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
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx,slow-unaligned-mem-32 | FileCheck %s
; RUN: llc -O0 < %s -mtriple=x86_64-unknown-unknown -mattr=avx,slow-unaligned-mem-32 | FileCheck %s -check-prefix=CHECK_O0
define void @test_256_load(double* nocapture %d, float* nocapture %f, <4 x i64>* nocapture %i) nounwind {
; CHECK-LABEL: test_256_load:
; CHECK: # %bb.0: # %entry
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: subq $96, %rsp
; CHECK-NEXT: movq %rdx, %r14
; CHECK-NEXT: movq %rsi, %r15
; CHECK-NEXT: movq %rdi, %rbx
; CHECK-NEXT: vmovaps (%rdi), %ymm0
; CHECK-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; CHECK-NEXT: vmovaps (%rsi), %ymm1
; CHECK-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; CHECK-NEXT: vmovaps (%rdx), %ymm2
; CHECK-NEXT: vmovups %ymm2, (%rsp) # 32-byte Spill
; CHECK-NEXT: callq dummy@PLT
; CHECK-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; CHECK-NEXT: vmovaps %ymm0, (%rbx)
; CHECK-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; CHECK-NEXT: vmovaps %ymm0, (%r15)
; CHECK-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
; CHECK-NEXT: vmovaps %ymm0, (%r14)
; CHECK-NEXT: addq $96, %rsp
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: popq %r14
; CHECK-NEXT: popq %r15
; CHECK-NEXT: vzeroupper
; CHECK-NEXT: retq
;
; CHECK_O0-LABEL: test_256_load:
; CHECK_O0: # %bb.0: # %entry
; CHECK_O0-NEXT: subq $184, %rsp
; CHECK_O0-NEXT: movq %rdi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK_O0-NEXT: movq %rsi, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK_O0-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK_O0-NEXT: vmovapd (%rdi), %ymm0
; CHECK_O0-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
; CHECK_O0-NEXT: vmovaps (%rsi), %ymm1
; CHECK_O0-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; CHECK_O0-NEXT: vmovdqa (%rdx), %ymm2
; CHECK_O0-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; CHECK_O0-NEXT: callq dummy@PLT
; CHECK_O0-NEXT: vmovups (%rsp), %ymm2 # 32-byte Reload
; CHECK_O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdi # 8-byte Reload
; CHECK_O0-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm1 # 32-byte Reload
; CHECK_O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rsi # 8-byte Reload
; CHECK_O0-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
; CHECK_O0-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rdx # 8-byte Reload
; CHECK_O0-NEXT: vmovapd %ymm2, (%rdi)
; CHECK_O0-NEXT: vmovaps %ymm1, (%rsi)
; CHECK_O0-NEXT: vmovdqa %ymm0, (%rdx)
; CHECK_O0-NEXT: addq $184, %rsp
; CHECK_O0-NEXT: vzeroupper
; CHECK_O0-NEXT: retq
entry:
%0 = bitcast double* %d to <4 x double>*
%tmp1.i = load <4 x double>, <4 x double>* %0, align 32
%1 = bitcast float* %f to <8 x float>*
%tmp1.i17 = load <8 x float>, <8 x float>* %1, align 32
%tmp1.i16 = load <4 x i64>, <4 x i64>* %i, align 32
