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
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=i686-unknown -mattr=+sse2 | FileCheck %s --check-prefix=X86
; RUN: llc < %s -mtriple=x86_64-unknown -mattr=+sse2 | FileCheck %s --check-prefix=X64
; test vector shifts converted to proper SSE2 vector shifts when the shift
; amounts are the same.
define void @shift1a(<2 x i64> %val, <2 x i64>* %dst) nounwind {
; X86-LABEL: shift1a:
; X86: # %bb.0: # %entry
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NEXT: psrlq $32, %xmm0
; X86-NEXT: movdqa %xmm0, (%eax)
; X86-NEXT: retl
;
; X64-LABEL: shift1a:
; X64: # %bb.0: # %entry
; X64-NEXT: psrlq $32, %xmm0
; X64-NEXT: movdqa %xmm0, (%rdi)
; X64-NEXT: retq
entry:
%lshr = lshr <2 x i64> %val, < i64 32, i64 32 >
store <2 x i64> %lshr, <2 x i64>* %dst
ret void
}
define void @shift1b(<2 x i64> %val, <2 x i64>* %dst, i64 %amt) nounwind {
; X86-LABEL: shift1b:
; X86: # %bb.0: # %entry
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NEXT: movq {{.*#+}} xmm1 = mem[0],zero
; X86-NEXT: psrlq %xmm1, %xmm0
; X86-NEXT: movdqa %xmm0, (%eax)
; X86-NEXT: retl
;
; X64-LABEL: shift1b:
; X64: # %bb.0: # %entry
; X64-NEXT: movq %rsi, %xmm1
; X64-NEXT: psrlq %xmm1, %xmm0
; X64-NEXT: movdqa %xmm0, (%rdi)
; X64-NEXT: retq
entry:
%0 = insertelement <2 x i64> undef, i64 %amt, i32 0
%1 = insertelement <2 x i64> %0, i64 %amt, i32 1
%lshr = lshr <2 x i64> %val, %1
store <2 x i64> %lshr, <2 x i64>* %dst
ret void
}
define void @shift2a(<4 x i32> %val, <4 x i32>* %dst) nounwind {
; X86-LABEL: shift2a:
; X86: # %bb.0: # %entry
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NEXT: psrld $17, %xmm0
; X86-NEXT: movdqa %xmm0, (%eax)
; X86-NEXT: retl
;
; X64-LABEL: shift2a:
; X64: # %bb.0: # %entry
; X64-NEXT: psrld $17, %xmm0
; X64-NEXT: movdqa %xmm0, (%rdi)
; X64-NEXT: retq
entry:
%lshr = lshr <4 x i32> %val, < i32 17, i32 17, i32 17, i32 17 >
store <4 x i32> %lshr, <4 x i32>* %dst
ret void
}
define void @shift2b(<4 x i32> %val, <4 x i32>* %dst, i32 %amt) nounwind {
; X86-LABEL: shift2b:
; X86: # %bb.0: # %entry
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NEXT: movd {{.*#+}} xmm1 = mem[0],zero,zero,zero
; X86-NEXT: psrld %xmm1, %xmm0
; X86-NEXT: movdqa %xmm0, (%eax)
; X86-NEXT: retl
;
; X64-LABEL: shift2b:
; X64: # %bb.0: # %entry
; X64-NEXT: movd %esi, %xmm1
; X64-NEXT: psrld %xmm1, %xmm0
; X64-NEXT: movdqa %xmm0, (%rdi)
; X64-NEXT: retq
entry:
%0 = insertelement <4 x i32> undef, i32 %amt, i32 0
%1 = insertelement <4 x i32> %0, i32 %amt, i32 1
%2 = insertelement <4 x i32> %1, i32 %amt, i32 2
%3 = insertelement <4 x i32> %2, i32 %amt, i32 3
%lshr = lshr <4 x i32> %val, %3
store <4 x i32> %lshr, <4 x i32>* %dst
ret void
}
define void @shift3a(<8 x i16> %val, <8 x i16>* %dst) nounwind {
; X86-LABEL: shift3a:
; X86: # %bb.0: # %entry
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NEXT: psrlw $5, %xmm0
; X86-NEXT: movdqa %xmm0, (%eax)
; X86-NEXT: retl
;
; X64-LABEL: shift3a:
; X64: # %bb.0: # %entry
; X64-NEXT: psrlw $5, %xmm0
; X64-NEXT: movdqa %xmm0, (%rdi)
; X64-NEXT: retq
entry:
%lshr = lshr <8 x i16> %val, < i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5 >
store <8 x i16> %lshr, <8 x i16>* %dst
ret void
}
; properly zero extend the shift amount
define void @shift3b(<8 x i16> %val, <8 x i16>* %dst, i16 %amt) nounwind {
; X86-LABEL: shift3b:
; X86: # %bb.0: # %entry
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
; X86-NEXT: movzwl {{[0-9]+}}(%esp), %ecx
; X86-NEXT: movd %ecx, %xmm1
; X86-NEXT: psrlw %xmm1, %xmm0
; X86-NEXT: movdqa %xmm0, (%eax)
; X86-NEXT: retl
;
; X64-LABEL: shift3b:
; X64: # %bb.0: # %entry
; X64-NEXT: movzwl %si, %eax
; X64-NEXT: movd %eax, %xmm1
; X64-NEXT: psrlw %xmm1, %xmm0
; X64-NEXT: movdqa %xmm0, (%rdi)
; X64-NEXT: retq
entry:
%0 = insertelement <8 x i16> undef, i16 %amt, i32 0
%1 = insertelement <8 x i16> %0, i16 %amt, i32 1
%2 = insertelement <8 x i16> %1, i16 %amt, i32 2
%3 = insertelement <8 x i16> %2, i16 %amt, i32 3
%4 = insertelement <8 x i16> %3, i16 %amt, i32 4
%5 = insertelement <8 x i16> %4, i16 %amt, i32 5
%6 = insertelement <8 x i16> %5, i16 %amt, i32 6
%7 = insertelement <8 x i16> %6, i16 %amt, i32 7
%lshr = lshr <8 x i16> %val, %7
store <8 x i16> %lshr, <8 x i16>* %dst
ret void
}
|
