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
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=i686-unknown -mattr=+sse4.1 | FileCheck %s --check-prefix=X32-SSE
; RUN: llc < %s -mtriple=i686-unknown -mattr=+avx | FileCheck %s --check-prefix=X32-AVX
; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+sse4.1 | FileCheck %s --check-prefix=X64-SSE
; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+avx | FileCheck %s --check-prefix=X64-AVX
define void @fptrunc_frommem2(<2 x double>* %in, <2 x float>* %out) {
; X32-SSE-LABEL: fptrunc_frommem2:
; X32-SSE: # BB#0: # %entry
; X32-SSE-NEXT: movl {{[0-9]+}}(%esp), %eax
; X32-SSE-NEXT: movl {{[0-9]+}}(%esp), %ecx
; X32-SSE-NEXT: cvtpd2ps (%ecx), %xmm0
; X32-SSE-NEXT: extractps $1, %xmm0, 4(%eax)
; X32-SSE-NEXT: movss %xmm0, (%eax)
; X32-SSE-NEXT: retl
;
; X32-AVX-LABEL: fptrunc_frommem2:
; X32-AVX: # BB#0: # %entry
; X32-AVX-NEXT: movl {{[0-9]+}}(%esp), %eax
; X32-AVX-NEXT: movl {{[0-9]+}}(%esp), %ecx
; X32-AVX-NEXT: vcvtpd2psx (%ecx), %xmm0
; X32-AVX-NEXT: vextractps $1, %xmm0, 4(%eax)
; X32-AVX-NEXT: vmovss %xmm0, (%eax)
; X32-AVX-NEXT: retl
;
; X64-SSE-LABEL: fptrunc_frommem2:
; X64-SSE: # BB#0: # %entry
; X64-SSE-NEXT: cvtpd2ps (%rdi), %xmm0
; X64-SSE-NEXT: movlpd %xmm0, (%rsi)
; X64-SSE-NEXT: retq
;
; X64-AVX-LABEL: fptrunc_frommem2:
; X64-AVX: # BB#0: # %entry
; X64-AVX-NEXT: vcvtpd2psx (%rdi), %xmm0
; X64-AVX-NEXT: vmovlpd %xmm0, (%rsi)
; X64-AVX-NEXT: retq
entry:
%0 = load <2 x double>, <2 x double>* %in
%1 = fptrunc <2 x double> %0 to <2 x float>
store <2 x float> %1, <2 x float>* %out, align 1
ret void
}
define void @fptrunc_frommem4(<4 x double>* %in, <4 x float>* %out) {
; X32-SSE-LABEL: fptrunc_frommem4:
; X32-SSE: # BB#0: # %entry
; X32-SSE-NEXT: movl {{[0-9]+}}(%esp), %eax
; X32-SSE-NEXT: movl {{[0-9]+}}(%esp), %ecx
; X32-SSE-NEXT: cvtpd2ps 16(%ecx), %xmm0
; X32-SSE-NEXT: cvtpd2ps (%ecx), %xmm1
; X32-SSE-NEXT: unpcklpd {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; X32-SSE-NEXT: movupd %xmm1, (%eax)
; X32-SSE-NEXT: retl
;
; X32-AVX-LABEL: fptrunc_frommem4:
; X32-AVX: # BB#0: # %entry
; X32-AVX-NEXT: movl {{[0-9]+}}(%esp), %eax
; X32-AVX-NEXT: movl {{[0-9]+}}(%esp), %ecx
; X32-AVX-NEXT: vcvtpd2psy (%ecx), %xmm0
; X32-AVX-NEXT: vmovupd %xmm0, (%eax)
; X32-AVX-NEXT: retl
;
; X64-SSE-LABEL: fptrunc_frommem4:
; X64-SSE: # BB#0: # %entry
; X64-SSE-NEXT: cvtpd2ps 16(%rdi), %xmm0
; X64-SSE-NEXT: cvtpd2ps (%rdi), %xmm1
; X64-SSE-NEXT: unpcklpd {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; X64-SSE-NEXT: movupd %xmm1, (%rsi)
; X64-SSE-NEXT: retq
;
; X64-AVX-LABEL: fptrunc_frommem4:
; X64-AVX: # BB#0: # %entry
; X64-AVX-NEXT: vcvtpd2psy (%rdi), %xmm0
; X64-AVX-NEXT: vmovupd %xmm0, (%rsi)
; X64-AVX-NEXT: retq
entry:
%0 = load <4 x double>, <4 x double>* %in
%1 = fptrunc <4 x double> %0 to <4 x float>
store <4 x float> %1, <4 x float>* %out, align 1
ret void
}
define void @fptrunc_frommem8(<8 x double>* %in, <8 x float>* %out) {
; X32-SSE-LABEL: fptrunc_frommem8:
; X32-SSE: # BB#0: # %entry
; X32-SSE-NEXT: movl {{[0-9]+}}(%esp), %eax
; X32-SSE-NEXT: movl {{[0-9]+}}(%esp), %ecx
; X32-SSE-NEXT: cvtpd2ps 16(%ecx), %xmm0
; X32-SSE-NEXT: cvtpd2ps (%ecx), %xmm1
; X32-SSE-NEXT: unpcklpd {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; X32-SSE-NEXT: cvtpd2ps 48(%ecx), %xmm0
; X32-SSE-NEXT: cvtpd2ps 32(%ecx), %xmm2
; X32-SSE-NEXT: unpcklpd {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; X32-SSE-NEXT: movupd %xmm2, 16(%eax)
; X32-SSE-NEXT: movupd %xmm1, (%eax)
; X32-SSE-NEXT: retl
;
; X32-AVX-LABEL: fptrunc_frommem8:
; X32-AVX: # BB#0: # %entry
; X32-AVX-NEXT: movl {{[0-9]+}}(%esp), %eax
; X32-AVX-NEXT: movl {{[0-9]+}}(%esp), %ecx
; X32-AVX-NEXT: vcvtpd2psy (%ecx), %xmm0
; X32-AVX-NEXT: vcvtpd2psy 32(%ecx), %xmm1
; X32-AVX-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
; X32-AVX-NEXT: vmovupd %ymm0, (%eax)
; X32-AVX-NEXT: vzeroupper
; X32-AVX-NEXT: retl
;
; X64-SSE-LABEL: fptrunc_frommem8:
; X64-SSE: # BB#0: # %entry
; X64-SSE-NEXT: cvtpd2ps 16(%rdi), %xmm0
; X64-SSE-NEXT: cvtpd2ps (%rdi), %xmm1
; X64-SSE-NEXT: unpcklpd {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; X64-SSE-NEXT: cvtpd2ps 48(%rdi), %xmm0
; X64-SSE-NEXT: cvtpd2ps 32(%rdi), %xmm2
; X64-SSE-NEXT: unpcklpd {{.*#+}} xmm2 = xmm2[0],xmm0[0]
; X64-SSE-NEXT: movupd %xmm2, 16(%rsi)
; X64-SSE-NEXT: movupd %xmm1, (%rsi)
; X64-SSE-NEXT: retq
;
; X64-AVX-LABEL: fptrunc_frommem8:
; X64-AVX: # BB#0: # %entry
; X64-AVX-NEXT: vcvtpd2psy (%rdi), %xmm0
; X64-AVX-NEXT: vcvtpd2psy 32(%rdi), %xmm1
; X64-AVX-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
; X64-AVX-NEXT: vmovupd %ymm0, (%rsi)
; X64-AVX-NEXT: vzeroupper
; X64-AVX-NEXT: retq
entry:
%0 = load <8 x double>, <8 x double>* %in
%1 = fptrunc <8 x double> %0 to <8 x float>
store <8 x float> %1, <8 x float>* %out, align 1
ret void
}
; FIXME: For exact truncations we should be able to fold this.
define <4 x float> @fptrunc_fromconst() {
; X32-SSE-LABEL: fptrunc_fromconst:
; X32-SSE: # BB#0: # %entry
; X32-SSE-NEXT: cvtpd2ps .LCPI3_0, %xmm1
; X32-SSE-NEXT: cvtpd2ps .LCPI3_1, %xmm0
; X32-SSE-NEXT: unpcklpd {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; X32-SSE-NEXT: retl
;
; X32-AVX-LABEL: fptrunc_fromconst:
; X32-AVX: # BB#0: # %entry
; X32-AVX-NEXT: vcvtpd2psy .LCPI3_0, %xmm0
; X32-AVX-NEXT: retl
;
; X64-SSE-LABEL: fptrunc_fromconst:
; X64-SSE: # BB#0: # %entry
; X64-SSE-NEXT: cvtpd2ps {{.*}}(%rip), %xmm1
; X64-SSE-NEXT: cvtpd2ps {{.*}}(%rip), %xmm0
; X64-SSE-NEXT: unpcklpd {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; X64-SSE-NEXT: retq
;
; X64-AVX-LABEL: fptrunc_fromconst:
; X64-AVX: # BB#0: # %entry
; X64-AVX-NEXT: vcvtpd2psy {{.*}}(%rip), %xmm0
; X64-AVX-NEXT: retq
entry:
%0 = insertelement <4 x double> undef, double 1.0, i32 0
%1 = insertelement <4 x double> %0, double -2.0, i32 1
%2 = insertelement <4 x double> %1, double +4.0, i32 2
%3 = insertelement <4 x double> %2, double -0.0, i32 3
%4 = fptrunc <4 x double> %3 to <4 x float>
ret <4 x float> %4
}
|
