aboutsummaryrefslogblamecommitdiff
path: root/test/CodeGen/AArch64/arm64-neon-vector-list-spill.ll
blob: 8262fe43a66c879e29b6fee3188b802f17279bab (plain) (tree) 
5ca98fd98791















































































































































































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















































































































































































                                                                                                                                                                                     
; RUN: llc < %s -verify-machineinstrs -mtriple=arm64-none-linux-gnu -mattr=+neon -fp-contract=fast | FileCheck %s

; 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(i32* %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]+}}.2d, v{{[0-9]+}}.2d }, [{{x[0-9]+|sp}}]
; CHECK: ld1 { v{{[0-9]+}}.2d, v{{[0-9]+}}.2d }, [{{x[0-9]+|sp}}]
entry:
  %vld = tail call { <2 x i32>, <2 x i32> } @llvm.aarch64.neon.ld2.v2i32.p0i32(i32* %arg1)
  %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(i16* %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]+}}.2d, v{{[0-9]+}}.2d, v{{[0-9]+}}.2d }, [{{x[0-9]+|sp}}]
; CHECK: ld1 { v{{[0-9]+}}.2d, v{{[0-9]+}}.2d, v{{[0-9]+}}.2d }, [{{x[0-9]+|sp}}]
entry:
  %vld = tail call { <4 x i16>, <4 x i16>, <4 x i16> } @llvm.aarch64.neon.ld3.v4i16.p0i16(i16* %arg1)
  %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(i16* %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]+}}.2d, v{{[0-9]+}}.2d, v{{[0-9]+}}.2d, v{{[0-9]+}}.2d }, [{{x[0-9]+|sp}}]
; CHECK: ld1 { v{{[0-9]+}}.2d, v{{[0-9]+}}.2d, v{{[0-9]+}}.2d, v{{[0-9]+}}.2d }, [{{x[0-9]+|sp}}]
entry:
  %vld = tail call { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> } @llvm.aarch64.neon.ld4.v4i16.p0i16(i16* %arg1)
  %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(i32* %arg1, i32 %arg2) {
; CHECK-LABEL: spill.QPairReg:
; CHECK: ld2 { v{{[0-9]+}}.4s, v{{[0-9]+}}.4s }, [{{x[0-9]+|sp}}]
; CHECK: st1 { v{{[0-9]+}}.2d, v{{[0-9]+}}.2d }, [{{x[0-9]+|sp}}]
; CHECK: ld1 { v{{[0-9]+}}.2d, v{{[0-9]+}}.2d }, [{{x[0-9]+|sp}}]
entry:
  %vld = tail call { <4 x i32>, <4 x i32> } @llvm.aarch64.neon.ld2.v4i32.p0i32(i32* %arg1)
  %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(float* %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]+}}.2d, v{{[0-9]+}}.2d, v{{[0-9]+}}.2d }, [{{x[0-9]+|sp}}]
; CHECK: ld1 { v{{[0-9]+}}.2d, v{{[0-9]+}}.2d, v{{[0-9]+}}.2d }, [{{x[0-9]+|sp}}]
entry:
  %vld3 = tail call { <4 x float>, <4 x float>, <4 x float> } @llvm.aarch64.neon.ld3.v4f32.p0f32(float* %arg1)
  %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]+}}.16b, v{{[0-9]+}}.16b, v{{[0-9]+}}.16b, v{{[0-9]+}}.16b }, [{{x[0-9]+|sp}}]
; CHECK: st1 { v{{[0-9]+}}.2d, v{{[0-9]+}}.2d, v{{[0-9]+}}.2d, v{{[0-9]+}}.2d }, [{{x[0-9]+|sp}}]
; CHECK: ld1 { v{{[0-9]+}}.2d, v{{[0-9]+}}.2d, v{{[0-9]+}}.2d, v{{[0-9]+}}.2d }, [{{x[0-9]+|sp}}]
entry:
  %vld = tail call { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } @llvm.aarch64.neon.ld4.v16i8.p0i8(i8* %arg1)
  %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.aarch64.neon.ld2.v2i32.p0i32(i32*)
declare { <4 x i16>, <4 x i16>, <4 x i16> } @llvm.aarch64.neon.ld3.v4i16.p0i16(i16*)
declare { <4 x i16>, <4 x i16>, <4 x i16>, <4 x i16> } @llvm.aarch64.neon.ld4.v4i16.p0i16(i16*)
declare { <4 x i32>, <4 x i32> } @llvm.aarch64.neon.ld2.v4i32.p0i32(i32*)
declare { <4 x float>, <4 x float>, <4 x float> } @llvm.aarch64.neon.ld3.v4f32.p0f32(float*)
declare { <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8> } @llvm.aarch64.neon.ld4.v16i8.p0i8(i8*)

