1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py 2; RUN: %if x86-registered-target %{ opt < %s -passes=slp-vectorizer -S -mtriple=x86_64-unknown-linux-gnu | FileCheck %s %} 3; RUN: %if aarch64-registered-target %{ opt < %s -passes=slp-vectorizer -S -mtriple=aarch64-unknown-linux-gnu | FileCheck %s %} 4 5@b = common global [4 x i32] zeroinitializer, align 16 6@c = common global [4 x i32] zeroinitializer, align 16 7@d = common global [4 x i32] zeroinitializer, align 16 8@e = common global [4 x i32] zeroinitializer, align 16 9@a = common global [4 x i32] zeroinitializer, align 16 10@fb = common global [4 x float] zeroinitializer, align 16 11@fc = common global [4 x float] zeroinitializer, align 16 12@fa = common global [4 x float] zeroinitializer, align 16 13@fd = common global [4 x float] zeroinitializer, align 16 14 15; Function Attrs: nounwind uwtable 16define void @addsub() #0 { 17; CHECK-LABEL: @addsub( 18; CHECK-NEXT: entry: 19; CHECK-NEXT: [[TMP0:%.*]] = load <4 x i32>, ptr @b, align 4 20; CHECK-NEXT: [[TMP1:%.*]] = load <4 x i32>, ptr @c, align 4 21; CHECK-NEXT: [[TMP2:%.*]] = add nsw <4 x i32> [[TMP0]], [[TMP1]] 22; CHECK-NEXT: [[TMP3:%.*]] = load <4 x i32>, ptr @d, align 4 23; CHECK-NEXT: [[TMP4:%.*]] = load <4 x i32>, ptr @e, align 4 24; CHECK-NEXT: [[TMP5:%.*]] = add nsw <4 x i32> [[TMP3]], [[TMP4]] 25; CHECK-NEXT: [[TMP6:%.*]] = add nsw <4 x i32> [[TMP2]], [[TMP5]] 26; CHECK-NEXT: [[TMP7:%.*]] = sub nsw <4 x i32> [[TMP2]], [[TMP5]] 27; CHECK-NEXT: [[TMP8:%.*]] = shufflevector <4 x i32> [[TMP6]], <4 x i32> [[TMP7]], <4 x i32> <i32 0, i32 5, i32 2, i32 7> 28; CHECK-NEXT: store <4 x i32> [[TMP8]], ptr @a, align 4 29; CHECK-NEXT: ret void 30; 31entry: 32 %0 = load i32, ptr @b, align 4 33 %1 = load i32, ptr @c, align 4 34 %add = add nsw i32 %0, %1 35 %2 = load i32, ptr @d, align 4 36 %3 = load i32, ptr @e, align 4 37 %add1 = add nsw i32 %2, %3 38 %add2 = add nsw i32 %add, %add1 39 store i32 %add2, ptr @a, align 4 40 %4 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @b, i32 0, i64 1), align 4 41 %5 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @c, i32 0, i64 1), align 4 42 %add3 = add nsw i32 %4, %5 43 %6 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @d, i32 0, i64 1), align 4 44 %7 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @e, i32 0, i64 1), align 4 45 %add4 = add nsw i32 %6, %7 46 %sub = sub nsw i32 %add3, %add4 47 store i32 %sub, ptr getelementptr inbounds ([4 x i32], ptr @a, i32 0, i64 1), align 4 48 %8 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @b, i32 0, i64 2), align 4 49 %9 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @c, i32 0, i64 2), align 4 50 %add5 = add nsw i32 %8, %9 51 %10 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @d, i32 0, i64 2), align 4 52 %11 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @e, i32 0, i64 2), align 4 53 %add6 = add nsw i32 %10, %11 54 %add7 = add nsw i32 %add5, %add6 55 store i32 %add7, ptr getelementptr inbounds ([4 x i32], ptr @a, i32 0, i64 2), align 4 56 %12 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @b, i32 0, i64 3), align 4 57 %13 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @c, i32 0, i64 3), align 4 58 %add8 = add nsw i32 %12, %13 59 %14 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @d, i32 0, i64 3), align 4 60 %15 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @e, i32 0, i64 3), align 4 61 %add9 = add nsw i32 %14, %15 62 %sub10 = sub nsw i32 %add8, %add9 63 store i32 %sub10, ptr getelementptr inbounds ([4 x i32], ptr @a, i32 0, i64 3), align 4 64 ret void 65} 66 67define void @addsub_freeze() #0 { 68; CHECK-LABEL: @addsub_freeze( 69; CHECK-NEXT: entry: 70; CHECK-NEXT: [[TMP0:%.*]] = load <4 x i32>, ptr @b, align 4 71; CHECK-NEXT: [[TMP1:%.*]] = load <4 x i32>, ptr @c, align 4 72; CHECK-NEXT: [[TMP2:%.*]] = add nsw <4 x i32> [[TMP0]], [[TMP1]] 73; CHECK-NEXT: [[TMP3:%.*]] = load <4 x i32>, ptr @d, align 4 74; CHECK-NEXT: [[TMP4:%.*]] = load <4 x i32>, ptr @e, align 4 75; CHECK-NEXT: [[TMP5:%.*]] = add nsw <4 x i32> [[TMP3]], [[TMP4]] 76; CHECK-NEXT: [[TMP6:%.*]] = add nsw <4 x i32> [[TMP2]], [[TMP5]] 77; CHECK-NEXT: [[TMP7:%.*]] = sub nsw <4 x i32> [[TMP2]], [[TMP5]] 78; CHECK-NEXT: [[TMP8:%.*]] = shufflevector <4 x i32> [[TMP6]], <4 x i32> [[TMP7]], <4 x i32> <i32 0, i32 5, i32 2, i32 7> 79; CHECK-NEXT: [[TMP9:%.*]] = freeze <4 x i32> [[TMP8]] 80; CHECK-NEXT: store <4 x i32> [[TMP9]], ptr @a, align 4 81; CHECK-NEXT: ret void 82; 83entry: 84 %0 = load i32, ptr @b, align 4 85 %1 = load i32, ptr @c, align 4 86 %add = add nsw i32 %0, %1 87 %2 = load i32, ptr @d, align 4 88 %3 = load i32, ptr @e, align 4 89 %add1 = add nsw i32 %2, %3 90 %add2 = add nsw i32 %add, %add1 91 %freeze.add2 = freeze i32 %add2 92 store i32 %freeze.add2, ptr @a, align 4 93 %4 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @b, i32 0, i64 1), align 4 94 %5 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @c, i32 0, i64 1), align 4 95 %add3 = add nsw i32 %4, %5 96 %6 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @d, i32 0, i64 1), align 4 97 %7 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @e, i32 0, i64 1), align 4 98 %add4 = add nsw i32 %6, %7 99 %sub = sub nsw i32 %add3, %add4 100 %freeze.sub = freeze i32 %sub 101 store i32 %freeze.sub, ptr getelementptr inbounds ([4 x i32], ptr @a, i32 0, i64 1), align 4 102 %8 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @b, i32 0, i64 2), align 4 103 %9 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @c, i32 0, i64 2), align 4 104 %add5 = add nsw i32 %8, %9 105 %10 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @d, i32 0, i64 2), align 4 106 %11 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @e, i32 0, i64 2), align 4 107 %add6 = add nsw i32 %10, %11 108 %add7 = add nsw i32 %add5, %add6 