1 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2 ; RUN: opt < %s -mtriple=x86_64-unknown -passes=slp-vectorizer,instcombine -S | FileCheck %s --check-prefix=CHECK --check-prefix=SSE
3 ; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=slm -passes=slp-vectorizer,instcombine -S | FileCheck %s --check-prefix=CHECK --check-prefix=SLM
4 ; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=corei7-avx -passes=slp-vectorizer,instcombine -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX
5 ; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=core-avx2 -passes=slp-vectorizer,instcombine -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX
6 ; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=knl -passes=slp-vectorizer,instcombine -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX512
7 ; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=skx -passes=slp-vectorizer,instcombine -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX512
9 define <8 x float> @fadd_fsub_v8f32(<8 x float> %a, <8 x float> %b) {
10 ; CHECK-LABEL: @fadd_fsub_v8f32(
11 ; CHECK-NEXT: [[TMP1:%.*]] = fadd <8 x float> [[A:%.*]], [[B:%.*]]
12 ; CHECK-NEXT: [[TMP2:%.*]] = fsub <8 x float> [[A]], [[B]]
13 ; CHECK-NEXT: [[TMP3:%.*]] = shufflevector <8 x float> [[TMP1]], <8 x float> [[TMP2]], <8 x i32> <i32 0, i32 9, i32 10, i32 3, i32 4, i32 13, i32 14, i32 7>
14 ; CHECK-NEXT: ret <8 x float> [[TMP3]]
16 %a0 = extractelement <8 x float> %a, i32 0
17 %a1 = extractelement <8 x float> %a, i32 1
18 %a2 = extractelement <8 x float> %a, i32 2
19 %a3 = extractelement <8 x float> %a, i32 3
20 %a4 = extractelement <8 x float> %a, i32 4
21 %a5 = extractelement <8 x float> %a, i32 5
22 %a6 = extractelement <8 x float> %a, i32 6
23 %a7 = extractelement <8 x float> %a, i32 7
24 %b0 = extractelement <8 x float> %b, i32 0
25 %b1 = extractelement <8 x float> %b, i32 1
26 %b2 = extractelement <8 x float> %b, i32 2
27 %b3 = extractelement <8 x float> %b, i32 3
28 %b4 = extractelement <8 x float> %b, i32 4
29 %b5 = extractelement <8 x float> %b, i32 5
30 %b6 = extractelement <8 x float> %b, i32 6
31 %b7 = extractelement <8 x float> %b, i32 7
32 %ab0 = fadd float %a0, %b0
33 %ab1 = fsub float %a1, %b1
34 %ab2 = fsub float %a2, %b2
35 %ab3 = fadd float %a3, %b3
36 %ab4 = fadd float %a4, %b4
37 %ab5 = fsub float %a5, %b5
38 %ab6 = fsub float %a6, %b6
39 %ab7 = fadd float %a7, %b7
40 %r0 = insertelement <8 x float> undef, float %ab0, i32 0
41 %r1 = insertelement <8 x float> %r0, float %ab1, i32 1
42 %r2 = insertelement <8 x float> %r1, float %ab2, i32 2
43 %r3 = insertelement <8 x float> %r2, float %ab3, i32 3
44 %r4 = insertelement <8 x float> %r3, float %ab4, i32 4
45 %r5 = insertelement <8 x float> %r4, float %ab5, i32 5
46 %r6 = insertelement <8 x float> %r5, float %ab6, i32 6
47 %r7 = insertelement <8 x float> %r6, float %ab7, i32 7
51 define <8 x float> @fmul_fdiv_v8f32(<8 x float> %a, <8 x float> %b) {
52 ; CHECK-LABEL: @fmul_fdiv_v8f32(
53 ; CHECK-NEXT: [[TMP1:%.*]] = fmul <8 x float> [[A:%.*]], [[B:%.*]]
54 ; CHECK-NEXT: [[TMP2:%.*]] = fdiv <8 x float> [[A]], [[B]]
55 ; CHECK-NEXT: [[TMP3:%.*]] = shufflevector <8 x float> [[TMP1]], <8 x float> [[TMP2]], <8 x i32> <i32 0, i32 9, i32 10, i32 3, i32 4, i32 13, i32 14, i32 7>
56 ; CHECK-NEXT: ret <8 x float> [[TMP3]]
58 %a0 = extractelement <8 x float> %a, i32 0
59 %a1 = extractelement <8 x float> %a, i32 1