declare void @foo()

; 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.
; check the spill for Register Class QPair_with_qsub_0_in_FPR128Lo
define <8 x i16> @test_2xFPR128Lo(i64 %got, i64* %ptr, <1 x i64> %a) {
  tail call void @llvm.aarch64.neon.st2lane.v1i64.p0i64(<1 x i64> zeroinitializer, <1 x i64> zeroinitializer, i64 0, i64* %ptr)
  tail call void @foo()
  %sv = shufflevector <1 x i64> zeroinitializer, <1 x i64> %a, <2 x i32> <i32 0, i32 1>
  %1 = bitcast <2 x i64> %sv to <8 x i16>
  %2 = shufflevector <8 x i16> %1, <8 x i16> undef, <8 x i32> <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2>
  %3 = mul <8 x i16> %2, %2
  ret <8 x i16> %3
}

; check the spill for Register Class QTriple_with_qsub_0_in_FPR128Lo
define <8 x i16> @test_3xFPR128Lo(i64 %got, i64* %ptr, <1 x i64> %a) {
  tail call void @llvm.aarch64.neon.st3lane.v1i64.p0i64(<1 x i64> zeroinitializer, <1 x i64> zeroinitializer, <1 x i64> zeroinitializer, i64 0, i64* %ptr)
  tail call void @foo()
  %sv = shufflevector <1 x i64> zeroinitializer, <1 x i64> %a, <2 x i32> <i32 0, i32 1>
  %1 = bitcast <2 x i64> %sv to <8 x i16>
  %2 = shufflevector <8 x i16> %1, <8 x i16> undef, <8 x i32> <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2>
  %3 = mul <8 x i16> %2, %2
  ret <8 x i16> %3
}

; check the spill for Register Class QQuad_with_qsub_0_in_FPR128Lo
define <8 x i16> @test_4xFPR128Lo(i64 %got, i64* %ptr, <1 x i64> %a) {
  tail call void @llvm.aarch64.neon.st4lane.v1i64.p0i64(<1 x i64> zeroinitializer, <1 x i64> zeroinitializer, <1 x i64> zeroinitializer, <1 x i64> zeroinitializer, i64 0, i64* %ptr)
  tail call void @foo()
  %sv = shufflevector <1 x i64> zeroinitializer, <1 x i64> %a, <2 x i32> <i32 0, i32 1>
  %1 = bitcast <2 x i64> %sv to <8 x i16>
  %2 = shufflevector <8 x i16> %1, <8 x i16> undef, <8 x i32> <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2>
  %3 = mul <8 x i16> %2, %2
  ret <8 x i16> %3
}

declare void @llvm.aarch64.neon.st2lane.v1i64.p0i64(<1 x i64>, <1 x i64>, i64, i64*)
declare void @llvm.aarch64.neon.st3lane.v1i64.p0i64(<1 x i64>, <1 x i64>, <1 x i64>, i64, i64*)
declare void @llvm.aarch64.neon.st4lane.v1i64.p0i64(<1 x i64>, <1 x i64>, <1 x i64>, <1 x i64>, i64, i64*)
generated by cgit v1.2.3 (git 2.25.1) at 2024-06-29 08:59:49 +0000