109 %freeze.add7 = freeze i32 %add7 110 store i32 %freeze.add7, ptr getelementptr inbounds ([4 x i32], ptr @a, i32 0, i64 2), align 4 111 %12 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @b, i32 0, i64 3), align 4 112 %13 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @c, i32 0, i64 3), align 4 113 %add8 = add nsw i32 %12, %13 114 %14 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @d, i32 0, i64 3), align 4 115 %15 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @e, i32 0, i64 3), align 4 116 %add9 = add nsw i32 %14, %15 117 %sub10 = sub nsw i32 %add8, %add9 118 %freeze.sub10 = freeze i32 %sub10 119 store i32 %freeze.sub10, ptr getelementptr inbounds ([4 x i32], ptr @a, i32 0, i64 3), align 4 120 ret void 121} 122 123; Function Attrs: nounwind uwtable 124define void @subadd() #0 { 125; CHECK-LABEL: @subadd( 126; CHECK-NEXT: entry: 127; CHECK-NEXT: [[TMP0:%.*]] = load <4 x i32>, ptr @b, align 4 128; CHECK-NEXT: [[TMP1:%.*]] = load <4 x i32>, ptr @c, align 4 129; CHECK-NEXT: [[TMP2:%.*]] = add nsw <4 x i32> [[TMP0]], [[TMP1]] 130; CHECK-NEXT: [[TMP3:%.*]] = load <4 x i32>, ptr @d, align 4 131; CHECK-NEXT: [[TMP4:%.*]] = load <4 x i32>, ptr @e, align 4 132; CHECK-NEXT: [[TMP5:%.*]] = add nsw <4 x i32> [[TMP3]], [[TMP4]] 133; CHECK-NEXT: [[TMP6:%.*]] = sub nsw <4 x i32> [[TMP2]], [[TMP5]] 134; CHECK-NEXT: [[TMP7:%.*]] = add nsw <4 x i32> [[TMP2]], [[TMP5]] 135; CHECK-NEXT: [[TMP8:%.*]] = shufflevector <4 x i32> [[TMP6]], <4 x i32> [[TMP7]], <4 x i32> <i32 0, i32 5, i32 2, i32 7> 136; CHECK-NEXT: store <4 x i32> [[TMP8]], ptr @a, align 4 137; CHECK-NEXT: ret void 138; 139entry: 140 %0 = load i32, ptr @b, align 4 141 %1 = load i32, ptr @c, align 4 142 %add = add nsw i32 %0, %1 143 %2 = load i32, ptr @d, align 4 144 %3 = load i32, ptr @e, align 4 145 %add1 = add nsw i32 %2, %3 146 %sub = sub nsw i32 %add, %add1 147 store i32 %sub, ptr @a, align 4 148 %4 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @b, i32 0, i64 1), align 4 149 %5 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @c, i32 0, i64 1), align 4 150 %add2 = add nsw i32 %4, %5 151 %6 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @d, i32 0, i64 1), align 4 152 %7 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @e, i32 0, i64 1), align 4 153 %add3 = add nsw i32 %6, %7 154 %add4 = add nsw i32 %add2, %add3 155 store i32 %add4, ptr getelementptr inbounds ([4 x i32], ptr @a, i32 0, i64 1), align 4 156 %8 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @b, i32 0, i64 2), align 4 157 %9 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @c, i32 0, i64 2), align 4 158 %add5 = add nsw i32 %8, %9 159 %10 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @d, i32 0, i64 2), align 4 160 %11 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @e, i32 0, i64 2), align 4 161 %add6 = add nsw i32 %10, %11 162 %sub7 = sub nsw i32 %add5, %add6 163 