1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py 2; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx | FileCheck %s 3 4define <8 x float> @A(<8 x float> %a) nounwind uwtable readnone ssp { 5; CHECK-LABEL: A: 6; CHECK: # %bb.0: 7; CHECK-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0 8; CHECK-NEXT: retq 9 %shuffle = shufflevector <8 x float> %a, <8 x float> undef, <8 x i32> <i32 8, i32 8, i32 8, i32 8, i32 0, i32 1, i32 2, i32 3> 10 ret <8 x float> %shuffle 11} 12 13define <4 x double> @B(<4 x double> %a) nounwind uwtable readnone ssp { 14; CHECK-LABEL: B: 15; CHECK: # %bb.0: 16; CHECK-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0 17; CHECK-NEXT: retq 18 %shuffle = shufflevector <4 x double> %a, <4 x double> undef, <4 x i32> <i32 4, i32 4, i32 0, i32 1> 19 ret <4 x double> %shuffle 20} 21 22declare <2 x double> @llvm.x86.sse2.min.pd(<2 x double>, <2 x double>) nounwind readnone 23declare <2 x double> @llvm.x86.sse2.min.sd(<2 x double>, <2 x double>) nounwind readnone 24 25define void @insert_crash() nounwind { 26; CHECK-LABEL: insert_crash: 27; CHECK: # %bb.0: # %allocas 28; CHECK-NEXT: vxorpd %xmm0, %xmm0, %xmm0 29; CHECK-NEXT: vminpd %xmm0, %xmm0, %xmm0 30; CHECK-NEXT: vminsd %xmm0, %xmm0, %xmm0 31; CHECK-NEXT: vcvtsd2ss %xmm0, %xmm0, %xmm0 32; CHECK-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[0,1,2,0] 33; CHECK-NEXT: vmovups %xmm0, (%rax) 34; CHECK-NEXT: retq 35allocas: 36 %v1.i.i451 = shufflevector <4 x double> zeroinitializer, <4 x double> undef, <4 x i32> <i32 2, i32 3, i32 undef, i32 undef> 37 %ret_0a.i.i.i452 = shufflevector <4 x double> %v1.i.i451, <4 x double> undef, <2 x i32> <i32 0, i32 1> 38 %vret_0.i.i.i454 = tail call <2 x double> @llvm.x86.sse2.min.pd(<2 x double> %ret_0a.i.i.i452, <2 x double> undef) nounwind 39 %ret_val.i.i.i463 = tail call <2 x double> @llvm.x86.sse2.min.sd(<2 x double> %vret_0.i.i.i454, <2 x double> undef) nounwind 40 %ret.i1.i.i464 = extractelement <2 x double> %ret_val.i.i.i463, i32 0 41 %double2float = fptrunc double %ret.i1.i.i464 to float 42 %smearinsert50 = insertelement <4 x float> undef, float %double2float, i32 3 43 %blendAsInt.i503 = bitcast <4 x float> %smearinsert50 to <4 x i32> 44 store <4 x i32> %blendAsInt.i503, <4 x i32>* undef, align 4 45 ret void 46} 47 48;; DAG Combine must remove useless vinsertf128 instructions 49 50define <4 x i32> @DAGCombineA(<4 x i32> %v1) nounwind readonly { 51; CHECK-LABEL: DAGCombineA: 52; CHECK: # %bb.0: 53; CHECK-NEXT: retq 54 %t1 = shufflevector <4 x i32> %v1, <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7> 55 %t2 = shufflevector <8 x i32> %t1, <8 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3> 56 ret <4 x i32> %t2 57} 58 59define <8 x i32> @DAGCombineB(<8 x i32> %v1, <8 x i32> %v2) nounwind readonly { 60; CHECK-LABEL: DAGCombineB: 61; CHECK: # %bb.0: 62; CHECK-NEXT: vextractf128 $1, %ymm1, %xmm2 63; CHECK-NEXT: vextractf128 $1, %ymm0, %xmm3 64; CHECK-NEXT: vpaddd %xmm3, %xmm2, %xmm2 65; CHECK-NEXT: vpaddd %xmm2, %xmm3, %xmm2 66; CHECK-NEXT: vpaddd %xmm0, %xmm1, %xmm1 67; CHECK-NEXT: vpaddd %xmm1, %xmm0, %xmm0 68; CHECK-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 69; CHECK-NEXT: retq 70 %t1 = add <8 x i32> %v1, %v2 71 %t2 = add <8 x i32> %t1, %v1 72 ret <8 x i32> %t2 73} 74 75define <4 x double> @insert_undef_pd(<4 x double> %a0, <2 x double> %a1) { 76; CHECK-LABEL: insert_undef_pd: 77; CHECK: # %bb.0: 78; CHECK-NEXT: # kill: def $xmm1 killed $xmm1 def $ymm1 79; CHECK-NEXT: vmovaps %ymm1, %ymm0 80; CHECK-NEXT: retq 81%res = call <4 x double> @llvm.x86.avx.vinsertf128.pd.256(<4 x double> undef, <2 x double> %a1, i8 0) 82ret <4 x double> %res 83} 84declare <4 x double> @llvm.x86.avx.vinsertf128.pd.256(<4 x double>, <2 x double>, i8) nounwind readnone 85 86define <8 x float> @insert_undef_ps(<8 x float> %a0, <4 x float> %a1) { 87; CHECK-LABEL: insert_undef_ps: 88; CHECK: # %bb.0: 89; CHECK-NEXT: # kill: def $xmm1 killed $xmm1 def $ymm1 90; CHECK-NEXT: vmovaps %ymm1, %ymm0 91; CHECK-NEXT: retq 92%res = call <8 x float> @llvm.x86.avx.vinsertf128.ps.256(<8 x float> undef, <4 x float> %a1, i8 0) 93ret <8 x float> %res 94} 95declare <8 x float> @llvm.x86.avx.vinsertf128.ps.256(<8 x float>, <4 x float>, i8) nounwind readnone 96 97define <8 x i32> @insert_undef_si(<8 x i32> %a0, <4 x i32> %a1) { 98; CHECK-LABEL: insert_undef_si: 99; CHECK: # %bb.0: 100; CHECK-NEXT: # kill: def $xmm1 killed $xmm1 def $ymm1 101; CHECK-NEXT: vmovaps %ymm1, %ymm0 102; CHECK-NEXT: retq 103%res = call <8 x i32> @llvm.x86.avx.vinsertf128.si.256(<8 x i32> undef, <4 x i32> %a1, i8 0) 104ret <8 x i32> %res 105} 106declare <8 x i32> @llvm.x86.avx.vinsertf128.si.256(<8 x i32>, <4 x i32>, i8) nounwind readnone 107 108; rdar://10643481 109define <8 x float> @vinsertf128_combine(float* nocapture %f) nounwind uwtable readonly ssp { 110; CHECK-LABEL: vinsertf128_combine: 111; CHECK: # %bb.0: 112; CHECK-NEXT: vinsertf128 $1, 16(%rdi), %ymm0, %ymm0 113; CHECK-NEXT: retq 114 %add.ptr = getelementptr inbounds float, float* %f, i64 4 115 %t0 = bitcast float* %add.ptr to <4 x float>* 116 %t1 = load <4 x float>, <4 x float>* %t0, align 16 117 %t2 = tail call <8 x float> @llvm.x86.avx.vinsertf128.ps.256(<8 x float> undef, <4 x float> %t1, i8 1) 118 ret <8 x float> %t2 119} 120 121; rdar://11076953 122define <8 x float> @vinsertf128_ucombine(float* nocapture %f) nounwind uwtable readonly ssp { 123; CHECK-LABEL: vinsertf128_ucombine: 124; CHECK: # %bb.0: 125; CHECK-NEXT: vinsertf128 $1, 16(%rdi), %ymm0, %ymm0 126; CHECK-NEXT: retq 127 %add.ptr = getelementptr inbounds float, float* %f, i64 4 128 %t0 = bitcast float* %add.ptr to <4 x float>* 129 %t1 = load <4 x float>, <4 x float>* %t0, align 8 130 %t2 = tail call <8 x float> @llvm.x86.avx.vinsertf128.ps.256(<8 x float> undef, <4 x float> %t1, i8 1) 131 ret <8 x float> %t2 132} 133 134