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
|
; RUN: llc < %s -verify-machineinstrs -mtriple=aarch64-none-linux-gnu -mattr=+neon -fp-contract=fast
; FIXME: We should not generate ld/st for such register spill/fill, because the
; test case seems very simple and the register pressure is not high. If the
; spill/fill algorithm is optimized, this test case may not be triggered. And
; then we can delete it.
define i32 @spill.DPairReg(i8* %arg1, i32 %arg2) {
; CHECK-LABEL: spill.DPairReg:
; CHECK: ld2 {v{{[0-9]+}}.2s, v{{[0-9]+}}.2s}, [{{x[0-9]+|sp}}]
; CHECK: st1 {v{{[0-9]+}}.16b, v{{[0-9]+}}.16b}, [{{x[0-9]+|sp}}]
; CHECK: ld1 {v{{[0-9]+}}.16b, v{{[0-9]+}}.16b}, [{{x[0-9]+|sp}}]
entry:
%vld = tail call { <2 x i32>, <2 x i32> } @llvm.arm.neon.vld2.v2i32(i8* %arg1, i32 4)
%cmp = icmp eq i32 %arg2, 0
br i1 %cmp, label %if.then, label %if.end
if.then:
tail call void @foo()
br label %if.end
if.end:
%vld.extract = extractvalue { <2 x i32>, <2 x i32> } %vld, 0
%res = extractelement <2 x i32> %vld.extract, i32 1
ret i32 %res
}
define i16 @spill.DTripleReg(i8* %arg1, i32 %arg2) {
; CHECK-LABEL: spill.DTripleReg:
; CHECK: ld3 {v{{[0-9]+}}.4h, v{{[0-9]+}}.4h, v{{[0-9]+}}.4h}, [{{x[0-9]+|sp}}]
; CHECK: st1 {v{{[0-9]+}}.16b, v{{[0-9]+}}.16b, v{{[0-9]+}}.16b}, [{{x[0-9]+|sp}}]
; CHECK: ld1 {v{{[0-9]+}}.16b, v{{[0-9]+}}.16b, v{{[0-9]+}}.16b}, [{{x[0-9]+|sp}}]
entry:
%vld = tail call { <4 x i16>, <4 x i16>, <4 x i16> } @llvm.arm.neon.vld3.v4i16(i8* %arg1, i32 4)
%cmp = icmp eq i32 %arg2, 0
br i1 %cmp, label %if.then, label %if.end
if.then:
tail call void @foo()
br label %if.end
if.end:
%vld.extract = extractvalue { <4 x i16>, <4 x i16>, <4 x i16> } %vld, 0
%res = extractelement <4 x i16> %vld.extract, i32 1
ret i16 %res
}
define i16 @spill.DQuadReg(i8* %arg1, i32 %arg2) {
; CHECK-LABEL: spill.DQuadReg:
; CHECK: ld4 {v{{[0-9]+}}.4h, v{{[0-9]+}}.4h, v{{[0-9]+}}.4h, v{{[0-9]+}}.4h}, [{{x[0-9]+|sp}}]
; CHECK: st1 {v{{[0-9]+}}.16b, v{{[0-9]+}}.16b, v{{[0-9]+}}.16b, v{{[0-9]+}}.16b}, [{{x[0-9]+|sp}}]
; CHECK: ld1 {v{{[0-9]+}}.16b, v{{[0-9]+}}.16b, v{{[0-9]+}}.16b, v{{[0-9]+}}.16b}, [{{x[0-9]+|sp}}]
entry:
%vld = tail call { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> } @llvm.arm.neon.vld4.v4i16(i8* %arg1, i32 4)
%cmp = icmp eq i32 %arg2, 0
br i1 %cmp, label %if.then, label %if.end
if.then:
tail call void @foo()
br label %if.end
if.end:
%vld.extract = extractvalue { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> } %vld, 0
%res = extractelement <4 x i16> %vld.extract, i32 0
ret i16 %res
}
define i32 @spill.QPairReg(i8* %arg1, i32 %arg2) {