tail call void @dummy(<4 x double> %tmp1.i, <8 x float> %tmp1.i17, <4 x i64> %tmp1.i16) nounwind
store <4 x double> %tmp1.i, <4 x double>* %0, align 32
store <8 x float> %tmp1.i17, <8 x float>* %1, align 32
store <4 x i64> %tmp1.i16, <4 x i64>* %i, align 32
ret void
}
declare void @dummy(<4 x double>, <8 x float>, <4 x i64>)
;;
;; The two tests below check that we must fold load + scalar_to_vector
;; + ins_subvec+ zext into only a single vmovss or vmovsd or vinsertps from memory
define <8 x float> @mov00(<8 x float> %v, float * %ptr) nounwind {
; CHECK-LABEL: mov00:
; CHECK: # %bb.0:
; CHECK-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; CHECK-NEXT: retq
;
; CHECK_O0-LABEL: mov00:
; CHECK_O0: # %bb.0:
; CHECK_O0-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; CHECK_O0-NEXT: # kill: def $ymm0 killed $xmm0
; CHECK_O0-NEXT: retq
%val = load float, float* %ptr
%i0 = insertelement <8 x float> zeroinitializer, float %val, i32 0
ret <8 x float> %i0
}
define <4 x double> @mov01(<4 x double> %v, double * %ptr) nounwind {
; CHECK-LABEL: mov01:
; CHECK: # %bb.0:
; CHECK-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; CHECK-NEXT: retq
;
; CHECK_O0-LABEL: mov01:
; CHECK_O0: # %bb.0:
; CHECK_O0-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; CHECK_O0-NEXT: # kill: def $ymm0 killed $xmm0
; CHECK_O0-NEXT: retq
%val = load double, double* %ptr
%i0 = insertelement <4 x double> zeroinitializer, double %val, i32 0
ret <4 x double> %i0
}
define void @storev16i16(<16 x i16> %a) nounwind {
; CHECK-LABEL: storev16i16:
; CHECK: # %bb.0:
; CHECK-NEXT: vmovaps %ymm0, (%rax)
;
; CHECK_O0-LABEL: storev16i16:
; CHECK_O0: # %bb.0:
; CHECK_O0-NEXT: # implicit-def: $rax
; CHECK_O0-NEXT: vmovdqa %ymm0, (%rax)
store <16 x i16> %a, <16 x i16>* undef, align 32
unreachable
}
define void @storev16i16_01(<16 x i16> %a) nounwind {
; CHECK-LABEL: storev16i16_01:
; CHECK: # %bb.0:
; CHECK-NEXT: vextractf128 $1, %ymm0, (%rax)
; CHECK-NEXT: vmovups %xmm0, (%rax)
;
; CHECK_O0-LABEL: storev16i16_01:
; CHECK_O0: # %bb.0:
; CHECK_O0-NEXT: # implicit-def: $rax
; CHECK_O0-NEXT: vmovdqu %ymm0, (%rax)
store <16 x i16> %a, <16 x i16>* undef, align 4
unreachable
}
define void @storev32i8(<32 x i8> %a) nounwind {
; CHECK-LABEL: storev32i8:
; CHECK: # %bb.0:
; CHECK-NEXT: vmovaps %ymm0, (%rax)
;
; CHECK_O0-LABEL: storev32i8:
; CHECK_O0: # %bb.0:
; CHECK_O0-NEXT: # implicit-def: $rax
; CHECK_O0-NEXT: vmovdqa %ymm0, (%rax)
store <32 x i8> %a, <32 x i8>* undef, align 32
unreachable
}
define void @storev32i8_01(<32 x i8> %a) nounwind {
; CHECK-LABEL: storev32i8_01:
; CHECK: # %bb.0:
; CHECK-NEXT: vextractf128 $1, %ymm0, (%rax)
; CHECK-NEXT: vmovups %xmm0, (%rax)
;
; CHECK_O0-LABEL: storev32i8_01:
; CHECK_O0: # %bb.0:
; CHECK_O0-NEXT: # implicit-def: $rax
; CHECK_O0-NEXT: vmovdqu %ymm0, (%rax)
store <32 x i8> %a, <32 x i8>* undef, align 4
unreachable
}
; It is faster to make two saves, if the data is already in xmm registers. For
; example, after making an integer operation.
