1; RUN: llc -verify-machineinstrs -mcpu=pwr9 \ 2; RUN: -mtriple=powerpc64le-unknown-unknown < %s | FileCheck %s 3 4; RUN: llc -verify-machineinstrs -mcpu=pwr9 \ 5; RUN: -mtriple=powerpc64-unknown-unknown < %s | FileCheck %s 6 7@uca = global <16 x i8> zeroinitializer, align 16 8@ucb = global <16 x i8> zeroinitializer, align 16 9@sca = global <16 x i8> zeroinitializer, align 16 10@scb = global <16 x i8> zeroinitializer, align 16 11@usa = global <8 x i16> zeroinitializer, align 16 12@usb = global <8 x i16> zeroinitializer, align 16 13@ssa = global <8 x i16> zeroinitializer, align 16 14@ssb = global <8 x i16> zeroinitializer, align 16 15@uia = global <4 x i32> zeroinitializer, align 16 16@uib = global <4 x i32> zeroinitializer, align 16 17@sia = global <4 x i32> zeroinitializer, align 16 18@sib = global <4 x i32> zeroinitializer, align 16 19@ulla = global <2 x i64> zeroinitializer, align 16 20@ullb = global <2 x i64> zeroinitializer, align 16 21@slla = global <2 x i64> zeroinitializer, align 16 22@sllb = global <2 x i64> zeroinitializer, align 16 23@uxa = global <1 x i128> zeroinitializer, align 16 24@uxb = global <1 x i128> zeroinitializer, align 16 25@sxa = global <1 x i128> zeroinitializer, align 16 26@sxb = global <1 x i128> zeroinitializer, align 16 27@vfa = global <4 x float> zeroinitializer, align 16 28@vfb = global <4 x float> zeroinitializer, align 16 29@vda = global <2 x double> zeroinitializer, align 16 30@vdb = global <2 x double> zeroinitializer, align 16 31 32define void @_Z4testv() { 33entry: 34; CHECK-LABEL: @_Z4testv 35 %0 = load <16 x i8>, <16 x i8>* @uca, align 16 36 %1 = load <16 x i8>, <16 x i8>* @ucb, align 16 37 %add.i = add <16 x i8> %1, %0 38 tail call void (...) @sink(<16 x i8> %add.i) 39; CHECK: lxvx 34, 0, 3 40; CHECK: lxvx 35, 0, 4 41; CHECK: vaddubm 2, 3, 2 42; CHECK: stxv 34, 43; CHECK: bl sink 44 %2 = load <16 x i8>, <16 x i8>* @sca, align 16 45 %3 = load <16 x i8>, <16 x i8>* @scb, align 16 46 %add.i22 = add <16 x i8> %3, %2 47 tail call void (...) @sink(<16 x i8> %add.i22) 48; CHECK: lxvx 34, 0, 3 49; CHECK: lxvx 35, 0, 4 50; CHECK: vaddubm 2, 3, 2 51; CHECK: stxv 34, 52; CHECK: bl sink 53 %4 = load <8 x i16>, <8 x i16>* @usa, align 16 54 %5 = load <8 x i16>, <8 x i16>* @usb, align 16 55 %add.i21 = add <8 x i16> %5, %4 56 tail call void (...) @sink(<8 x i16> %add.i21) 57; CHECK: lxvx 34, 0, 3 58; CHECK: lxvx 35, 0, 4 59; CHECK: vadduhm 2, 3, 2 60; CHECK: stxv 34, 61; CHECK: bl sink 62 %6 = load <8 x i16>, <8 x i16>* @ssa, align 16 63 %7 = load <8 x i16>, <8 x i16>* @ssb, align 16 64 %add.i20 = add <8 x i16> %7, %6 65 tail call void (...) @sink(<8 x i16> %add.i20) 66; CHECK: lxvx 34, 0, 3 67; CHECK: lxvx 35, 0, 4 68; CHECK: vadduhm 2, 3, 2 69; CHECK: stxv 34, 70; CHECK: bl sink 71 %8 = load <4 x i32>, <4 x i32>* @uia, align 16 72 %9 = load <4 x i32>, <4 x i32>* @uib, align 16 73 %add.i19 = add <4 x i32> %9, %8 74 tail call void (...) @sink(<4 x i32> %add.i19) 75; CHECK: lxvx 34, 0, 3 76; CHECK: lxvx 35, 0, 4 77; CHECK: vadduwm 2, 