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
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
; RUN: llc -mtriple=aarch64-linux-gnu -o - %s | FileCheck %s
; This file tests eliding stack slots when lowering the FSINCOS ISD node.
define { float, float } @sincos_f32_value_return(float %x) {
; CHECK-LABEL: sincos_f32_value_return:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: str x30, [sp, #-16]! // 8-byte Folded Spill
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: .cfi_offset w30, -16
; CHECK-NEXT: add x0, sp, #12
; CHECK-NEXT: add x1, sp, #8
; CHECK-NEXT: bl sincosf
; CHECK-NEXT: ldp s1, s0, [sp, #8]
; CHECK-NEXT: ldr x30, [sp], #16 // 8-byte Folded Reload
; CHECK-NEXT: ret
entry:
%sin = tail call float @llvm.sin.f32(float %x)
%cos = tail call float @llvm.cos.f32(float %x)
%ret_0 = insertvalue { float, float } poison, float %sin, 0
%ret_1 = insertvalue { float, float } %ret_0, float %cos, 1
ret { float, float } %ret_1
}
define void @sincos_f32_ptr_return(float %x, ptr noalias %out_sin, ptr noalias %out_cos) {
; CHECK-LABEL: sincos_f32_ptr_return:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: str x30, [sp, #-16]! // 8-byte Folded Spill
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: .cfi_offset w30, -16
; CHECK-NEXT: bl sincosf
; CHECK-NEXT: ldr x30, [sp], #16 // 8-byte Folded Reload
; CHECK-NEXT: ret
entry:
%sin = tail call float @llvm.sin.f32(float %x)
%cos = tail call float @llvm.cos.f32(float %x)
store float %sin, ptr %out_sin, align 4
store float %cos, ptr %out_cos, align 4
ret void
}
define float @sincos_f32_mixed_return(float %x, ptr %out_sin) {
; CHECK-LABEL: sincos_f32_mixed_return:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: str x30, [sp, #-16]! // 8-byte Folded Spill
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: .cfi_offset w30, -16
; CHECK-NEXT: add x1, sp, #12
; CHECK-NEXT: bl sincosf
; CHECK-NEXT: ldr s0, [sp, #12]
; CHECK-NEXT: ldr x30, [sp], #16 // 8-byte Folded Reload
; CHECK-NEXT: ret
entry:
%sin = tail call float @llvm.sin.f32(float %x)
%cos = tail call float @llvm.cos.f32(float %x)
store float %sin, ptr %out_sin, align 4
ret float %cos
}
define { double, double } @sincos_f64_value_return(double %x) {
; CHECK-LABEL: sincos_f64_value_return:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: sub sp, sp, #32
; CHECK-NEXT: str x30, [sp, #16] // 8-byte Folded Spill
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: .cfi_offset w30, -16
; CHECK-NEXT: add x0, sp, #24
; CHECK-NEXT: add x1, sp, #8
; CHECK-NEXT: bl sincos
; CHECK-NEXT: ldr d0, [sp, #24]
; CHECK-NEXT: ldr d1, [sp, #8]
; CHECK-NEXT: ldr x30, [sp, #16] // 8-byte Folded Reload
; CHECK-NEXT: add sp, sp, #32
; CHECK-NEXT: ret
entry:
%sin = tail call double @llvm.sin.f64(double %x)
%cos = tail call double @llvm.cos.f64(double %x)
%ret_0 = insertvalue { double, double } poison, double %sin, 0
%ret_1 = insertvalue { double, double } %ret_0, double %cos, 1
ret { double, double } %ret_1
}
define void @sincos_f64_ptr_return(double %x, ptr noalias %out_sin, ptr noalias %out_cos) {
; CHECK-LABEL: sincos_f64_ptr_return:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: str x30, [sp, #-16]! // 8-byte Folded Spill
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: .cfi_offset w30, -16
; CHECK-NEXT: bl sincos
; CHECK-NEXT: ldr x30, [sp], #16 // 8-byte Folded Reload
; CHECK-NEXT: ret
entry:
%sin = tail call double @llvm.sin.f64(double %x)
%cos = tail call double @llvm.cos.f64(double %x)
store double %sin, ptr %out_sin, align 8
store double %cos, ptr %out_cos, align 8
ret void
}
define double @sincos_f64_mixed_return(double %x, ptr %out_sin) {
; CHECK-LABEL: sincos_f64_mixed_return:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: str x30, [sp, #-16]! // 8-byte Folded Spill
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: .cfi_offset w30, -16
; CHECK-NEXT: add x1, sp, #8
; CHECK-NEXT: bl sincos
; CHECK-NEXT: ldr d0, [sp, #8]
; CHECK-NEXT: ldr x30, [sp], #16 // 8-byte Folded Reload
; CHECK-NEXT: ret
entry:
%sin = tail call double @llvm.sin.f64(double %x)
%cos = tail call double @llvm.cos.f64(double %x)
store double %sin, ptr %out_sin, align 8
ret double %cos
}
; Here %out_sin and %out_cos may alias so we can't replace both stores with the
; call to sincosf (as the order of stores in sincosf is not defined).
