:%.*]] = phi i32 [ 0, [[WHILE_COND:%.*]] ], [ [[ADD:%.*]], [[FOR_COND]] ] 348; AUTOVF-NEXT: [[CONV:%.*]] = and i32 [[D_0]], 65535 349; AUTOVF-NEXT: [[CMP:%.*]] = icmp ult i32 [[CONV]], 4 350; AUTOVF-NEXT: [[ADD]] = add nuw nsw i32 [[CONV]], 1 351; AUTOVF-NEXT: br i1 [[CMP]], label [[FOR_COND]], label [[WHILE_COND_LOOPEXIT:%.*]] 352; AUTOVF: while.cond.loopexit: 353; AUTOVF-NEXT: ret i32 0 354; 355while.cond: 356 br label %for.cond 357 358for.cond: 359 %d.0 = phi i32 [ 0, %while.cond ], [ %add, %for.cond ] 360 %conv = and i32 %d.0, 65535 361 %cmp = icmp ult i32 %conv, 4 362 %add = add nuw nsw i32 %conv, 1 363 br i1 %cmp, label %for.cond, label %while.cond.loopexit 364 365while.cond.loopexit: 366 ret i32 0 367} 368