define void @double_save(<4 x i32> %A, <4 x i32> %B, <8 x i32>* %P) nounwind ssp {
; CHECK-LABEL: double_save:
; CHECK: # %bb.0:
; CHECK-NEXT: vmovaps %xmm1, 16(%rdi)
; CHECK-NEXT: vmovaps %xmm0, (%rdi)
; CHECK-NEXT: retq
;
; CHECK_O0-LABEL: double_save:
; CHECK_O0: # %bb.0:
; CHECK_O0-NEXT: vmovaps %xmm0, %xmm2
; CHECK_O0-NEXT: # implicit-def: $ymm0
; CHECK_O0-NEXT: vmovaps %xmm2, %xmm0
; CHECK_O0-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
; CHECK_O0-NEXT: vmovdqu %ymm0, (%rdi)
; CHECK_O0-NEXT: vzeroupper
; CHECK_O0-NEXT: retq
%Z = shufflevector <4 x i32>%A, <4 x i32>%B, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
store <8 x i32> %Z, <8 x i32>* %P, align 16
ret void
}
define void @double_save_volatile(<4 x i32> %A, <4 x i32> %B, <8 x i32>* %P) nounwind {
; CHECK-LABEL: double_save_volatile:
; CHECK: # %bb.0:
; CHECK-NEXT: # kill: def $xmm0 killed $xmm0 def $ymm0
; CHECK-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
; CHECK-NEXT: vmovups %ymm0, (%rdi)
; CHECK-NEXT: vzeroupper
; CHECK-NEXT: retq
;
; CHECK_O0-LABEL: double_save_volatile:
; CHECK_O0: # %bb.0:
; CHECK_O0-NEXT: vmovaps %xmm0, %xmm2
; CHECK_O0-NEXT: # implicit-def: $ymm0
; CHECK_O0-NEXT: vmovaps %xmm2, %xmm0
; CHECK_O0-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
; CHECK_O0-NEXT: vmovdqu %ymm0, (%rdi)
; CHECK_O0-NEXT: vzeroupper
; CHECK_O0-NEXT: retq
%Z = shufflevector <4 x i32>%A, <4 x i32>%B, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
store volatile <8 x i32> %Z, <8 x i32>* %P, align 16
ret void
}
declare void @llvm.x86.avx.maskstore.ps.256(i8*, <8 x i32>, <8 x float>) nounwind
define void @f_f() nounwind {
; CHECK-LABEL: f_f:
; CHECK: # %bb.0: # %allocas
; CHECK-NEXT: xorl %eax, %eax
; CHECK-NEXT: testb %al, %al
; CHECK-NEXT: jne .LBB9_2
; CHECK-NEXT: # %bb.1: # %cif_mask_all
; CHECK-NEXT: .LBB9_2: # %cif_mask_mixed
; CHECK-NEXT: xorl %eax, %eax
; CHECK-NEXT: testb %al, %al
; CHECK-NEXT: jne .LBB9_4
; CHECK-NEXT: # %bb.3: # %cif_mixed_test_all
; CHECK-NEXT: vmovaps {{.*#+}} xmm0 = [4294967295,0,0,0]
; CHECK-NEXT: vmaskmovps %ymm0, %ymm0, (%rax)
; CHECK-NEXT: .LBB9_4: # %cif_mixed_test_any_check
;
; CHECK_O0-LABEL: f_f:
; CHECK_O0: # %bb.0: # %allocas
; CHECK_O0-NEXT: # implicit-def: $al
; CHECK_O0-NEXT: testb $1, %al
; CHECK_O0-NEXT: jne .LBB9_1
; CHECK_O0-NEXT: jmp .LBB9_2
; CHECK_O0-NEXT: .LBB9_1: # %cif_mask_all
; CHECK_O0-NEXT: .LBB9_2: # %cif_mask_mixed
; CHECK_O0-NEXT: # implicit-def: $al