store i32 %sub7, ptr getelementptr inbounds ([4 x i32], ptr @a, i32 0, i64 2), align 4 164 %12 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @b, i32 0, i64 3), align 4 165 %13 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @c, i32 0, i64 3), align 4 166 %add8 = add nsw i32 %12, %13 167 %14 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @d, i32 0, i64 3), align 4 168 %15 = load i32, ptr getelementptr inbounds ([4 x i32], ptr @e, i32 0, i64 3), align 4 169 %add9 = add nsw i32 %14, %15 170 %add10 = add nsw i32 %add8, %add9 171 store i32 %add10, ptr getelementptr inbounds ([4 x i32], ptr @a, i32 0, i64 3), align 4 172 ret void 173} 174 175; Function Attrs: nounwind uwtable 176define void @faddfsub() #0 { 177; CHECK-LABEL: @faddfsub( 178; CHECK-NEXT: entry: 179; CHECK-NEXT: [[TMP0:%.*]] = load <4 x float>, ptr @fb, align 4 180; CHECK-NEXT: [[TMP1:%.*]] = load <4 x float>, ptr @fc, align 4 181; CHECK-NEXT: [[TMP2:%.*]] = fadd <4 x float> [[TMP0]], [[TMP1]] 182; CHECK-NEXT: [[TMP3:%.*]] = fsub <4 x float> [[TMP0]], [[TMP1]] 183; CHECK-NEXT: [[TMP4:%.*]] = shufflevector <4 x float> [[TMP2]], <4 x float> [[TMP3]], <4 x i32> <i32 0, i32 5, i32 2, i32 7> 184; CHECK-NEXT: store <4 x float> [[TMP4]], ptr @fa, align 4 185; CHECK-NEXT: ret void 186; 187entry: 188 %0 = load float, ptr @fb, align 4 189 %1 = load float, ptr @fc, align 4 190 %add = fadd float %0, %1 191 store float %add, ptr @fa, align 4 192 %2 = load float, ptr getelementptr inbounds ([4 x float], ptr @fb, i32 0, i64 1), align 4 193 %3 = load float, ptr getelementptr inbounds ([4 x float], ptr @fc, i32 0, i64 1), align 4 194 %sub = fsub float %2, %3 195 store float %sub, ptr getelementptr inbounds ([4 x float], ptr @fa, i32 0, i64 1), align 4 196 %4 = load float, ptr getelementptr inbounds ([4 x float], ptr @fb, i32 0, i64 2), align 4 197 %5 = load float, ptr getelementptr inbounds ([4 x float], ptr @fc, i32 0, i64 2), align 4 198 %add1 = fadd float %4, %5 199 store float %add1, ptr getelementptr inbounds ([4 x float], ptr @fa, i32 0, i64 2), align 4 200 %6 = load float, ptr getelementptr inbounds ([4 x float], ptr @fb, i32 0, i64 3), align 4 201 %7 = load float, ptr getelementptr inbounds ([4 x float], ptr @fc, i32 0, i64 3), align 4 202 %sub2 = fsub float %6, %7 203 store float %sub2, ptr getelementptr inbounds ([4 x float], ptr @fa, i32 0, i64 3), align 4 204 ret void 205} 206 207; Function Attrs: nounwind uwtable 208define void @fsubfadd() #0 { 209; CHECK-LABEL: @fsubfadd( 210; CHECK-NEXT: entry: 211; CHECK-NEXT: [[TMP0:%.*]] = load <4 x float>, ptr @fb, align 4 212; CHECK-NEXT: [[TMP1:%.*]] = load <4 x float>, ptr @fc, align 4 213; CHECK-NEXT: [[TMP2:%.*]] = fsub <4 x float> [[TMP0]], [[TMP1]] 214; CHECK-NEXT: [[TMP3:%.*]] = fadd <4 x float> [[TMP0]], [[TMP1]] 215; CHECK-NEXT: [[TMP4:%.