3, 2 78; CHECK: stxv 34, 79; CHECK: bl sink 80 %10 = load <4 x i32>, <4 x i32>* @sia, align 16 81 %11 = load <4 x i32>, <4 x i32>* @sib, align 16 82 %add.i18 = add <4 x i32> %11, %10 83 tail call void (...) @sink(<4 x i32> %add.i18) 84; CHECK: lxvx 34, 0, 3 85; CHECK: lxvx 35, 0, 4 86; CHECK: vadduwm 2, 3, 2 87; CHECK: stxv 34, 88; CHECK: bl sink 89 %12 = load <2 x i64>, <2 x i64>* @ulla, align 16 90 %13 = load <2 x i64>, <2 x i64>* @ullb, align 16 91 %add.i17 = add <2 x i64> %13, %12 92 tail call void (...) @sink(<2 x i64> %add.i17) 93; CHECK: lxvx 34, 0, 3 94; CHECK: lxvx 35, 0, 4 95; CHECK: vaddudm 2, 3, 2 96; CHECK: stxv 34, 97; CHECK: bl sink 98 %14 = load <2 x i64>, <2 x i64>* @slla, align 16 99 %15 = load <2 x i64>, <2 x i64>* @sllb, align 16 100 %add.i16 = add <2 x i64> %15, %14 101 tail call void (...) @sink(<2 x i64> %add.i16) 102; CHECK: lxvx 34, 0, 3 103; CHECK: lxvx 35, 0, 4 104; CHECK: vaddudm 2, 3, 2 105; CHECK: stxv 34, 106; CHECK: bl sink 107 %16 = load <1 x i128>, <1 x i128>* @uxa, align 16 108 %17 = load <1 x i128>, <1 x i128>* @uxb, align 16 109 %add.i15 = add <1 x i128> %17, %16 110 tail call void (...) @sink(<1 x i128> %add.i15) 111; CHECK: lxvx 34, 0, 3 112; CHECK: lxvx 35, 0, 4 113; CHECK: vadduqm 2, 3, 2 114; CHECK: stxv 34, 115; CHECK: bl sink 116 %18 = load <1 x i128>, <1 x i128>* @sxa, align 16 117 %19 = load <1 x i128>, <1 x i128>* @sxb, align 16 118 %add.i14 = add <1 x i128> %19, %18 119 tail call void (...) @sink(<1 x i128> %add.i14) 120; CHECK: lxvx 34, 0, 3 121; CHECK: lxvx 35, 0, 4 122; CHECK: vadduqm 2, 3, 2 123; CHECK: stxv 34, 124; CHECK: bl sink 125 %20 = load <4 x float>, <4 x float>* @vfa, align 16 126 %21 = load <4 x float>, <4 x float>* @vfb, align 16 127 %add.i13 = fadd <4 x float> %20, %21 128 tail call void (...) @sink(<4 x float> %add.i13) 129; CHECK: lxvx 0, 0, 3 130; CHECK: lxvx 1, 0, 4 131; CHECK: xvaddsp 34, 0, 1 132; CHECK: stxv 34, 133; CHECK: bl sink 134 %22 = load <2 x double>, <2 x double>* @vda, align 16 135 %23 = load <2 x double>, <2 x double>* @vdb, align 16 136 %add.i12 = fadd <2 x double> %22, %23 137 tail call void (...) @sink(<2 x double> %add.i12) 138; CHECK: lxvx 0, 0, 3 139; CHECK: lxvx 1, 0, 4 140; CHECK: xvadddp 0, 0, 1 141; CHECK: stxv 0, 142; CHECK: bl sink 143 ret void 144} 145 146; Function Attrs: nounwind readnone 147define <4 x float> @testXVIEXPSP(<4 x i32> %a, <4 x i32> %b) { 148entry: 149 %0 = tail call <4 x float> @llvm.ppc.vsx.xviexpsp(<4 x i32> %a, <4 x i32> %b) 150 ret <4 x float> %0 151; CHECK-LABEL: testXVIEXPSP 152; CHECK: xviexpsp 34, 34, 35 153; CHECK: blr 154} 155; Function Attrs: nounwind readnone 156declare <4 x float> @llvm.ppc.vsx.xviexpsp(<4 x i32>, <4 x i32>) 157 158; Function Attrs: nounwind readnone 159define <2 x double> @testXVIEXPDP(<2 x i64> %a, <2 x i64> %b) { 160entry: 161 %0 = tail call <2 x double> @llvm.ppc.vsx.xviexpdp(<2 x i64> %a, <2 x i64> %b) 162 ret <2 x double> %0 163; CHECK-LABEL: testXVIEXPDP 164; CHECK: xviexpdp 34, 34, 35 165; CHECK: blr 166} 