1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py 2; RUN: llc -mcpu=pwr9 -verify-machineinstrs -mtriple=powerpc64le-unknown-linux-gnu < %s | FileCheck %s 3 4; void foo(float *data, float d) { 5; long i; 6; for (i = 0; i < 8000; i++) 7; data[i] = d; 8; } 9; 10; This loop will be unrolled by 96 and vectorized on power9. 11; icmp for loop iteration index and loop trip count(384) has LSRUse for 'reg({0,+,384})'. 12; Make sure above icmp does not impact LSR choose best formulae sets based on 'reg({(192 + %0),+,384})' 13 14define void @foo(float* nocapture %data, float %d) { 15; CHECK-LABEL: foo: 16; CHECK: .LBB0_1: # %vector.body 17; CHECK: add 5, 3, 4 18; CHECK-NEXT: stxvx 0, 3, 4 19; CHECK-NEXT: addi 4, 4, 384 20; CHECK-NEXT: stxv 0, 16(5) 21; CHECK-NEXT: stxv 0, 32(5) 22; CHECK-NEXT: stxv 0, 48(5) 23; CHECK-NEXT: stxv 0, 64(5) 24; CHECK-NEXT: stxv 0, 80(5) 25; CHECK-NEXT: stxv 0, 96(5) 26; CHECK-NEXT: stxv 0, 112(5) 27; CHECK-NEXT: stxv 0, 128(5) 28; CHECK-NEXT: stxv 0, 144(5) 29; CHECK-NEXT: stxv 0, 160(5) 30; CHECK-NEXT: stxv 0, 176(5) 31; CHECK-NEXT: stxv 0, 192(5) 32; CHECK-NEXT: stxv 0, 208(5) 33; CHECK-NEXT: stxv 0, 224(5) 34; CHECK-NEXT: stxv 0, 240(5) 35; CHECK-NEXT: stxv 0, 256(5) 36; CHECK-NEXT: stxv 0, 272(5) 37; CHECK-NEXT: stxv 0, 288(5) 38; CHECK-NEXT: stxv 0, 304(5) 39; CHECK-NEXT: stxv 0, 320(5) 40; CHECK-NEXT: stxv 0, 336(5) 41; CHECK-NEXT: stxv 0, 352(5) 42; CHECK-NEXT: stxv 0, 368(5) 43; CHECK-NEXT: bdnz .LBB0_1 44 45entry: 46 %broadcast.splatinsert16 = insertelement <4 x float> undef, float %d, i32 0 47 %broadcast.splat17 = shufflevector <4 x float> %broadcast.splatinsert16, <4 x float> undef, <4 x i32> zeroinitializer 48 %broadcast.splatinsert18 = insertelement <4 x float> undef, float %d, i32 0 49 %broadcast.splat19 = shufflevector <4 x float> %broadcast.splatinsert18, <4 x float> undef, <4 x i32> zeroinitializer 50 %broadcast.splatinsert20 = insertelement <4 x float> undef, float %d, i32 0 51 %broadcast.splat21 = shufflevector <4 x float> %broadcast.splatinsert20, <4 x float> undef, <4 x i32> zeroinitializer 52 %broadcast.splatinsert22 = insertelement <4 x float> undef, float %d, i32 0 53 %broadcast.splat23 = shufflevector <4 x float> %broadcast.splatinsert22, <4 x float> undef, <4 x i32> zeroinitializer 54 %broadcast.splatinsert24 = insertelement <4 x float> undef, float %d, i32 0 55 %broadcast.splat25 = shufflevector <4 x float> %broadcast.splatinsert24, <4 x float> undef, <4 x i32> zeroinitializer 56 %broadcast.splatinsert26 = insertelement <4 x float> undef, float %d, i32 0 57 %broadcast.splat27 = shufflevector <4 x float> %broadcast.splatinsert26, <4 x float> undef, <4 x i32> zeroinitializer 58 %broadcast.splatinsert28 = insertelement <4 x float> undef, float %d, i32 0 59 %broadcast.splat29 = shufflevector <4 x float> %broadcast.splatinsert28, <4 x float> undef, <4 x i32> zeroinitializer 60 %broadcast.splatinsert30 = insertelement <4 x float> undef, float %d, i32 0 61 %broadcast.splat31 = shufflevector <4 x float> %broadcast.splatinsert30, <4 x float> undef, <4 x i32> zeroinitializer 62 %broadcast.splatinsert32 = insertelement <4 x float> undef, float %d, i32 0 63 %broadcast.splat33 = shufflevector <4 x float> %broadcast.splatinsert32, <4 x float> undef, <4 x i32> zeroinitializer 64 %broadcast.splatinsert34 = insertelement <4 x float> undef, float %d, i32 0 65 %broadcast.splat35 = shufflevector <4 x float> %broadcast.splatinsert34, <4 x float> undef, <4 x i32> zeroinitializer 66 %broadcast.splatinsert36 = insertelement <4 x float> undef, float %d, i32 0 67 %broadcast.splat37 = shufflevector <4 x float> %broadcast.splatinsert36, <4 x float> undef, <4 x i32> zeroinitializer 68 %broadcast.splatinsert38 = insertelement <4 x float> undef, float %d, i32 0 69 %broadcast.splat39 = shufflevector <4 x float> %broadcast.splatinsert38, <4 x float> undef, <4 x i32> zeroinitializer 70 br label %vector.body 71 72vector.body: ; preds = %vector.body, %entry 73 %index = phi i64 [ 0, %entry ], [ %index.next.1, %vector.body ] 74 %0 = getelementptr inbounds float, float* %data, i64 %index 75 %1 = bitcast float* %0 to <4 x float>* 76 store <4 x float> %broadcast.splat17, <4 x float>* %1, align 4 77 %2 = getelementptr inbounds float, float* %0, i64 4 78 %3 = bitcast float* %2 to <4 x float>* 79 store <4 x float> %broadcast.splat19, <4 x float>* %3, align 4 80 %4 = getelementptr inbounds float, float* %0, i64 8 81 %5 = bitcast float* %4 to <4 x float>* 82 store <4 x float> %broadcast.splat21, <4 x float>* %5, align 4 83 %6 = getelementptr inbounds float, float* %0, i64 12 84 %7 = bitcast float* %6 to <4 x float>* 85 store <4 x float> %broadcast.splat23, <4 x float>* %7, align 4 86 %8 = getelementptr inbounds float, float* %0, i64 16 87 %9 = bitcast float* %8 to <4 x float>* 88 store <4 x float> %broadcast.splat25, <4 x float>* %9, align 4 89 %10 = getelementptr inbounds float, float* %0, i64 20 90 %11 = bitcast float* %10 to <4 x float>* 91 store <4 x float> %broadcast.splat27, <4 x float>* %11, align 4 92 %12 = getelementptr inbounds float, float* %0, i64 24 93 %13 = bitcast float* %12 to <4 x float>* 94 store <4 x float> %broadcast.splat29, <4 x float>* %13, align 4 95 %14 = getelementptr inbounds float, float* %0, i64 28 96 %15 = bitcast float* %14 to <4 x float>* 97 store <4 x float> %broadcast.splat31, <4 x float>* %15, align 4 98 %16 = getelementptr inbounds float, float* %0, i64 32 99 %17 = bitcast float* %16 to <4 x float>* 100 store <4 x float> %broadcast.splat33, <4 x float>* %17, align 4 101 %18 = getelementptr inbounds float, float* %0, i64 36 102 %19 = bitcast float* %18 to <4 x float>* 103 store <4 x float> %broadcast.splat35, <4 x float>* %19, align 4 104 %20 = getelementptr inbounds float, float* %0, i64 40 105 %21 = bitcast float* %20 to <4 x float>* 106 store <4 x float> %broadcast.splat37, <4 x float>* %21, align 4 107 %22 = getelementptr inbounds float, float* %0, i64 44 108 %23 = bitcast float* %22 to <4 x float>* 109 store <4 x float> %broadcast.splat39, <4 x float>* %23, align 4 110 %index.next = add nuw nsw i64 %index, 48 111 %24 = getelementptr inbounds float, float* %data, i64 %index.next 112 %25 = bitcast float* %24 to <4 x float>* 113 store <4 x float> %broadcast.splat17, <4 x float>* %25, align 4 114 %26 = getelementptr inbounds float, float* %24, i64 4 115 %27 = bitcast float* %26 to <4 x float>* 116 store <4 x float> %broadcast.splat19, <4 x float>* %27, align 4 117 %28 = getelementptr inbounds float, float* %24, i64 8 118 %29 = bitcast float* %28 to <4 x float>* 119 store <4 x float> %broadcast.splat21, <4 x float>* %29, align 4 120 %30 = getelementptr inbounds float, float* %24, i64 12 121 %31 = bitcast float* %30 to <4 x float>* 122 store <4 x float> %broadcast.splat23, <4 x float>* %31, align 4 123 %32 = getelementptr inbounds float, float* %24, i64 16 124 %33 = bitcast float* %32 to <4 x float>* 125 store <4 x float> %broadcast.splat25, <4 x float>* %33, align 4 126 %34 = getelementptr inbounds float, float* %24, i64 20 127 %35 = bitcast float* %34 to <4 x float>* 128 store <4 x float> %broadcast.splat27, <4 x float>* %35, align 4 129 %36 = getelementptr inbounds float, float* %24, i64 24 130 %37 = bitcast float* %36 to <4 x float>* 131 store <4 x float> %broadcast.splat29, <4 x float>* %37, align 4 132 %38 = getelementptr inbounds float, float* %24, i64 28 133 %39 = bitcast float* %38 to <4 x float>* 134 store <4 x float> %broadcast.splat31, <4 x float>* %39, align 4 135 %40 = getelementptr inbounds float, float* %24, i64 32 136 %41 = bitcast float* %40 to <4 x float>* 137 store <4 x float> %broadcast.splat33, <4 x float>* %41, align 4 138 %42 = getelementptr inbounds float, float* %24, i64 36 139 %43 = bitcast float* %42 to <4 x float>* 140 store <4 x float> %broadcast.splat35, <4 x float>* %43, align 4 141 %44 = getelementptr inbounds float, float* %24, i64 40 142 %45 = bitcast float* %44 to <4 x float>* 143 store <4 x float> %broadcast.splat37, <4 x float>* %45, align 4 144 %46 = getelementptr inbounds float, float* %24, i64 44 145 %47 = bitcast float* %46 to <4 x float>* 146 store <4 x float> %broadcast.splat39, <4 x