*]] = shufflevector <4 x float> [[TMP2]], <4 x float> [[TMP3]], <4 x i32> <i32 0, i32 5, i32 2, i32 7> 216; CHECK-NEXT: store <4 x float> [[TMP4]], ptr @fa, align 4 217; CHECK-NEXT: ret void 218; 219entry: 220 %0 = load float, ptr @fb, align 4 221 %1 = load float, ptr @fc, align 4 222 %sub = fsub float %0, %1 223 store float %sub, ptr @fa, align 4 224 %2 = load float, ptr getelementptr inbounds ([4 x float], ptr @fb, i32 0, i64 1), align 4 225 %3 = load float, ptr getelementptr inbounds ([4 x float], ptr @fc, i32 0, i64 1), align 4 226 %add = fadd float %2, %3 227 store float %add, ptr getelementptr inbounds ([4 x float], ptr @fa, i32 0, i64 1), align 4 228 %4 = load float, ptr getelementptr inbounds ([4 x float], ptr @fb, i32 0, i64 2), align 4 229 %5 = load float, ptr getelementptr inbounds ([4 x float], ptr @fc, i32 0, i64 2), align 4 230 %sub1 = fsub float %4, %5 231 store float %sub1, ptr getelementptr inbounds ([4 x float], ptr @fa, i32 0, i64 2), align 4 232 %6 = load float, ptr getelementptr inbounds ([4 x float], ptr @fb, i32 0, i64 3), align 4 233 %7 = load float, ptr getelementptr inbounds ([4 x float], ptr @fc, i32 0, i64 3), align 4 234 %add2 = fadd float %6, %7 235 store float %add2, ptr getelementptr inbounds ([4 x float], ptr @fa, i32 0, i64 3), align 4 236 ret void 237} 238 239; Function Attrs: nounwind uwtable 240define void @faddfsub_select() #0 { 241; CHECK-LABEL: @faddfsub_select( 242; CHECK-NEXT: entry: 243; CHECK-NEXT: [[TMP0:%.*]] = load <4 x float>, ptr @fb, align 4 244; CHECK-NEXT: [[TMP1:%.*]] = load <4 x float>, ptr @fc, align 4 245; CHECK-NEXT: [[TMP2:%.*]] = fadd <4 x float> [[TMP0]], [[TMP1]] 246; CHECK-NEXT: [[TMP3:%.*]] = fsub <4 x float> [[TMP0]], [[TMP1]] 247; CHECK-NEXT: [[TMP4:%.*]] = shufflevector <4 x float> [[TMP2]], <4 x float> [[TMP3]], <4 x i32> <i32 0, i32 1, i32 2, i32 7> 248; CHECK-NEXT: store <4 x float> [[TMP4]], ptr @fa, align 4 249; CHECK-NEXT: ret void 250; 251entry: 252 %0 = load float, ptr @fb, align 4 253 %1 = load float, ptr @fc, align 4 254 %add = fadd float %0, %1 255 store float %add, ptr @fa, align 4 256 %2 = load float, ptr getelementptr inbounds ([4 x float], ptr @fb, i32 0, i64 1), align 4 257 %3 = load float, ptr getelementptr inbounds ([4 x float], ptr @fc, i32 0, i64 1), align 4 258 %add1 = fadd float %2, %3 259 store float %add1, ptr getelementptr inbounds ([4 x float], ptr @fa, i32 0, i64 1), align 4 260 %4 = load float, ptr getelementptr inbounds ([4 x float], ptr @fb, i32 0, i64 2), align 4 261 %5 = load float, ptr getelementptr inbounds ([4 x float], ptr @fc, i32 0, i64 2), align 4 262 %add2 = fadd float %4, %5 263 store float %add2, ptr getelementptr inbounds ([4 x float], ptr @fa, i32 0, i64 2), align 4 264 %6 = load float, ptr getelementptr inbounds ([4 x float], ptr @fb, i32 0, i64 3), align 4 265 %7 = load float, ptr getelementptr inbounds ([4 x float], ptr @fc, i32 0, i64 3), align 4 266 %sub = fsub float %6, %7 267 store float %sub, ptr getelementptr inbounds ([4 x float], ptr @fa, i32 0, i64 3), align 4 268 ret void 269} 270 271; Check vectorization of following code for float data type- 272; fc[0] = fb[0]+fa[0]; //swapped fb and fa 273; fc[1] = fa[1]-fb[1]; 274; fc[2] = fa[2]+fb[2]; 275; fc[3] = fa[3]-fb[3]; 276 277define void @reorder_alt() #0 { 278; CHECK-LABEL: @reorder_alt( 279; CHECK-NEXT: [[TMP1:%.