1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+sse2 | FileCheck %s --check-prefix=SSE2
; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+ssse3 | FileCheck %s --check-prefixes=SSSE3,SSSE3-SLOW
; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+ssse3,fast-hops | FileCheck %s --check-prefixes=SSSE3,SSSE3-FAST
; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx | FileCheck %s --check-prefixes=AVX1,AVX1-SLOW
; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx,fast-hops | FileCheck %s --check-prefixes=AVX1,AVX1-FAST
; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx2 | FileCheck %s --check-prefix=AVX2
define float @pr26491(<4 x float> %a0) {
; SSE2-LABEL: pr26491:
; SSE2: # %bb.0:
; SSE2-NEXT: movaps %xmm0, %xmm1
; SSE2-NEXT: shufps {{.*#+}} xmm1 = xmm1[1,1],xmm0[3,3]
; SSE2-NEXT: addps %xmm1, %xmm0
; SSE2-NEXT: movaps %xmm0, %xmm1
; SSE2-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSE2-NEXT: addss %xmm1, %xmm0
; SSE2-NEXT: retq
;
; SSSE3-SLOW-LABEL: pr26491:
; SSSE3-SLOW: # %bb.0:
; SSSE3-SLOW-NEXT: movshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; SSSE3-SLOW-NEXT: addps %xmm1, %xmm0
; SSSE3-SLOW-NEXT: movaps %xmm0, %xmm1
; SSSE3-SLOW-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
; SSSE3-SLOW-NEXT: addss %xmm1, %xmm0
; SSSE3-SLOW-NEXT: retq
;
; SSSE3-FAST-LABEL: pr26491:
; SSSE3-FAST: # %bb.0:
; SSSE3-FAST-NEXT: haddps %xmm0, %xmm0
; SSSE3-FAST-NEXT: movshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; SSSE3-FAST-NEXT: addss %xmm1, %xmm0
; SSSE3-FAST-NEXT: retq
;
; AVX1-SLOW-LABEL: pr26491:
; AVX1-SLOW: # %bb.0:
; AVX1-SLOW-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX1-SLOW-NEXT: vaddps %xmm0, %xmm1, %xmm0
; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddss %xmm0, %xmm1, %xmm0
; AVX1-SLOW-NEXT: retq
;
; AVX1-FAST-LABEL: pr26491:
; AVX1-FAST: # %bb.0:
; AVX1-FAST-NEXT: vhaddps %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX1-FAST-NEXT: vaddss %xmm0, %xmm1, %xmm0
; AVX1-FAST-NEXT: retq
;
; AVX2-LABEL: pr26491:
; AVX2: # %bb.0:
; AVX2-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX2-NEXT: vaddps %xmm0, %xmm1, %xmm0
; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX2-NEXT: vaddss %xmm0, %xmm1, %xmm0
; AVX2-NEXT: retq
%1 = shufflevector <4 x float> %a0, <4 x float> undef, <4 x i32> <i32 1, i32 1, i32 3, i32 3>
%2 = fadd <4 x float> %1, %a0
%3 = extractelement <4 x float> %2, i32 2
%4 = extractelement <4 x float> %2, i32 0
%5 = fadd float %3, %4
ret float %5
}
; When simplifying away a splat (broadcast), the hop type must match the shuffle type.