167; Function Attrs: nounwind readnone 168declare <2 x double> @llvm.ppc.vsx.xviexpdp(<2 x i64>, <2 x i64>) 169 170define <16 x i8> @testVSLV(<16 x i8> %a, <16 x i8> %b) { 171entry: 172 %0 = tail call <16 x i8> @llvm.ppc.altivec.vslv(<16 x i8> %a, <16 x i8> %b) 173 ret <16 x i8> %0 174; CHECK-LABEL: testVSLV 175; CHECK: vslv 2, 2, 3 176; CHECK: blr 177} 178; Function Attrs: nounwind readnone 179declare <16 x i8> @llvm.ppc.altivec.vslv(<16 x i8>, <16 x i8>) 180 181; Function Attrs: nounwind readnone 182define <16 x i8> @testVSRV(<16 x i8> %a, <16 x i8> %b) { 183entry: 184 %0 = tail call <16 x i8> @llvm.ppc.altivec.vsrv(<16 x i8> %a, <16 x i8> %b) 185 ret <16 x i8> %0 186; CHECK-LABEL: testVSRV 187; CHECK: vsrv 2, 2, 3 188; CHECK: blr 189} 190; Function Attrs: nounwind readnone 191declare <16 x i8> @llvm.ppc.altivec.vsrv(<16 x i8>, <16 x i8>) 192 193; Function Attrs: nounwind readnone 194define <8 x i16> @testXVCVSPHP(<4 x float> %a) { 195entry: 196; CHECK-LABEL: testXVCVSPHP 197; CHECK: xvcvsphp 34, 34 198; CHECK: blr 199 %0 = tail call <4 x float> @llvm.ppc.vsx.xvcvsphp(<4 x float> %a) 200 %1 = bitcast <4 x float> %0 to <8 x i16> 201 ret <8 x i16> %1 202} 203 204; Function Attrs: nounwind readnone 205define <4 x i32> @testVRLWMI(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { 206entry: 207; CHECK-LABEL: testVRLWMI 208; CHECK: vrlwmi 3, 2, 4 209; CHECK: blr 210 %0 = tail call <4 x i32> @llvm.ppc.altivec.vrlwmi(<4 x i32> %a, <4 x i32> %c, <4 x i32> %b) 211 ret <4 x i32> %0 212} 213 214; Function Attrs: nounwind readnone 215define <2 x i64> @testVRLDMI(<2 x i64> %a, <2 x i64> %b, <2 x i64> %c) { 216entry: 217; CHECK-LABEL: testVRLDMI 218; CHECK: vrldmi 3, 2, 4 219; CHECK: blr 220 %0 = tail call <2 x i64> @llvm.ppc.altivec.vrldmi(<2 x i64> %a, <2 x i64> %c, <2 x i64> %b) 221 ret <2 x i64> %0 222} 223 224; Function Attrs: nounwind readnone 225define <4 x i32> @testVRLWNM(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { 226entry: 227 %0 = tail call <4 x i32> @llvm.ppc.altivec.vrlwnm(<4 x i32> %a, <4 x i32> %b) 228 %and.i = and <4 x i32> %0, %c 229 ret <4 x i32> %and.i 230; CHECK-LABEL: testVRLWNM 231; CHECK: vrlwnm 2, 2, 3 232; CHECK: xxland 34, 34, 36 233; CHECK: blr 234} 235 236; Function Attrs: nounwind readnone 237define <2 x i64> @testVRLDNM(<2 x i64> %a, <2 x i64> %b, <2 x i64> %c) { 238entry: 239 %0 = tail call <2 x i64> @llvm.ppc.altivec.vrldnm(<2 x i64> %a, <2 x i64> %b) 240 %and.i = and <2 x i64> %0, %c 241 ret <2 x i64> %and.i 242; CHECK-LABEL: testVRLDNM 243; CHECK: vrldnm 2, 2, 3 244; CHECK: xxland 34, 34, 36 245; CHECK: blr 246} 247 248; Function Attrs: nounwind readnone 249declare <4 x float> @llvm.ppc.vsx.xvcvsphp(<4 x float>) 250 251; Function Attrs: nounwind readnone 252declare <4 x i32> @llvm.ppc.altivec.vrlwmi(<4 x i32>, <4 x i32>, <4 x i32>) 253 254; Function Attrs: nounwind readnone 255declare <2 x i64> @llvm.ppc.altivec.vrldmi(<2 x i64>, <2 x i64>, <2 x i64>) 256 257; Function Attrs: nounwind readnone 258declare <4 x