1; RUN: opt -S -loop-reroll %s | FileCheck %s 2target triple = "aarch64--linux-gnu" 3 4define void @test(i32 %n, float* %arrayidx200, float* %arrayidx164, float* %arrayidx172) { 5entry: 6 %rem.i = srem i32 %n, 4 7 %t22 = load float, float* %arrayidx172, align 4 8 %cmp.9 = icmp eq i32 %n, 0 9 %t7 = sext i32 %n to i64 10 br i1 %cmp.9, label %while.end, label %while.body.preheader 11 12while.body.preheader: 13 br label %while.body 14 15while.body: 16;CHECK-LABEL: while.body: 17;CHECK-NEXT: %indvars.iv.i423 = phi i64 [ %indvars.iv.next.i424, %while.body ], [ 0, %while.body.preheader ] 18;CHECK-NEXT: [[T1:%[0-9]+]] = trunc i64 %indvars.iv.i423 to i32 19;CHECK-NEXT: %arrayidx62.i = getelementptr inbounds float, float* %arrayidx200, i64 %indvars.iv.i423 20;CHECK-NEXT: %t1 = load float, float* %arrayidx62.i, align 4 21;CHECK-NEXT: %arrayidx64.i = getelementptr inbounds float, float* %arrayidx164, i64 %indvars.iv.i423 22;CHECK-NEXT: %t2 = load float, float* %arrayidx64.i, align 4 23;CHECK-NEXT: %mul65.i = fmul fast float %t2, %t22 24;CHECK-NEXT: %add66.i = fadd fast float %mul65.i, %t1 25;CHECK-NEXT: store float %add66.i, float* %arrayidx62.i, align 4 26;CHECK-NEXT: %indvars.iv.next.i424 = add i64 %indvars.iv.i423, 1 27;CHECK-NEXT: [[T2:%[0-9]+]] = sext i32 [[T1]] to i64 28;CHECK-NEXT: %exitcond = icmp eq i64 [[T2]], %{{[0-9]+}} 29;CHECK-NEXT: br i1 %exitcond, label %while.end.loopexit, label %while.body 30 31 %indvars.iv.i423 = phi i64 [ %indvars.iv.next.i424, %while.body ], [ 0, %while.body.preheader ] 32 %i.22.i = phi i32 [ %add103.i, %while.body ], [ %rem.i, %while.body.preheader ] 33 %arrayidx62.i = getelementptr inbounds float, float* %arrayidx200, i64 %indvars.iv.i423 34 %t1 = load float, float* %arrayidx62.i, align 4 35 %arrayidx64.i = getelementptr inbounds float, float* %arrayidx164, i64 %indvars.iv.i423 36 %t2 = load float, float* %arrayidx64.i, align 4 37 %mul65.i = fmul fast float %t2, %t22 38 %add66.i = fadd fast float %mul65.i, %t1 39 store float %add66.i, float* %arrayidx62.i, align 4 40 %t3 = add nsw i64 %indvars.iv.i423, 1 41 %arrayidx71.i = getelementptr inbounds float, float* %arrayidx200, i64 %t3 42 %t4 = load float, float* %arrayidx71.i, align 4 43 %arrayidx74.i = getelementptr inbounds float, float* %arrayidx164, i64 %t3 44 %t5 = load float, float* %arrayidx74.i, align 4 45 %mul75.i = fmul fast float %t5, %t22 46 %add76.i = fadd fast float %mul75.i, %t4 47 store float %add76.i, float* %arrayidx71.i, align 4 48 %add103.i = add nsw i32 %i.22.i, 2 49 %t6 = sext i32 %add103.i to i64 50 %cmp58.i = icmp slt i64 %t6, %t7 51 %indvars.iv.next.i424 = add i64 %indvars.iv.i423, 2 52 br i1 %cmp58.i, label %while.body, label %while.end.loopexit 53 54while.end.loopexit: 55 br label %while.end 56 57while.end: 58 ret void 59} 60 61; Function Attrs: noinline norecurse nounwind 62define i32 @test2(i64 %n, i32* nocapture %x, i32* nocapture readonly %y) { 63entry: 64 %cmp18 = icmp sgt i64 %n, 0 65 br i1 %cmp18, label %for.body.preheader, label %for.end 66 67for.body.preheader: ; preds = %entry 68 br label %for.body 69 70for.body: ; preds = %for.body.preheader, %for.body 71 72;CHECK: for.body: 73;CHECK-NEXT: %indvar = phi i64 [ %indvar.next, %for.body ], [ 0, %for.body.preheader ] 74;CHECK-NEXT: %arrayidx = getelementptr inbounds i32, i32* %y, i64 %indvar 75;CHECK-NEXT: [[T1:%[0-9]+]] = load i32, i32* %arrayidx, align 4 76;CHECK-NEXT: %arrayidx3 = getelementptr inbounds i32, i32* %x, i64 %indvar 77;CHECK-NEXT: store i32 [[T1]], i32* %arrayidx3, align 4 78;CHECK-NEXT: %indvar.next = add i64 %indvar, 1 79;CHECK-NEXT: %exitcond = icmp eq i64 %indvar, %{{[0-9]+}} 80;CHECK-NEXT: br i1 %exitcond, label %for.end.loopexit, label %for.body 81 82 %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %for.body.preheader ] 83 %arrayidx = getelementptr inbounds i32, i32* %y, i64 %indvars.iv 84 %0 = load i32, i32* %arrayidx, align 4 85 %arrayidx3 = getelementptr inbounds i32, i32* %x, i64 %indvars.iv 86 store i32 %0, i32* %arrayidx3, align 4 87 %1 = or i64 %indvars.iv, 1 88 %arrayidx5 = getelementptr inbounds i32, i32* %y, i64 %1 89 %2 = load i32, i32* %arrayidx5, align 4 90 %arrayidx8 = getelementptr inbounds i32, i32* %x, i64 %1 91 store i32 %2, i32* %arrayidx8, align 4 92 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 2 93 %cmp = icmp slt i64 %indvars.iv.next, %n 94 br i1 %cmp, label %for.body, label %for.end.loopexit 95 96for.end.loopexit: ; preds = %for.body 97 br label %for.end 98 99for.end: ; preds = %for.end.loopexit, %entry 100 ret i32 0 101} 102 103; Function Attrs: noinline norecurse nounwind 104define i32 @test3(i32 %n, i32* nocapture %x, i32* nocapture readonly %y) { 105entry: 106 %cmp21 = icmp sgt i32 %n, 0 107 br i1 %cmp21, label %for.body.preheader, label %for.end 108 109for.body.preheader: ; preds = %entry 110 br label %for.body 111 112for.body: ; preds = %for.body.preheader, %for.body 113 114;CHECK: for.body: 115;CHECK: %add12 = add i8 %i.022, 2 116;CHECK-NEXT: %conv = sext i8 %add12 to i32 117;CHECK-NEXT: %cmp = icmp slt i32 %conv, %n 118;CHECK-NEXT: br i1 %cmp, label %for.body, label %for.end.loopexit 119 120 %conv23 = phi i32 [ %conv, %for.body ], [ 0, %for.body.preheader ] 121 %i.022 = phi i8 [ %add12, %for.body ], [ 0, %for.body.preheader ] 122 %idxprom = sext i8 %i.022 to i64 123 %arrayidx = getelementptr inbounds i32, i32* %y, i64 %idxprom 124 %0 = load i32, i32* %arrayidx, align 4 125 %arrayidx3 = getelementptr inbounds i32, i32* %x, i64 %idxprom 126 store i32 %0, i32* %arrayidx3, align 4 127 %add = or i32 %conv23, 1 128 %idxprom5 = sext i32 %add to i64 129 %arrayidx6 = getelementptr inbounds i32, i32* %y, i64 %idxprom5 130 %1 = load i32, i32* %arrayidx6, align 4 131 %arrayidx10 = getelementptr inbounds i32, i32* %x, i64 %idxprom5 132 store i32 %1, i32* %arrayidx10, align 4 133 %add12 = add i8 %i.022, 2 134 %conv = sext i8 %add12 to i32 135 %cmp = icmp slt i32 %conv, %n 136 br i1 %cmp, label %for.body, label %for.end.loopexit 137 138for.end.loopexit: ; preds = %for.body 139 br label %for.end 140 141for.end: ; preds = %for.end.loopexit, %entry 142 ret i32 0 143} 144 145; Function Attrs: noinline norecurse nounwind 146define i32 @test4(i64 %n, i32* nocapture %x, i32* nocapture readonly %y) { 147entry: 148 %cmp18 = icmp eq i64 %n, 0 149 br i1 %cmp18, label %for.end, label %for.body.preheader 150 151for.body.preheader: ; preds = %entry 152 br label %for.body 153 154for.body: ; preds = %for.body.preheader, %for.body 155 156;CHECK: for.body: 157;CHECK-NEXT: %indvar = phi i64 [ %indvar.next, %for.body ], [ 0, %for.body.preheader ] 158;CHECK-NEXT: %arrayidx = getelementptr inbounds i32, i32* %y, i64 %indvar 159;CHECK-NEXT: [[T1:%[0-9]+]] = load i32, i32* %arrayidx, align 4 160;CHECK-NEXT: %arrayidx3 = getelementptr inbounds i32, i32* %x, i64 %indvar 161;CHECK-NEXT: store i32 [[T1]], i32* %arrayidx3, align 4 162;CHECK-NEXT: %indvar.next = add i64 %indvar, 1 163;CHECK-NEXT: %exitcond = icmp eq i64 %indvar, %{{[0-9]+}} 164;CHECK-NEXT: br i1 %exitcond, label %for.end.loopexit, label %for.body 165 166 %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %for.body.preheader ] 167 %arrayidx = getelementptr inbounds i32, i32* %y, i64 %indvars.iv 168 %0 = load i32, i32* %arrayidx, align 4 169 %arrayidx3 = getelementptr inbounds i32, i32* %x, i64 %indvars.iv 170 store i32 %0, i32* %arrayidx3, align 4 171 %1 = or i64 %indvars.iv, 1 172 %arrayidx5 = getelementptr inbounds i32, i32* %y, i64 %1 173 %2 = load i32, i32* %arrayidx5, align 4 174 %arrayidx8 = getelementptr inbounds i32, i32* %x, i64 %1 175 store i32 %2, i32* %arrayidx8, align 4 176 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 2 177 %cmp = icmp ult i64 %indvars.iv.next, %n 178 br i1 %cmp, label %for.body, label %for.end.loopexit 179 180for.end.loopexit: ; preds = %for.body 181 br label %for.end 182 183for.end: ; preds = %for.end.loopexit, %entry 184 ret i32 0 185} 186 187