float>* %47, align 4 147 %index.next.1 = add nuw nsw i64 %index, 96 148 %48 = icmp eq i64 %index.next.1, 7968 149 br i1 %48, label %for.body, label %vector.body 150 151for.body: ; preds = %vector.body 152 %arrayidx = getelementptr inbounds float, float* %data, i64 7968 153 store float %d, float* %arrayidx, align 4 154 %arrayidx.1 = getelementptr inbounds float, float* %data, i64 7969 155 store float %d, float* %arrayidx.1, align 4 156 %arrayidx.2 = getelementptr inbounds float, float* %data, i64 7970 157 store float %d, float* %arrayidx.2, align 4 158 %arrayidx.3 = getelementptr inbounds float, float* %data, i64 7971 159 store float %d, float* %arrayidx.3, align 4 160 %arrayidx.4 = getelementptr inbounds float, float* %data, i64 7972 161 store float %d, float* %arrayidx.4, align 4 162 %arrayidx.5 = getelementptr inbounds float, float* %data, i64 7973 163 store float %d, float* %arrayidx.5, align 4 164 %arrayidx.6 = getelementptr inbounds float, float* %data, i64 7974 165 store float %d, float* %arrayidx.6, align 4 166 %arrayidx.7 = getelementptr inbounds float, float* %data, i64 7975 167 store float %d, float* %arrayidx.7, align 4 168 %arrayidx.8 = getelementptr inbounds float, float* %data, i64 7976 169 store float %d, float* %arrayidx.8, align 4 170 %arrayidx.9 = getelementptr inbounds float, float* %data, i64 7977 171 store float %d, float* %arrayidx.9, align 4 172 %arrayidx.10 = getelementptr inbounds float, float* %data, i64 7978 173 store float %d, float* %arrayidx.10, align 4 174 %arrayidx.11 = getelementptr inbounds float, float* %data, i64 7979 175 store float %d, float* %arrayidx.11, align 4 176 %arrayidx.12 = getelementptr inbounds float, float* %data, i64 7980 177 store float %d, float* %arrayidx.12, align 4 178 %arrayidx.13 = getelementptr inbounds float, float* %data, i64 7981 179 store float %d, float* %arrayidx.13, align 4 180 %arrayidx.14 = getelementptr inbounds float, float* %data, i64 7982 181 store float %d, float* %arrayidx.14, align 4 182 %arrayidx.15 = getelementptr inbounds float, float* %data, i64 7983 183 store float %d, float* %arrayidx.15, align 4 184 %arrayidx.16 = getelementptr inbounds float, float* %data, i64 7984 185 store float %d, float* %arrayidx.16, align 4 186 %arrayidx.17 = getelementptr inbounds float, float* %data, i64 7985 187 store float %d, float* %arrayidx.17, align 4 188 %arrayidx.18 = getelementptr inbounds float, float* %data, i64 7986 189 store float %d, float* %arrayidx.18, align 4 190 %arrayidx.19 = getelementptr inbounds float, float* %data, i64 7987 191 store float %d, float* %arrayidx.19, align 4 192 %arrayidx.20 = getelementptr inbounds float, float* %data, i64 7988 193 store float %d, float* %arrayidx.20, align 4 194 %arrayidx.21 = getelementptr inbounds float, float* %data, i64 7989 195 store float %d, float* %arrayidx.21, align 4 196 %arrayidx.22 = getelementptr inbounds float, float* %data, i64 7990 197 store float %d, float* %arrayidx.22, align 4 198 %arrayidx.23 = getelementptr inbounds float, float* %data, i64 7991 199 store float %d, float* %arrayidx.23, align 4 200 %arrayidx.24 = getelementptr inbounds float, float* %data, i64 7992 201 store float %d, float* %arrayidx.24, align 4 202 %arrayidx.25 = getelementptr inbounds float, float* %data, i64 7993 203 store float %d, float* %arrayidx.25, align 4 204 %arrayidx.26 = getelementptr inbounds float, float* %data, i64 7994 205 store float %d, float* %arrayidx.26, align 4 206 %arrayidx.27 = getelementptr inbounds float, float* %data, i64 7995 207 store float %d, float* %arrayidx.27, align 4 208 %arrayidx.28 = getelementptr inbounds float, float* %data, i64 7996 209 store float %d, float* %arrayidx.28, align 4 210 %arrayidx.29 = getelementptr inbounds float, float* %data, i64 7997 211 store float %d, float* %arrayidx.29, align 4 212 %arrayidx.30 = getelementptr inbounds float, float* %data, i64 7998 213 store float %d, float* %arrayidx.30, align 4 214 %arrayidx.31 = getelementptr inbounds float, float* %data, i64 7999 215 store float %d, float* %arrayidx.31, align 4 216 ret void 217} 218