*]] = load <4 x float>, ptr @fa, align 4 280; CHECK-NEXT: [[TMP2:%.*]] = load <4 x float>, ptr @fb, align 4 281; CHECK-NEXT: [[TMP3:%.*]] = fadd <4 x float> [[TMP1]], [[TMP2]] 282; CHECK-NEXT: [[TMP4:%.*]] = fsub <4 x float> [[TMP1]], [[TMP2]] 283; CHECK-NEXT: [[TMP5:%.*]] = shufflevector <4 x float> [[TMP3]], <4 x float> [[TMP4]], <4 x i32> <i32 0, i32 5, i32 2, i32 7> 284; CHECK-NEXT: store <4 x float> [[TMP5]], ptr @fc, align 4 285; CHECK-NEXT: ret void 286; 287 %1 = load float, ptr @fb, align 4 288 %2 = load float, ptr @fa, align 4 289 %3 = fadd float %1, %2 290 store float %3, ptr @fc, align 4 291 %4 = load float, ptr getelementptr inbounds ([4 x float], ptr @fa, i32 0, i64 1), align 4 292 %5 = load float, ptr getelementptr inbounds ([4 x float], ptr @fb, i32 0, i64 1), align 4 293 %6 = fsub float %4, %5 294 store float %6, ptr getelementptr inbounds ([4 x float], ptr @fc, i32 0, i64 1), align 4 295 %7 = load float, ptr getelementptr inbounds ([4 x float], ptr @fa, i32 0, i64 2), align 4 296 %8 = load float, ptr getelementptr inbounds ([4 x float], ptr @fb, i32 0, i64 2), align 4 297 %9 = fadd float %7, %8 298 store float %9, ptr getelementptr inbounds ([4 x float], ptr @fc, i32 0, i64 2), align 4 299 %10 = load float, ptr getelementptr inbounds ([4 x float], ptr @fa, i32 0, i64 3), align 4 300 %11 = load float, ptr getelementptr inbounds ([4 x float], ptr @fb, i32 0, i64 3), align 4 301 %12 = fsub float %10, %11 302 store float %12, ptr getelementptr inbounds ([4 x float], ptr @fc, i32 0, i64 3), align 4 303 ret void 304} 305 306; Check vectorization of following code for float data type- 307; fc[0] = fa[0]+(fb[0]-fd[0]); 308; fc[1] = fa[1]-(fb[1]+fd[1]); 309; fc[2] = fa[2]+(fb[2]-fd[2]); 310; fc[3] = fa[3]-(fd[3]+fb[3]); //swapped fd and fb 311 312define void @reorder_alt_subTree() #0 { 313; CHECK-LABEL: @reorder_alt_subTree( 314; CHECK-NEXT: [[TMP1:%.*]] = load <4 x float>, ptr @fa, align 4 315; CHECK-NEXT: [[TMP2:%.*]] = load <4 x float>, ptr @fd, align 4 316; CHECK-NEXT: [[TMP3:%.*]] = load <4 x float>, ptr @fb, align 4 317; CHECK-NEXT: [[TMP4:%.*]] = fsub <4 x float> [[TMP3]], [[TMP2]] 318; CHECK-NEXT: [[TMP5:%.*]] = fadd <4 x float> [[TMP3]], [[TMP2]] 319; CHECK-NEXT: [[TMP6:%.*]] = shufflevector <4 x float> [[TMP4]], <4 x float> [[TMP5]], <4 x i32> <i32 0, i32 5, i32 2, i32 7> 320; CHECK-NEXT: [[TMP7:%.*]] = fadd <4 x float> [[TMP1]], [[TMP6]] 321; CHECK-NEXT: [[TMP8:%.