; CHECK-LABEL: spill.QPairReg:
; CHECK: ld3 {v{{[0-9]+}}.4s, v{{[0-9]+}}.4s}, [{{x[0-9]+|sp}}]
; CHECK: st1 {v{{[0-9]+}}.16b, v{{[0-9]+}}.16b}, [{{x[0-9]+|sp}}]
; CHECK: ld1 {v{{[0-9]+}}.16b, v{{[0-9]+}}.16b}, [{{x[0-9]+|sp}}]
entry:
%vld = tail call { <4 x i32>, <4 x i32> } @llvm.arm.neon.vld2.v4i32(i8* %arg1, i32 4)
%cmp = icmp eq i32 %arg2, 0
br i1 %cmp, label %if.then, label %if.end
if.then:
tail call void @foo()
br label %if.end
if.end:
%vld.extract = extractvalue { <4 x i32>, <4 x i32> } %vld, 0
%res = extractelement <4 x i32> %vld.extract, i32 1
ret i32 %res
}
define float @spill.QTripleReg(i8* %arg1, i32 %arg2) {
; CHECK-LABEL: spill.QTripleReg:
; CHECK: ld3 {v{{[0-9]+}}.4s, v{{[0-9]+}}.4s, v{{[0-9]+}}.4s}, [{{x[0-9]+|sp}}]
; CHECK: st1 {v{{[0-9]+}}.16b, v{{[0-9]+}}.16b, v{{[0-9]+}}.16b}, [{{x[0-9]+|sp}}]
; CHECK: ld1 {v{{[0-9]+}}.16b, v{{[0-9]+}}.16b, v{{[0-9]+}}.16b}, [{{x[0-9]+|sp}}]
entry:
%vld3 = tail call { <4 x float>, <4 x float>, <4 x float> } @llvm.arm.neon.vld3.v4f32(i8* %arg1, i32 4)
%cmp = icmp eq i32 %arg2, 0
br i1 %cmp, label %if.then, label %if.end
if.then:
tail call void @foo()
br label %if.end
if.end:
%vld3.extract = extractvalue { <4 x float>, <4 x float>, <4 x float> } %vld3, 0
%res = extractelement <4 x float> %vld3.extract, i32 1
ret float %res
}
define i8 @spill.QQuadReg(i8* %arg1, i32 %arg2) {
; CHECK-LABEL: spill.QQuadReg:
; CHECK: ld4 {v{{[0-9]+}}.2d, v{{[0-9]+}}.2d, v{{[0-9]+}}.2d, v{{[0-9]+}}.2d}, [{{x[0-9]+|sp}}]
; CHECK: st1 {v{{[0-9]+}}.16b, v{{[0-9]+}}.16b, v{{[0-9]+}}.16b, v{{[0-9]+}}.16b}, [{{x[0-9]+|sp}}]
; CHECK: ld1 {v{{[0-9]+}}.16b, v{{[0-9]+}}.16b, v{{[0-9]+}}.16b, v{{[0-9]+}}.16b}, [{{x[0-9]+|sp}}]
entry:
%vld = tail call { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } @llvm.arm.neon.vld4.v16i8(i8* %arg1, i32 4)
%cmp = icmp eq i32 %arg2, 0
br i1 %cmp, label %if.then, label %if.end
if.then:
tail call void @foo()
br label %if.end
if.end:
%vld.extract = extractvalue { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } %vld, 0
%res = extractelement <16 x i8> %vld.extract, i32 1
ret i8 %res
}
declare { <2 x i32>, <2 x i32> } @llvm.arm.neon.vld2.v2i32(i8*, i32)
declare { <4 x i16>, <4 x i16>, <4 x i16> } @llvm.arm.neon.vld3.v4i16(i8*, i32)
declare { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> } @llvm.arm.neon.vld4.v4i16(i8*, i32)
declare { <4 x i32>, <4 x i32> } @llvm.arm.neon.vld2.v4i32(i8*, i32)
declare { <4 x float>, <4 x float>, <4 x float> } @llvm.arm.neon.vld3.v4f32(i8*, i32)
declare { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } @llvm.arm.neon.vld4.v16i8(i8*, i32)
declare void @foo()
|
