1 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2 ; RUN: opt < %s -instcombine -S | FileCheck %s
3 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
5 define <4 x float> @test_vfmadd_ss(<4 x float> %a, <4 x float> %b, <4 x float> %c) {
6 ; CHECK-LABEL: @test_vfmadd_ss(
7 ; CHECK-NEXT: [[TMP1:%.*]] = extractelement <4 x float> [[A:%.*]], i64 0
8 ; CHECK-NEXT: [[TMP2:%.*]] = extractelement <4 x float> [[B:%.*]], i32 0
9 ; CHECK-NEXT: [[TMP3:%.*]] = extractelement <4 x float> [[C:%.*]], i32 0
10 ; CHECK-NEXT: [[TMP4:%.*]] = call float @llvm.fma.f32(float [[TMP1]], float [[TMP2]], float [[TMP3]])
11 ; CHECK-NEXT: [[TMP5:%.*]] = insertelement <4 x float> [[A]], float [[TMP4]], i64 0
12 ; CHECK-NEXT: ret <4 x float> [[TMP5]]
14 %1 = insertelement <4 x float> %b, float 1.000000e+00, i32 1
15 %2 = insertelement <4 x float> %1, float 2.000000e+00, i32 2
16 %3 = insertelement <4 x float> %2, float 3.000000e+00, i32 3
17 %4 = insertelement <4 x float> %c, float 4.000000e+00, i32 1
18 %5 = insertelement <4 x float> %4, float 5.000000e+00, i32 2
19 %6 = insertelement <4 x float> %5, float 6.000000e+00, i32 3
20 %7 = extractelement <4 x float> %a, i64 0
21 %8 = extractelement <4 x float> %3, i64 0
22 %9 = extractelement <4 x float> %6, i64 0
23 %10 = call float @llvm.fma.f32(float %7, float %8, float %9)
24 %11 = insertelement <4 x float> %a, float %10, i64 0
28 define float @test_vfmadd_ss_0(<4 x float> %a, <4 x float> %b, <4 x float> %c) {
29 ; CHECK-LABEL: @test_vfmadd_ss_0(
30 ; CHECK-NEXT: [[TMP1:%.*]] = extractelement <4 x float> [[A:%.*]], i32 0
31 ; CHECK-NEXT: [[TMP2:%.*]] = extractelement <4 x float> [[B:%.*]], i64 0
32 ; CHECK-NEXT: [[TMP3:%.*]] = extractelement <4 x float> [[C:%.*]], i64 0
33 ; CHECK-NEXT: [[TMP4:%.*]] = call float @llvm.fma.f32(float [[TMP1]], float [[TMP2]], float [[TMP3]])
34 ; CHECK-NEXT: ret float [[TMP4]]
36 %1 = insertelement <4 x float> %a, float 1.000000e+00, i32 1
37 %2 = insertelement <4 x float> %1, float 2.000000e+00, i32 2
38 %3 = insertelement <4 x float> %2, float 3.000000e+00, i32 3
39 %4 = extractelement <4 x float> %3, i64 0
40 %5 = extractelement <4 x float> %b, i64 0
41 %6 = extractelement <4 x float> %c, i64 0
42 %7 = call float @llvm.fma.f32(float %4, float %5, float %6)
43 %8 = insertelement <4 x float> %3, float %7, i64 0
44 %9 = extractelement <4 x float> %8, i32 0
48 define float @test_vfmadd_ss_1(<4 x float> %a, <4 x float> %b, <4 x float> %c) {
49 ; CHECK-LABEL: @test_vfmadd_ss_1(
50 ; CHECK-NEXT: ret float 1.000000e+00
52 %1 = insertelement <4 x float> %a, float 1.000000e+00, i32 1
53 %2 = insertelement <4 x float> %1, float 2.000000e+00, i32 2
54 %3 = insertelement <4 x float> %2, float 3.000000e+00, i32 3
55 %4 = extractelement <4 x float> %3, i64 0