*]] = fsub <4 x float> [[TMP1]], [[TMP6]] 322; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <4 x float> [[TMP7]], <4 x float> [[TMP8]], <4 x i32> <i32 0, i32 5, i32 2, i32 7> 323; CHECK-NEXT: store <4 x float> [[TMP9]], ptr @fc, align 4 324; CHECK-NEXT: ret void 325; 326 %1 = load float, ptr @fa, align 4 327 %2 = load float, ptr @fb, align 4 328 %3 = load float, ptr @fd, align 4 329 %4 = fsub float %2, %3 330 %5 = fadd float %1, %4 331 store float %5, ptr @fc, align 4 332 %6 = load float, ptr getelementptr inbounds ([4 x float], ptr @fa, i32 0, i64 1), align 4 333 %7 = load float, ptr getelementptr inbounds ([4 x float], ptr @fb, i32 0, i64 1), align 4 334 %8 = load float, ptr getelementptr inbounds ([4 x float], ptr @fd, i32 0, i64 1), align 4 335 %9 = fadd float %7, %8 336 %10 = fsub float %6, %9 337 store float %10, ptr getelementptr inbounds ([4 x float], ptr @fc, i32 0, i64 1), align 4 338 %11 = load float, ptr getelementptr inbounds ([4 x float], ptr @fa, i32 0, i64 2), align 4 339 %12 = load float, ptr getelementptr inbounds ([4 x float], ptr @fb, i32 0, i64 2), align 4 340 %13 = load float, ptr getelementptr inbounds ([4 x float], ptr @fd, i32 0, i64 2), align 4 341 %14 = fsub float %12, %13 342 %15 = fadd float %11, %14 343 store float %15, ptr getelementptr inbounds ([4 x float], ptr @fc, i32 0, i64 2), align 4 344 %16 = load float, ptr getelementptr inbounds ([4 x float], ptr @fa, i32 0, i64 3), align 4 345 %17 = load float, ptr getelementptr inbounds ([4 x float], ptr @fd, i32 0, i64 3), align 4 346 %18 = load float, ptr getelementptr inbounds ([4 x float], ptr @fb, i32 0, i64 3), align 4 347 %19 = fadd float %17, %18 348 %20 = fsub float %16, %19 349 store float %20, ptr getelementptr inbounds ([4 x float], ptr @fc, i32 0, i64 3), align 4 350 ret void 351} 352 353; Check vectorization of following code for double data type- 354; c[0] = (a[0]+b[0])-d[0]; 355; c[1] = d[1]+(a[1]+b[1]); //swapped d[1] and (a[1]+b[1]) 356 357define void @reorder_alt_rightsubTree(ptr nocapture %c, ptr noalias nocapture readonly %a, ptr noalias nocapture readonly %b, ptr noalias nocapture readonly %d) { 358; CHECK-LABEL: @reorder_alt_rightsubTree( 359; CHECK-NEXT: [[TMP1:%.*]] = load <2 x double>, ptr [[D:%.*]], align 8 360; CHECK-NEXT: [[TMP2:%.*]] = load <2 x double>, ptr [[A:%.*]], align 8 361; CHECK-NEXT: [[TMP3:%.*]] = load <2 x double>, ptr [[B:%.*]], align 8 362; CHECK-NEXT: [[TMP4:%.*]] = fadd <2 x double> [[TMP2]], [[TMP3]] 363; CHECK-NEXT: [[TMP5:%.*]] = fsub <2 x double> [[TMP4]], [[TMP1]] 364; CHECK-NEXT: [[TMP6:%.*]] = fadd <2 x double> [[TMP4]], [[TMP1]] 365; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <2 x double> [[TMP5]], <2 x double> [[TMP6]], <2 x i32> <i32 0, i32 3> 366; CHECK-NEXT: store <2 x double> [[TMP7]], ptr [[C:%.