define <4 x double> @PR41414(i64 %x, <4 x double> %y) {
; SSE2-LABEL: PR41414:
; SSE2: # %bb.0:
; SSE2-NEXT: movq %rdi, %xmm2
; SSE2-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],mem[0],xmm2[1],mem[1]
; SSE2-NEXT: subpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2
; SSE2-NEXT: movapd %xmm2, %xmm3
; SSE2-NEXT: unpcklpd {{.*#+}} xmm3 = xmm3[0],xmm2[0]
; SSE2-NEXT: unpckhpd {{.*#+}} xmm2 = xmm2[1,1]
; SSE2-NEXT: addpd %xmm3, %xmm2
; SSE2-NEXT: divpd %xmm2, %xmm1
; SSE2-NEXT: divpd %xmm2, %xmm0
; SSE2-NEXT: xorpd %xmm2, %xmm2
; SSE2-NEXT: addpd %xmm2, %xmm0
; SSE2-NEXT: addpd %xmm2, %xmm1
; SSE2-NEXT: retq
;
; SSSE3-LABEL: PR41414:
; SSSE3: # %bb.0:
; SSSE3-NEXT: movq %rdi, %xmm2
; SSSE3-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],mem[0],xmm2[1],mem[1]
; SSSE3-NEXT: subpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2
; SSSE3-NEXT: haddpd %xmm2, %xmm2
; SSSE3-NEXT: divpd %xmm2, %xmm1
; SSSE3-NEXT: divpd %xmm2, %xmm0
; SSSE3-NEXT: xorpd %xmm2, %xmm2
; SSSE3-NEXT: addpd %xmm2, %xmm0
; SSSE3-NEXT: addpd %xmm2, %xmm1
; SSSE3-NEXT: retq
;
; AVX1-LABEL: PR41414:
; AVX1: # %bb.0:
; AVX1-NEXT: vmovq %rdi, %xmm1
; AVX1-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
; AVX1-NEXT: vsubpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
; AVX1-NEXT: vhaddpd %xmm1, %xmm1, %xmm1
; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm1, %ymm1
; AVX1-NEXT: vdivpd %ymm1, %ymm0, %ymm0
; AVX1-NEXT: vxorpd %xmm1, %xmm1, %xmm1
; AVX1-NEXT: vaddpd %ymm1, %ymm0, %ymm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: PR41414:
; AVX2: # %bb.0:
; AVX2-NEXT: vmovq %rdi, %xmm1
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
; AVX2-NEXT: vsubpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
; AVX2-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX2-NEXT: vaddsd %xmm1, %xmm2, %xmm1
; AVX2-NEXT: vbroadcastsd %xmm1, %ymm1
; AVX2-NEXT: vdivpd %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vxorpd %xmm1, %xmm1, %xmm1
; AVX2-NEXT: vaddpd %ymm1, %ymm0, %ymm0
; AVX2-NEXT: retq
%conv = uitofp i64 %x to double
%t0 = insertelement <4 x double> undef, double %conv, i32 0
%t1 = shufflevector <4 x double> %t0, <4 x double> undef, <4 x i32> zeroinitializer
%t2 = fdiv <4 x double> %y, %t1
%t3 = fadd <4 x double> zeroinitializer, %t2
ret <4 x double> %t3
}
define <4 x float> @PR48823(<4 x float> %0, <4 x float> %1) {
; SSE2-LABEL: PR48823:
; SSE2: # %bb.0:
; SSE2-NEXT: movaps %xmm0, %xmm2
; SSE2-NEXT: shufps {{.*#+}} xmm2 = xmm2[1,1],xmm1[2,3]
; SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,2]
; SSE2-NEXT: subps %xmm2, %xmm0
; SSE2-NEXT: retq
;
; SSSE3-SLOW-LABEL: PR48823:
; SSSE3-SLOW: # %bb.0:
; SSSE3-SLOW-NEXT: movaps %xmm0, %xmm2
; SSSE3-SLOW-NEXT: shufps {{.*#+}} xmm2 = xmm2[1,1],xmm1[2,3]
; SSSE3-SLOW-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,2]
; SSSE3-SLOW-NEXT: subps %xmm2, %xmm0
; SSSE3-SLOW-NEXT: retq
;
; SSSE3-FAST-LABEL: PR48823:
; SSSE3-FAST: # %bb.0:
; SSSE3-FAST-NEXT: hsubps %xmm1, %xmm0
; SSSE3-FAST-NEXT: retq
;
; AVX1-SLOW-LABEL: PR48823:
; AVX1-SLOW: # %bb.0:
; AVX1-SLOW-NEXT: vshufps {{.*#+}} xmm2 = xmm0[1,1],xmm1[2,3]
; AVX1-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,2]
; AVX1-SLOW-NEXT: vsubps %xmm2, %xmm0, %xmm0
; AVX1-SLOW-NEXT: retq
;
; AVX1-FAST-LABEL: PR48823:
; AVX1-FAST: # %bb.0:
; AVX1-FAST-NEXT: vhsubps %xmm1, %xmm0, %xmm0
; AVX1-FAST-NEXT: retq
;
; AVX2-LABEL: PR48823:
; AVX2: # %bb.0:
; AVX2-NEXT: vshufps {{.*#+}} xmm2 = xmm0[1,1],xmm1[2,3]
; AVX2-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,2]
; AVX2-NEXT: vsubps %xmm2, %xmm0, %xmm0
; AVX2-NEXT: retq
%3 = shufflevector <4 x float> %0, <4 x float> poison, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
%4 = fsub <4 x float> %0, %3
%5 = shufflevector <4 x float> %1, <4 x float> poison, <4 x i32> <i32 undef, i32 undef, i32 undef, i32 2>
%6 = fsub <4 x float> %5, %1
%7 = shufflevector <4 x float> %4, <4 x float> %6, <4 x i32> <i32 0, i32 undef, i32 undef, i32 7>
ret <4 x float> %7
}
|