define void @sincos_may_alias(float %x, ptr %out_sin, ptr %out_cos) {
; CHECK-LABEL: sincos_may_alias:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: sub sp, sp, #32
; CHECK-NEXT: stp x30, x19, [sp, #16] // 16-byte Folded Spill
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: .cfi_offset w19, -8
; CHECK-NEXT: .cfi_offset w30, -16
; CHECK-NEXT: mov x19, x1
; CHECK-NEXT: add x1, sp, #12
; CHECK-NEXT: bl sincosf
; CHECK-NEXT: ldr s0, [sp, #12]
; CHECK-NEXT: str s0, [x19]
; CHECK-NEXT: ldp x30, x19, [sp, #16] // 16-byte Folded Reload
; CHECK-NEXT: add sp, sp, #32
; CHECK-NEXT: ret
entry:
%sin = tail call float @llvm.sin.f32(float %x)
%cos = tail call float @llvm.cos.f32(float %x)
store float %sin, ptr %out_sin, align 4
store float %cos, ptr %out_cos, align 4
ret void
}
; Here %out is used for both sin and cos (with the final value stored being cos).
define float @sincos_multiple_uses(float %x, ptr %out) {
; CHECK-LABEL: sincos_multiple_uses:
; CHECK: // %bb.0:
; CHECK-NEXT: str x30, [sp, #-16]! // 8-byte Folded Spill
; CHECK-NEXT: .cfi_def_cfa_offset 16
; CHECK-NEXT: .cfi_offset w30, -16
; CHECK-NEXT: mov x1, x0
; CHECK-NEXT: add x0, sp, #12
; CHECK-NEXT: bl sincosf
; CHECK-NEXT: ldr s0, [sp, #12]
; CHECK-NEXT: ldr x30, [sp], #16 // 8-byte Folded Reload
; CHECK-NEXT: ret
%sin = call float @llvm.sin.f32(float %x)
store float %sin, ptr %out, align 4
%reload = load float, ptr %out, align 4
%cos = call float @llvm.cos.f32(float %x)
store float %cos, ptr %out, align 4
ret float %reload
}
; Negative test. We can't fold volatile stores into the library call.
define void @sincos_volatile_result_stores(float %x, ptr noalias %out_sin, ptr noalias %out_cos) {
; CHECK-LABEL: sincos_volatile_result_stores:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: str x30, [sp, #-32]! // 8-byte Folded Spill
; CHECK-NEXT: stp x20, x19, [sp, #16] // 16-byte Folded Spill
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: .cfi_offset w19, -8
; CHECK-NEXT: .cfi_offset w20, -16
; CHECK-NEXT: .cfi_offset w30, -32
; CHECK-NEXT: mov x19, x1
; CHECK-NEXT: mov x20, x0
; CHECK-NEXT: add x0, sp, #12
; CHECK-NEXT: add x1, sp, #8
; CHECK-NEXT: bl sincosf
; CHECK-NEXT: ldp s1, s0, [sp, #8]
; CHECK-NEXT: str s0, [x20]
; CHECK-NEXT: str s1, [x19]
; CHECK-NEXT: ldp x20, x19, [sp, #16] // 16-byte Folded Reload
; CHECK-NEXT: ldr x30, [sp], #32 // 8-byte Folded Reload
; CHECK-NEXT: ret
entry:
%sin = tail call float @llvm.sin.f32(float %x)
%cos = tail call float @llvm.cos.f32(float %x)
store volatile float %sin, ptr %out_sin, align 4
store volatile float %cos, ptr %out_cos, align 4
ret void
}
; Negative test. We can't fold atomic stores into the library call.