i32> @llvm.ppc.altivec.vrlwnm(<4 x i32>, <4 x i32>) 259 260; Function Attrs: nounwind readnone 261declare <2 x i64> @llvm.ppc.altivec.vrldnm(<2 x i64>, <2 x i64>) 262 263define <4 x i32> @testXVXEXPSP(<4 x float> %a) { 264entry: 265 %0 = tail call <4 x i32> @llvm.ppc.vsx.xvxexpsp(<4 x float> %a) 266 ret <4 x i32> %0 267; CHECK-LABEL: testXVXEXPSP 268; CHECK: xvxexpsp 34, 34 269; CHECK: blr 270} 271; Function Attrs: nounwind readnone 272declare <4 x i32> @llvm.ppc.vsx.xvxexpsp(<4 x float>) 273 274; Function Attrs: nounwind readnone 275define <2 x i64> @testXVXEXPDP(<2 x double> %a) { 276entry: 277 %0 = tail call <2 x i64> @llvm.ppc.vsx.xvxexpdp(<2 x double> %a) 278 ret <2 x i64> %0 279; CHECK-LABEL: testXVXEXPDP 280; CHECK: xvxexpdp 34, 34 281; CHECK: blr 282} 283; Function Attrs: nounwind readnone 284declare <2 x i64>@llvm.ppc.vsx.xvxexpdp(<2 x double>) 285 286; Function Attrs: nounwind readnone 287define <4 x i32> @testXVXSIGSP(<4 x float> %a) { 288entry: 289 %0 = tail call <4 x i32> @llvm.ppc.vsx.xvxsigsp(<4 x float> %a) 290 ret <4 x i32> %0 291; CHECK-LABEL: testXVXSIGSP 292; CHECK: xvxsigsp 34, 34 293; CHECK: blr 294} 295; Function Attrs: nounwind readnone 296declare <4 x i32> @llvm.ppc.vsx.xvxsigsp(<4 x float>) 297 298; Function Attrs: nounwind readnone 299define <2 x i64> @testXVXSIGDP(<2 x double> %a) { 300entry: 301 %0 = tail call <2 x i64> @llvm.ppc.vsx.xvxsigdp(<2 x double> %a) 302 ret <2 x i64> %0 303; CHECK-LABEL: testXVXSIGDP 304; CHECK: xvxsigdp 34, 34 305; CHECK: blr 306} 307; Function Attrs: nounwind readnone 308declare <2 x i64> @llvm.ppc.vsx.xvxsigdp(<2 x double>) 309 310; Function Attrs: nounwind readnone 311define <4 x i32> @testXVTSTDCSP(<4 x float> %a) { 312entry: 313 %0 = tail call <4 x i32> @llvm.ppc.vsx.xvtstdcsp(<4 x float> %a, i32 127) 314 ret <4 x i32> %0 315; CHECK-LABEL: testXVTSTDCSP 316; CHECK: xvtstdcsp 34, 34, 127 317; CHECK: blr 318} 319; Function Attrs: nounwind readnone 320declare <4 x i32> @llvm.ppc.vsx.xvtstdcsp(<4 x float> %a, i32 %b) 321 322; Function Attrs: nounwind readnone 323define <2 x i64> @testXVTSTDCDP(<2 x double> %a) { 324entry: 325 %0 = tail call <2 x i64> @llvm.ppc.vsx.xvtstdcdp(<2 x double> %a, i32 127) 326 ret <2 x i64> %0 327; CHECK-LABEL: testXVTSTDCDP 328; CHECK: xvtstdcdp 34, 34, 127 329; CHECK: blr 330} 331; Function Attrs: nounwind readnone 332declare <2 x i64> @llvm.ppc.vsx.xvtstdcdp(<2 x double> %a, i32 %b) 333 334define <4 x float> @testXVCVHPSP(<8 x i16> %a) { 335entry: 336 %0 = tail call <4 x float>@llvm.ppc.vsx.xvcvhpsp(<8 x i16> %a) 337 ret <4 x float> %0 338; CHECK-LABEL: testXVCVHPSP 339; CHECK: xvcvhpsp 34, 34 340; CHECK: blr 341} 342; Function Attrs: nounwind readnone 343declare <4 x float>@llvm.ppc.vsx.xvcvhpsp(<8 x i16>) 344 345; Function Attrs: nounwind readnone 346define <4 x i32> @testLXVL(i8* %a, i64 %b) { 347entry: 348 %0 = tail call <4 x i32> @llvm.ppc.vsx.lxvl(i8* %a, i64 %b) 349 ret <4 x i32> %0 350; CHECK-LABEL: testLXVL 351; CHECK: lxvl 34, 3, 4 352; CHECK: blr 353} 354; Function Attrs: nounwind readnone 355declare <4 x i32> @llvm.ppc.vsx.lxvl(i8*, i64) 356 357define void @testSTXVL(<4 x i32> %a, i8* %b, i64 %c) { 358entry: 359 tail call void @llvm.ppc.vsx.stxvl(<4 x i32> %a, i8* %b, i64 %c) 360 ret void 361; CHECK-LABEL: testSTXVL 362; CHECK: stxvl 34, 5, 6 363; CHECK: blr 364} 365; Function Attrs: nounwind readnone 366declare void @llvm.ppc.vsx.stxvl(<4 x i32>, i8*, i64) 367 368; Function Attrs: nounwind readnone 369define <4 x i32> @testLXVLL(i8* %a, i64 %b) { 370entry: 371 %0 = tail call <4 x i32> @llvm.ppc.vsx.lxvll(i8* %a, i64 %b) 372 ret <4 x i32> %0 373; CHECK-LABEL: testLXVLL 374; CHECK: lxvll 34, 3, 4 375; CHECK: blr 376} 377; Function Attrs: nounwind readnone 378declare <4 x i32> @llvm.ppc.vsx.lxvll(i8*, i64) 379 380define void @testSTXVLL(<4 x i32> %a, i8* %b, i64 %c) { 381entry: 382 tail call void @llvm.ppc.vsx.stxvll(<4 x i32> %a, i8* %b, i64 %c) 383 ret void 384; CHECK-LABEL: testSTXVLL 385; CHECK: stxvll 34, 5, 6 386; CHECK: blr 387} 388; Function Attrs: nounwind readnone 389declare void @llvm.ppc.vsx.stxvll(<4 x i32>, i8*, i64) 390 391define <4 x i32> @test0(<4 x i32> %a) local_unnamed_addr #0 { 392entry: 393 %sub.i = sub <4 x i32> zeroinitializer, %a 394 ret <4 x i32> %sub.i 395 396; CHECK-LABEL: @test0 397; CHECK: vnegw 2, 2 398; CHECK: blr 399 400} 401 402define <2 x i64> @test1(<2 x i64> %a) local_unnamed_addr #0 { 403entry: 404 %sub.i = sub <2 x i64> zeroinitializer, %a 405 ret <2 x i64> %sub.i 406 407; CHECK-LABEL: @test1 408; CHECK: vnegd 2, 2 409; CHECK: blr 410 411} 412 413declare void @sink(...) 414 415; stack object should be accessed using D-form load/store instead of X-form 416define signext i32 @func1() { 417; CHECK-LABEL: @func1 418; CHECK-NOT: stxvx 419; CHECK: stxv {{[0-9]+}}, {{[0-9]+}}(1) 420; CHECK-NOT: stxvx 421; CHECK: blr 422entry: 423 %a = alloca [4 x i32], align 4 424 %0 = bitcast [4 x i32]* %a to i8* 425 call void @llvm.memset.p0i8.i64(i8* nonnull align 4 %0, i8 0, i64 16, i1 false) 426 %arraydecay = getelementptr inbounds [4 x i32], [4 x i32]* %a, i64 0, i64 0 427 %call = call signext i32 @callee(i32* nonnull %arraydecay) #3 428 ret i32 %call 429} 430 431; stack object should be accessed using D-form load/store instead of X-form 432define signext i32 @func2() { 433; CHECK-LABEL: @func2 434; CHECK-NOT: stxvx 435; CHECK: stxv [[ZEROREG:[0-9]+]], {{[0-9]+}}(1) 436; CHECK: stxv [[ZEROREG]], {{[0-9]+}}(1) 437; CHECK: stxv [[ZEROREG]], {{[0-9]+}}(1) 438; CHECK: stxv [[ZEROREG]], {{[0-9]+}}(1) 439; CHECK-NOT: stxvx 440; CHECK: blr 441entry: 442 %a = alloca [16 x i32], align 4 443 %0 = bitcast [16 x i32]* %a to i8* 444 call void @llvm.memset.p0i8.i64(i8* nonnull align 4 %0, i8 0, i64 64, i1 false) 445 %arraydecay = getelementptr inbounds [16 x i32], [16 x i32]* %a, i64 0, i64 0 446 %call = call signext i32 @callee(i32* nonnull %arraydecay) #3 447 ret i32 %call 448} 449 450declare void @llvm.memset.p0i8.i64(i8* nocapture writeonly, i8, i64, i1) #1 451declare signext i32 @callee(i32*) local_unnamed_addr #2 452