60 %a2 = extractelement <8 x float> %a, i32 2
61 %a3 = extractelement <8 x float> %a, i32 3
62 %a4 = extractelement <8 x float> %a, i32 4
63 %a5 = extractelement <8 x float> %a, i32 5
64 %a6 = extractelement <8 x float> %a, i32 6
65 %a7 = extractelement <8 x float> %a, i32 7
66 %b0 = extractelement <8 x float> %b, i32 0
67 %b1 = extractelement <8 x float> %b, i32 1
68 %b2 = extractelement <8 x float> %b, i32 2
69 %b3 = extractelement <8 x float> %b, i32 3
70 %b4 = extractelement <8 x float> %b, i32 4
71 %b5 = extractelement <8 x float> %b, i32 5
72 %b6 = extractelement <8 x float> %b, i32 6
73 %b7 = extractelement <8 x float> %b, i32 7
74 %ab0 = fmul float %a0, %b0
75 %ab1 = fdiv float %a1, %b1
76 %ab2 = fdiv float %a2, %b2
77 %ab3 = fmul float %a3, %b3
78 %ab4 = fmul float %a4, %b4
79 %ab5 = fdiv float %a5, %b5
80 %ab6 = fdiv float %a6, %b6
81 %ab7 = fmul float %a7, %b7
82 %r0 = insertelement <8 x float> undef, float %ab0, i32 0
83 %r1 = insertelement <8 x float> %r0, float %ab1, i32 1
84 %r2 = insertelement <8 x float> %r1, float %ab2, i32 2
85 %r3 = insertelement <8 x float> %r2, float %ab3, i32 3
86 %r4 = insertelement <8 x float> %r3, float %ab4, i32 4
87 %r5 = insertelement <8 x float> %r4, float %ab5, i32 5
88 %r6 = insertelement <8 x float> %r5, float %ab6, i32 6
89 %r7 = insertelement <8 x float> %r6, float %ab7, i32 7
93 define <4 x float> @fmul_fdiv_v4f32_const(<4 x float> %a) {
94 ; SSE-LABEL: @fmul_fdiv_v4f32_const(
95 ; SSE-NEXT: [[TMP1:%.*]] = fmul <4 x float> [[A:%.*]], <float 2.000000e+00, float 1.000000e+00, float 1.000000e+00, float 2.000000e+00>
96 ; SSE-NEXT: ret <4 x float> [[TMP1]]
98 ; SLM-LABEL: @fmul_fdiv_v4f32_const(
99 ; SLM-NEXT: [[A2:%.*]] = extractelement <4 x float> [[A:%.*]], i64 2
100 ; SLM-NEXT: [[A3:%.*]] = extractelement <4 x float> [[A]], i64 3
101 ; SLM-NEXT: [[TMP1:%.*]] = shufflevector <4 x float> [[A]], <4 x float> poison, <2 x i32> <i32 0, i32 1>
102 ; SLM-NEXT: [[TMP2:%.*]] = fmul <2 x float> [[TMP1]], <float 2.000000e+00, float 1.000000e+00>
103 ; SLM-NEXT: [[AB3:%.*]] = fmul float [[A3]], 2.000000e+00
104 ; SLM-NEXT: [[TMP3:%.*]] = shufflevector <2 x float> [[TMP2]], <2 x float> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
105 ; SLM-NEXT: [[R2:%.*]] = insertelement <4 x float> [[TMP3]], float [[A2]], i64 2
106 ; SLM-NEXT: [[R3:%.*]] = insertelement <4 x float> [[R2]], float [[AB3]], i64 3
107 ; SLM-NEXT: ret <4 x float> [[R3]]
109 ; AVX-LABEL: @fmul_fdiv_v4f32_const(
110 ; AVX-NEXT: [[TMP1:%.*]] = fmul <4 x float> [[A:%.*]], <float 2.000000e+00, float 1.000000e+00, float 1.000000e+00, float 2.000000e+00>
111 ; AVX-NEXT: ret <4 x float> [[TMP1]]
113 ; AVX512-LABEL: @fmul_fdiv_v4f32_const(
114 ; AVX512-NEXT: [[TMP1:%.*]] = fmul <4 x float> [[A:%.*]], <float 2.000000e+00, float 1.000000e+00, float 1.000000e+00, float 2.000000e+00>
115 ; AVX512-NEXT: ret <4 x float> [[TMP1]]
117 %a0 = extractelement <4 x float> %a, i32 0
118 %a1 = extractelement <4 x float> %a, i32 1
119 %a2 = extractelement <4 x float> %a, i32 2
120 %a3 = extractelement <4 x float> %a, i32 3
121 %ab0 = fmul float %a0, 2.0
122 %ab1 = fmul float %a1, 1.0
123 %ab2 = fdiv float %a2, 1.0
124 %ab3 = fdiv float %a3, 0.5
125 %r0 = insertelement <4 x float> undef, float %ab0, i32 0
126 %r1 = insertelement <4 x float> %r0, float %ab1, i32 1
127 %r2 = insertelement <4 x float> %r1, float %ab2, i32 2
128 %r3 = insertelement <4 x float> %r2, float %ab3, i32 3