1; RUN: opt -S -loop-vectorize -force-vector-width=4 -force-vector-interleave=1 < %s | FileCheck %s 2 3target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128" 4 5; PR39417 6; Check that the need for overflow check prevents vectorizing a loop with tiny 7; trip count (which implies opt for size). 8; CHECK-LABEL: @func_34 9; CHECK-NOT: vector.scevcheck 10; CHECK-NOT: vector.body: 11; CHECK-LABEL: bb67: 12define void @func_34() { 13bb1: 14 br label %bb67 15 16bb67: 17 %storemerge2 = phi i32 [ 0, %bb1 ], [ %_tmp2300, %bb67 ] 18 %sext = shl i32 %storemerge2, 16 19 %_tmp2299 = ashr exact i32 %sext, 16 20 %_tmp2300 = add nsw i32 %_tmp2299, 1 21 %_tmp2310 = trunc i32 %_tmp2300 to i16 22 %_tmp2312 = icmp slt i16 %_tmp2310, 3 23 br i1 %_tmp2312, label %bb67, label %bb68 24 25bb68: 26 ret void 27} 28 29; Check that a loop under opt-for-size is vectorized, w/o checking for 30; stride==1. 31; NOTE: Some assertions have been autogenerated by utils/update_test_checks.py 32define void @scev4stride1(i32* noalias nocapture %a, i32* noalias nocapture readonly %b, i32 %k) #0 { 33; CHECK-LABEL: @scev4stride1( 34; CHECK-NEXT: for.body.preheader: 35; CHECK-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 36; CHECK: vector.ph: 37; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> undef, i32 [[K:%.*]], i32 0 38; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> undef, <4 x i32> zeroinitializer 39; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 40; CHECK: vector.body: 41; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 42; CHECK-NEXT: [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ] 43; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[INDEX]], 0 44; CHECK-NEXT: [[TMP1:%.*]] = add i32 [[INDEX]], 1 45; CHECK-NEXT: [[TMP2:%.*]] = add i32 [[INDEX]], 2 46; CHECK-NEXT: [[TMP3:%.*]] = add i32 [[INDEX]], 3 47; CHECK-NEXT: [[TMP4:%.*]] = mul nsw <4 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]] 48; CHECK-NEXT: [[TMP5:%.*]] = extractelement <4 x i32> [[TMP4]], i32 0 49; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], i32 [[TMP5]] 50; CHECK-NEXT: [[TMP7:%.*]] = extractelement <4 x i32> [[TMP4]], i32 1 51; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds i32, i32* [[B]], i32 [[TMP7]] 52; CHECK-NEXT: [[TMP9:%.*]] = extractelement <4 x i32> [[TMP4]], i32 2 53; CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds i32, i32* [[B]], i32 [[TMP9]] 54; CHECK-NEXT: [[TMP11:%.*]] = extractelement <4 x i32> [[TMP4]], i32 3 55; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds i32, i32* [[B]], i32 [[TMP11]] 56; CHECK-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP6]], align 4 57; CHECK-NEXT: [[TMP14:%.*]] = load i32, i32* [[TMP8]], align 4 58; CHECK-NEXT: [[TMP15:%.*]] = load i32, i32* [[TMP10]], align 4 59; CHECK-NEXT: [[TMP16:%.*]] = load i32, i32* [[TMP12]], align 4 60; CHECK-NEXT: [[TMP17:%.*]] = insertelement <4 x i32> undef, i32 [[TMP13]], i32 0 61; CHECK-NEXT: [[TMP18:%.*]] = insertelement <4 x i32> [[TMP17]], i32 [[TMP14]], i32 1 62; CHECK-NEXT: [[TMP19:%.*]] = insertelement <4 x i32> [[TMP18]], i32 [[TMP15]], i32 2 63; CHECK-NEXT: [[TMP20:%.*]] = insertelement <4 x i32> [[TMP19]], i32 [[TMP16]], i32 3 64; CHECK-NEXT: [[TMP21:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i32 [[TMP0]] 65; CHECK-NEXT: [[TMP22:%.*]] = getelementptr inbounds i32, i32* [[TMP21]], i32 0 66; CHECK-NEXT: [[TMP23:%.*]] = bitcast i32* [[TMP22]] to <4 x i32>* 67; CHECK-NEXT: store <4 x i32> [[TMP20]], <4 x i32>* [[TMP23]], align 4 68; CHECK-NEXT: [[INDEX_NEXT]] = add i32 [[INDEX]], 4 69; CHECK-NEXT: [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], <i32 4, i32 4, i32 4, i32 4> 70; CHECK-NEXT: [[TMP24:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1024 71; CHECK-NEXT: br i1 [[TMP24]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop !0 72; CHECK: middle.block: 73; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 1024, 1024 74; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]] 75; CHECK: scalar.ph: 76; CHECK: for.body: 77; CHECK: for.end.loopexit: 78; CHECK-NEXT: ret void 79; 80for.body.preheader: 81 br label %for.body 82 83for.body: 84 %i.07 = phi i32 [ %inc, %for.body ], [ 0, %for.body.preheader ] 85 %mul = mul nsw i32 %i.07, %k 86 %arrayidx = getelementptr inbounds i32, i32* %b, i32 %mul 87 %0 = load i32, i32* %arrayidx, align 4 88 %arrayidx1 = getelementptr inbounds i32, i32* %a, i32 %i.07 89 store i32 %0, i32* %arrayidx1, align 4 90 %inc = add nuw nsw i32 %i.07, 1 91 %exitcond = icmp eq i32 %inc, 1024 92 br i1 %exitcond, label %for.end.loopexit, label %for.body 93 94for.end.loopexit: 95 ret void 96} 97 98attributes #0 = { optsize } 99