define void @sincos_atomic_result_stores(float %x, ptr noalias %out_sin, ptr noalias %out_cos) {
; CHECK-LABEL: sincos_atomic_result_stores:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: str x30, [sp, #-32]! // 8-byte Folded Spill
; CHECK-NEXT: stp x20, x19, [sp, #16] // 16-byte Folded Spill
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: .cfi_offset w19, -8
; CHECK-NEXT: .cfi_offset w20, -16
; CHECK-NEXT: .cfi_offset w30, -32
; CHECK-NEXT: mov x19, x1
; CHECK-NEXT: mov x20, x0
; CHECK-NEXT: add x0, sp, #12
; CHECK-NEXT: add x1, sp, #8
; CHECK-NEXT: bl sincosf
; CHECK-NEXT: ldr w8, [sp, #12]
; CHECK-NEXT: str w8, [x20]
; CHECK-NEXT: ldr w8, [sp, #8]
; CHECK-NEXT: str w8, [x19]
; CHECK-NEXT: ldp x20, x19, [sp, #16] // 16-byte Folded Reload
; CHECK-NEXT: ldr x30, [sp], #32 // 8-byte Folded Reload
; CHECK-NEXT: ret
entry:
%sin = tail call float @llvm.sin.f32(float %x)
%cos = tail call float @llvm.cos.f32(float %x)
store atomic float %sin, ptr %out_sin unordered, align 4
store atomic float %cos, ptr %out_cos unordered, align 4
ret void
}
; Negative test. We can't fold misaligned stores into the library call.
define void @sincos_misaligned_result_stores(double %x, ptr noalias %out_sin, ptr noalias %out_cos) {
; CHECK-LABEL: sincos_misaligned_result_stores:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: sub sp, sp, #48
; CHECK-NEXT: str x30, [sp, #16] // 8-byte Folded Spill
; CHECK-NEXT: stp x20, x19, [sp, #32] // 16-byte Folded Spill
; CHECK-NEXT: .cfi_def_cfa_offset 48
; CHECK-NEXT: .cfi_offset w19, -8
; CHECK-NEXT: .cfi_offset w20, -16
; CHECK-NEXT: .cfi_offset w30, -32
; CHECK-NEXT: mov x19, x1
; CHECK-NEXT: mov x20, x0
; CHECK-NEXT: add x0, sp, #24
; CHECK-NEXT: add x1, sp, #8
; CHECK-NEXT: bl sincos
; CHECK-NEXT: ldr d0, [sp, #24]
; CHECK-NEXT: ldr d1, [sp, #8]
; CHECK-NEXT: ldr x30, [sp, #16] // 8-byte Folded Reload
; CHECK-NEXT: str d0, [x20]
; CHECK-NEXT: str d1, [x19]
; CHECK-NEXT: ldp x20, x19, [sp, #32] // 16-byte Folded Reload
; CHECK-NEXT: add sp, sp, #48
; CHECK-NEXT: ret
entry:
%sin = tail call double @llvm.sin.f64(double %x)
%cos = tail call double @llvm.cos.f64(double %x)
store double %sin, ptr %out_sin, align 4
store double %cos, ptr %out_cos, align 4
ret void
}
declare void @foo(ptr, ptr)
define void @can_fold_with_call_in_chain(float %x, ptr noalias %a, ptr noalias %b) {
; CHECK-LABEL: can_fold_with_call_in_chain:
; CHECK: // %bb.0: // %entry
; CHECK-NEXT: str d8, [sp, #-32]! // 8-byte Folded Spill
; CHECK-NEXT: str x30, [sp, #8] // 8-byte Folded Spill
; CHECK-NEXT: stp x20, x19, [sp, #16] // 16-byte Folded Spill
; CHECK-NEXT: .cfi_def_cfa_offset 32
; CHECK-NEXT: .cfi_offset w19, -8
; CHECK-NEXT: .cfi_offset w20, -16
; CHECK-NEXT: .cfi_offset w30, -24
; CHECK-NEXT: .cfi_offset b8, -32
; CHECK-NEXT: mov x19, x1
; CHECK-NEXT: mov x20, x0
; CHECK-NEXT: fmov s8, s0
; CHECK-NEXT: bl foo
; CHECK-NEXT: fmov s0, s8
; CHECK-NEXT: mov x0, x20
; CHECK-NEXT: mov x1, x19
; CHECK-NEXT: bl sincosf
; CHECK-NEXT: ldp x20, x19, [sp, #16] // 16-byte Folded Reload
; CHECK-NEXT: ldr x30, [sp, #8] // 8-byte Folded Reload
; CHECK-NEXT: ldr d8, [sp], #32 // 8-byte Folded Reload
; CHECK-NEXT: ret
entry:
%sin = tail call float @llvm.sin.f32(float %x)
%cos = tail call float @llvm.cos.f32(float %x)
call void @foo(ptr %a, ptr %b)
store float %sin, ptr %a, align 4
store float %cos, ptr %b, align 4
ret void
}
|
