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
4 define <8 x float> @sitofp00(<8 x i32> %a) nounwind {
5 ; CHECK-LABEL: sitofp00:
7 ; CHECK-NEXT: vcvtdq2ps %ymm0, %ymm0
9 %b = sitofp <8 x i32> %a to <8 x float>
13 define <8 x i32> @fptosi00(<8 x float> %a) nounwind {
14 ; CHECK-LABEL: fptosi00:
16 ; CHECK-NEXT: vcvttps2dq %ymm0, %ymm0
18 %b = fptosi <8 x float> %a to <8 x i32>
22 define <4 x double> @sitofp01(<4 x i32> %a) {
23 ; CHECK-LABEL: sitofp01:
25 ; CHECK-NEXT: vcvtdq2pd %xmm0, %ymm0
27 %b = sitofp <4 x i32> %a to <4 x double>
31 define <8 x float> @sitofp02(<8 x i16> %a) {
32 ; CHECK-LABEL: sitofp02:
34 ; CHECK-NEXT: vpmovsxwd %xmm0, %xmm1
35 ; CHECK-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
36 ; CHECK-NEXT: vpmovsxwd %xmm0, %xmm0
37 ; CHECK-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
38 ; CHECK-NEXT: vcvtdq2ps %ymm0, %ymm0
40 %b = sitofp <8 x i16> %a to <8 x float>
44 define <4 x i32> @fptosi01(<4 x double> %a) {
45 ; CHECK-LABEL: fptosi01:
47 ; CHECK-NEXT: vcvttpd2dq %ymm0, %xmm0
48 ; CHECK-NEXT: vzeroupper
50 %b = fptosi <4 x double> %a to <4 x i32>
54 define <8 x float> @fptrunc00(<8 x double> %b) nounwind {
55 ; CHECK-LABEL: fptrunc00:
57 ; CHECK-NEXT: vcvtpd2ps %ymm0, %xmm0
58 ; CHECK-NEXT: vcvtpd2ps %ymm1, %xmm1
59 ; CHECK-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
61 %a = fptrunc <8 x double> %b to <8 x float>
65 define <4 x float> @fptrunc01(<2 x double> %a0, <4 x float> %a1) nounwind {
66 ; CHECK-LABEL: fptrunc01:
68 ; CHECK-NEXT: vcvtsd2ss %xmm0, %xmm1, %xmm0
70 %ext = extractelement <2 x double> %a0, i32 0
71 %cvt = fptrunc double %ext to float
72 %res = insertelement <4 x float> %a1, float %cvt, i32 0
76 define <4 x double> @fpext00(<4 x float> %b) nounwind {
77 ; CHECK-LABEL: fpext00:
79 ; CHECK-NEXT: vcvtps2pd %xmm0, %ymm0
81 %a = fpext <4 x float> %b to <4 x double>
85 define <2 x double> @fpext01(<2 x double> %a0, <4 x float> %a1) nounwind {
86 ; CHECK-LABEL: fpext01:
88 ; CHECK-NEXT: vcvtss2sd %xmm1, %xmm0, %xmm0
90 %ext = extractelement <4 x float> %a1, i32 0
91 %cvt = fpext float %ext to double
92 %res = insertelement <2 x double> %a0, double %cvt, i32 0
96 define double @funcA(i64* nocapture %e) nounwind uwtable readonly ssp {
99 ; CHECK-NEXT: vcvtsi2sdq (%rdi), %xmm0, %xmm0
101 %tmp1 = load i64, i64* %e, align 8
102 %conv = sitofp i64 %tmp1 to double
106 define double @funcB(i32* nocapture %e) nounwind uwtable readonly ssp {
107 ; CHECK-LABEL: funcB:
109 ; CHECK-NEXT: vcvtsi2sdl (%rdi), %xmm0, %xmm0
111 %tmp1 = load i32, i32* %e, align 4
112 %conv = sitofp i32 %tmp1 to double
116 define float @funcC(i32* nocapture %e) nounwind uwtable readonly ssp {
117 ; CHECK-LABEL: funcC:
119 ; CHECK-NEXT: vcvtsi2ssl (%rdi), %xmm0, %xmm0
121 %tmp1 = load i32, i32* %e, align 4
122 %conv = sitofp i32 %tmp1 to float
126 define float @funcD(i64* nocapture %e) nounwind uwtable readonly ssp {
127 ; CHECK-LABEL: funcD:
129 ; CHECK-NEXT: vcvtsi2ssq (%rdi), %xmm0, %xmm0
131 %tmp1 = load i64, i64* %e, align 8
132 %conv = sitofp i64 %tmp1 to float
136 define void @fpext() nounwind uwtable {
137 ; CHECK-LABEL: fpext:
139 ; CHECK-NEXT: vcvtss2sd -{{[0-9]+}}(%rsp), %xmm0, %xmm0
140 ; CHECK-NEXT: vmovsd %xmm0, -{{[0-9]+}}(%rsp)
142 %f = alloca float, align 4
143 %d = alloca double, align 8
144 %tmp = load float, float* %f, align 4
145 %conv = fpext float %tmp to double
146 store double %conv, double* %d, align 8
150 define double @nearbyint_f64(double %a) {
151 ; CHECK-LABEL: nearbyint_f64:
153 ; CHECK-NEXT: vroundsd $12, %xmm0, %xmm0, %xmm0
155 %res = call double @llvm.nearbyint.f64(double %a)
158 declare double @llvm.nearbyint.f64(double %p)
160 define float @floor_f32(float %a) {
161 ; CHECK-LABEL: floor_f32:
163 ; CHECK-NEXT: vroundss $9, %xmm0, %xmm0, %xmm0
165 %res = call float @llvm.floor.f32(float %a)
168 declare float @llvm.floor.f32(float %p)