*]], align 8 367; CHECK-NEXT: ret void 368; 369 %1 = load double, ptr %a 370 %2 = load double, ptr %b 371 %3 = fadd double %1, %2 372 %4 = load double, ptr %d 373 %5 = fsub double %3, %4 374 store double %5, ptr %c 375 %6 = getelementptr inbounds double, ptr %d, i64 1 376 %7 = load double, ptr %6 377 %8 = getelementptr inbounds double, ptr %a, i64 1 378 %9 = load double, ptr %8 379 %10 = getelementptr inbounds double, ptr %b, i64 1 380 %11 = load double, ptr %10 381 %12 = fadd double %9, %11 382 %13 = fadd double %7, %12 383 %14 = getelementptr inbounds double, ptr %c, i64 1 384 store double %13, ptr %14 385 ret void 386} 387 388define void @vec_shuff_reorder() #0 { 389; CHECK-LABEL: @vec_shuff_reorder( 390; CHECK-NEXT: [[TMP1:%.*]] = load <2 x float>, ptr @fa, align 4 391; CHECK-NEXT: [[TMP2:%.*]] = load <2 x float>, ptr @fb, align 4 392; CHECK-NEXT: [[TMP3:%.*]] = load <2 x float>, ptr getelementptr inbounds ([4 x float], ptr @fb, i32 0, i64 2), align 4 393; CHECK-NEXT: [[TMP4:%.*]] = load <2 x float>, ptr getelementptr inbounds ([4 x float], ptr @fa, i32 0, i64 2), align 4 394; CHECK-NEXT: [[TMP5:%.*]] = call <4 x float> @llvm.vector.insert.v4f32.v2f32(<4 x float> poison, <2 x float> [[TMP1]], i64 0) 395; CHECK-NEXT: [[TMP6:%.*]] = call <4 x float> @llvm.vector.insert.v4f32.v2f32(<4 x float> [[TMP5]], <2 x float> [[TMP3]], i64 2) 396; CHECK-NEXT: [[TMP7:%.*]] = call <4 x float> @llvm.vector.insert.v4f32.v2f32(<4 x float> poison, <2 x float> [[TMP2]], i64 0) 397; CHECK-NEXT: [[TMP8:%.*]] = call <4 x float> @llvm.vector.insert.v4f32.v2f32(<4 x float> [[TMP7]], <2 x float> [[TMP4]], i64 2) 398; CHECK-NEXT: [[TMP9:%.*]] = fadd <4 x float> [[TMP6]], [[TMP8]] 399; CHECK-NEXT: [[TMP10:%.*]] = fsub <4 x float> [[TMP6]], [[TMP8]] 400; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <4 x float> [[TMP9]], <4 x float> [[TMP10]], <4 x i32> <i32 0, i32 5, i32 2, i32 7> 401; CHECK-NEXT: store <4 x float> [[TMP11]], ptr @fc, align 4 402; CHECK-NEXT: ret void 403; 404 %1 = load float, ptr @fb, align 4 405 %2 = load float, ptr @fa, align 4 406 %3 = fadd float %1, %2 407 store float %3, ptr @fc, align 4 408 %4 = load float, ptr getelementptr inbounds ([4 x float], ptr @fa, i32 0, i64 1), align 4 409 %5 = load float, ptr getelementptr inbounds ([4 x float], ptr @fb, i32 0, i64 1), align 4 410 %6 = fsub float %4, %5 411 store float %6, ptr getelementptr inbounds ([4 x float], ptr @fc, i32 0, i64 1), align 4 412 %7 = load float, ptr getelementptr inbounds ([4 x float], ptr @fa, i32 0, i64 2), align 4 413 %8 = load float, ptr getelementptr inbounds ([4 x float], ptr @fb, i32 0, i64 2), align 4 414 %9 = fadd float %7, %8 415 store float %9, ptr getelementptr inbounds ([4 x float], ptr @fc, i32 0, i64 2), align 4 416 %10 = load float, ptr getelementptr inbounds ([4 x float], ptr @fb, i32 0, i64 3), align 4 417 %11 = load float, ptr getelementptr inbounds ([4 x float], ptr @fa, i32 0, i64 3), align 4 418 %12 = fsub float %10, %11 419 store float %12, ptr getelementptr inbounds ([4 x float], ptr @fc, i32 0, i64 3), align 4 420 ret void 421} 422 423 424attributes #0 = { nounwind } 425 426