1; RUN: opt < %s -basicaa -slp-vectorizer -slp-threshold=-999 -dce -S -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s 2 3target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128" 4target triple = "x86_64-apple-macosx10.8.0" 5 6declare double @sin(double) 7declare double @cos(double) 8declare double @pow(double, double) 9declare double @exp2(double) 10declare i64 @round(i64) 11 12 13; CHECK: sin_libm 14; CHECK: call <2 x double> @llvm.sin.v2f64 15; CHECK: ret void 16define void @sin_libm(double* %a, double* %b, double* %c) { 17entry: 18 %i0 = load double, double* %a, align 8 19 %i1 = load double, double* %b, align 8 20 %mul = fmul double %i0, %i1 21 %call = tail call double @sin(double %mul) nounwind readnone 22 %arrayidx3 = getelementptr inbounds double, double* %a, i64 1 23 %i3 = load double, double* %arrayidx3, align 8 24 %arrayidx4 = getelementptr inbounds double, double* %b, i64 1 25 %i4 = load double, double* %arrayidx4, align 8 26 %mul5 = fmul double %i3, %i4 27 %call5 = tail call double @sin(double %mul5) nounwind readnone 28 store double %call, double* %c, align 8 29 %arrayidx5 = getelementptr inbounds double, double* %c, i64 1 30 store double %call5, double* %arrayidx5, align 8 31 ret void 32} 33 34; CHECK: cos_libm 35; CHECK: call <2 x double> @llvm.cos.v2f64 36; CHECK: ret void 37define void @cos_libm(double* %a, double* %b, double* %c) { 38entry: 39 %i0 = load double, double* %a, align 8 40 %i1 = load double, double* %b, align 8 41 %mul = fmul double %i0, %i1 42 %call = tail call double @cos(double %mul) nounwind readnone 43 %arrayidx3 = getelementptr inbounds double, double* %a, i64 1 44 %i3 = load double, double* %arrayidx3, align 8 45 %arrayidx4 = getelementptr inbounds double, double* %b, i64 1 46 %i4 = load double, double* %arrayidx4, align 8 47 %mul5 = fmul double %i3, %i4 48 %call5 = tail call double @cos(double %mul5) nounwind readnone 49 store double %call, double* %c, align 8 50 %arrayidx5 = getelementptr inbounds double, double* %c, i64 1 51 store double %call5, double* %arrayidx5, align 8 52 ret void 53} 54 55; CHECK: pow_libm 56; CHECK: call <2 x double> @llvm.pow.v2f64 57; CHECK: ret void 58define void @pow_libm(double* %a, double* %b, double* %c) { 59entry: 60 %i0 = load double, double* %a, align 8 61 %i1 = load double, double* %b, align 8 62 %mul = fmul double %i0, %i1 63 %call = tail call double @pow(double %mul,double %mul) nounwind readnone 64 %arrayidx3 = getelementptr inbounds double, double* %a, i64 1 65 %i3 = load double, double* %arrayidx3, align 8 66 %arrayidx4 = getelementptr inbounds double, double* %b, i64 1 67 %i4 = load double, double* %arrayidx4, align 8 68 %mul5 = fmul double %i3, %i4 69 %call5 = tail call double @pow(double %mul5,double %mul5) nounwind readnone 70 store double %call, double* %c, align 8 71 %arrayidx5 = getelementptr inbounds double, double* %c, i64 1 72 store double %call5, double* %arrayidx5, align 8 73 ret void 74} 75 76 77; CHECK: exp2_libm 78; CHECK: call <2 x double> @llvm.exp2.v2f64 79; CHECK: ret void 80define void @exp2_libm(double* %a, double* %b, double* %c) { 81entry: 82 %i0 = load double, double* %a, align 8 83 %i1 = load double, double* %b, align 8 84 %mul = fmul double %i0, %i1 85 %call = tail call double @exp2(double %mul) nounwind readnone 86 %arrayidx3 = getelementptr inbounds double, double* %a, i64 1 87 %i3 = load double, double* %arrayidx3, align 8 88 %arrayidx4 = getelementptr inbounds double, double* %b, i64 1 89 %i4 = load double, double* %arrayidx4, align 8 90 %mul5 = fmul double %i3, %i4 91 %call5 = tail call double @exp2(double %mul5) nounwind readnone 92 store double %call, double* %c, align 8 93 %arrayidx5 = getelementptr inbounds double, double* %c, i64 1 94 store double %call5, double* %arrayidx5, align 8 95 ret void 96} 97 98 99; Negative test case 100; CHECK: round_custom 101; CHECK-NOT: load <4 x i64> 102; CHECK: ret void 103define void @round_custom(i64* %a, i64* %b, i64* %c) { 104entry: 105 %i0 = load i64, i64* %a, align 8 106 %i1 = load i64, i64* %b, align 8 107 %mul = mul i64 %i0, %i1 108 %call = tail call i64 @round(i64 %mul) nounwind readnone 109 %arrayidx3 = getelementptr inbounds i64, i64* %a, i64 1 110 %i3 = load i64, i64* %arrayidx3, align 8 111 %arrayidx4 = getelementptr inbounds i64, i64* %b, i64 1 112 %i4 = load i64, i64* %arrayidx4, align 8 113 %mul5 = mul i64 %i3, %i4 114 %call5 = tail call i64 @round(i64 %mul5) nounwind readnone 115 store i64 %call, i64* %c, align 8 116 %arrayidx5 = getelementptr inbounds i64, i64* %c, i64 1 117 store i64 %call5, i64* %arrayidx5, align 8 118 ret void 119} 120 121 122; CHECK: declare <2 x double> @llvm.sin.v2f64(<2 x double>) [[ATTR0:#[0-9]+]] 123; CHECK: declare <2 x double> @llvm.cos.v2f64(<2 x double>) [[ATTR0]] 124; CHECK: declare <2 x double> @llvm.pow.v2f64(<2 x double>, <2 x double>) [[ATTR0]] 125; CHECK: declare <2 x double> @llvm.exp2.v2f64(<2 x double>) [[ATTR0]] 126 127; CHECK: attributes [[ATTR0]] = { nounwind readnone } 128 129