56 %5 = extractelement <4 x float> %b, i64 0
57 %6 = extractelement <4 x float> %c, i64 0
58 %7 = call float @llvm.fma.f32(float %4, float %5, float %6)
59 %8 = insertelement <4 x float> %3, float %7, i64 0
60 %9 = extractelement <4 x float> %8, i32 1
64 define <2 x double> @test_vfmadd_sd(<2 x double> %a, <2 x double> %b, <2 x double> %c) {
65 ; CHECK-LABEL: @test_vfmadd_sd(
66 ; CHECK-NEXT: [[TMP1:%.*]] = extractelement <2 x double> [[A:%.*]], i64 0
67 ; CHECK-NEXT: [[TMP2:%.*]] = extractelement <2 x double> [[B:%.*]], i64 0
68 ; CHECK-NEXT: [[TMP3:%.*]] = extractelement <2 x double> [[C:%.*]], i64 0
69 ; CHECK-NEXT: [[TMP4:%.*]] = call double @llvm.fma.f64(double [[TMP1]], double [[TMP2]], double [[TMP3]])
70 ; CHECK-NEXT: [[TMP5:%.*]] = insertelement <2 x double> [[A]], double [[TMP4]], i64 0
71 ; CHECK-NEXT: ret <2 x double> [[TMP5]]
73 %1 = insertelement <2 x double> %b, double 1.000000e+00, i32 1
74 %2 = insertelement <2 x double> %c, double 2.000000e+00, i32 1
75 %3 = extractelement <2 x double> %a, i64 0
76 %4 = extractelement <2 x double> %1, i64 0
77 %5 = extractelement <2 x double> %2, i64 0
78 %6 = call double @llvm.fma.f64(double %3, double %4, double %5)
79 %7 = insertelement <2 x double> %a, double %6, i64 0
83 define double @test_vfmadd_sd_0(<2 x double> %a, <2 x double> %b, <2 x double> %c) {
84 ; CHECK-LABEL: @test_vfmadd_sd_0(
85 ; CHECK-NEXT: [[TMP1:%.*]] = extractelement <2 x double> [[A:%.*]], i64 0
86 ; CHECK-NEXT: [[TMP2:%.*]] = extractelement <2 x double> [[B:%.*]], i64 0
87 ; CHECK-NEXT: [[TMP3:%.*]] = extractelement <2 x double> [[C:%.*]], i64 0
88 ; CHECK-NEXT: [[TMP4:%.*]] = call double @llvm.fma.f64(double [[TMP1]], double [[TMP2]], double [[TMP3]])
89 ; CHECK-NEXT: ret double [[TMP4]]
91 %1 = insertelement <2 x double> %a, double 1.000000e+00, i32 1
92 %2 = extractelement <2 x double> %1, i64 0
93 %3 = extractelement <2 x double> %b, i64 0
94 %4 = extractelement <2 x double> %c, i64 0
95 %5 = call double @llvm.fma.f64(double %2, double %3, double %4)
96 %6 = insertelement <2 x double> %1, double %5, i64 0
97 %7 = extractelement <2 x double> %6, i32 0
101 define double @test_vfmadd_sd_1(<2 x double> %a, <2 x double> %b, <2 x double> %c) {
102 ; CHECK-LABEL: @test_vfmadd_sd_1(
103 ; CHECK-NEXT: ret double 1.000000e+00
105 %1 = insertelement <2 x double> %a, double 1.000000e+00, i32 1
106 %2 = extractelement <2 x double> %1, i64 0
107 %3 = extractelement <2 x double> %b, i64 0
108 %4 = extractelement <2 x double> %c, i64 0
109 %5 = call double @llvm.fma.f64(double %2, double %3, double %4)
110 %6 = insertelement <2 x double> %1, double %5, i64 0
111 %7 = extractelement <2 x double> %6, i32 1
115 declare float @llvm.fma.f32(float, float, float)
116 declare double @llvm.fma.f64(double, double, double)