; CHECK_O0-NEXT: testb $1, %al
; CHECK_O0-NEXT: jne .LBB9_3
; CHECK_O0-NEXT: jmp .LBB9_4
; CHECK_O0-NEXT: .LBB9_3: # %cif_mixed_test_all
; CHECK_O0-NEXT: vmovdqa {{.*#+}} xmm0 = [4294967295,0,0,0]
; CHECK_O0-NEXT: vmovdqa %xmm0, %xmm0
; CHECK_O0-NEXT: # kill: def $ymm0 killed $xmm0
; CHECK_O0-NEXT: # implicit-def: $rax
; CHECK_O0-NEXT: # implicit-def: $ymm1
; CHECK_O0-NEXT: vmaskmovps %ymm1, %ymm0, (%rax)
; CHECK_O0-NEXT: .LBB9_4: # %cif_mixed_test_any_check
allocas:
br i1 undef, label %cif_mask_all, label %cif_mask_mixed
cif_mask_all:
unreachable
cif_mask_mixed:
br i1 undef, label %cif_mixed_test_all, label %cif_mixed_test_any_check
cif_mixed_test_all:
call void @llvm.x86.avx.maskstore.ps.256(i8* undef, <8 x i32> <i32 -1, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>, <8 x float> undef) nounwind
unreachable
cif_mixed_test_any_check:
unreachable
}
define void @add8i32(<8 x i32>* %ret, <8 x i32>* %bp) nounwind {
; CHECK-LABEL: add8i32:
; CHECK: # %bb.0:
; CHECK-NEXT: vmovups (%rsi), %xmm0
; CHECK-NEXT: vmovups 16(%rsi), %xmm1
; CHECK-NEXT: vmovups %xmm1, 16(%rdi)
; CHECK-NEXT: vmovups %xmm0, (%rdi)
; CHECK-NEXT: retq
;
; CHECK_O0-LABEL: add8i32:
; CHECK_O0: # %bb.0:
; CHECK_O0-NEXT: vmovdqu (%rsi), %xmm2
; CHECK_O0-NEXT: vmovdqu 16(%rsi), %xmm1
; CHECK_O0-NEXT: # implicit-def: $ymm0
; CHECK_O0-NEXT: vmovaps %xmm2, %xmm0
; CHECK_O0-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
; CHECK_O0-NEXT: vmovdqu %ymm0, (%rdi)
; CHECK_O0-NEXT: vzeroupper
; CHECK_O0-NEXT: retq
%b = load <8 x i32>, <8 x i32>* %bp, align 1
%x = add <8 x i32> zeroinitializer, %b
store <8 x i32> %x, <8 x i32>* %ret, align 1
ret void
}
define void @add4i64a64(<4 x i64>* %ret, <4 x i64>* %bp) nounwind {
; CHECK-LABEL: add4i64a64:
; CHECK: # %bb.0:
; CHECK-NEXT: vmovaps (%rsi), %ymm0
; CHECK-NEXT: vmovaps %ymm0, (%rdi)
; CHECK-NEXT: vzeroupper
; CHECK-NEXT: retq
;
; CHECK_O0-LABEL: add4i64a64:
; CHECK_O0: # %bb.0:
; CHECK_O0-NEXT: vmovaps (%rsi), %ymm0
; CHECK_O0-NEXT: vmovdqa %ymm0, (%rdi)
; CHECK_O0-NEXT: vzeroupper
; CHECK_O0-NEXT: retq
%b = load <4 x i64>, <4 x i64>* %bp, align 64
%x = add <4 x i64> zeroinitializer, %b
store <4 x i64> %x, <4 x i64>* %ret, align 64
ret void
}
define void @add4i64a16(<4 x i64>* %ret, <4 x i64>* %bp) nounwind {
; CHECK-LABEL: add4i64a16:
; CHECK: # %bb.0:
; CHECK-NEXT: vmovaps (%rsi), %xmm0
; CHECK-NEXT: vmovaps 16(%rsi), %xmm1
; CHECK-NEXT: vmovaps %xmm1, 16(%rdi)
; CHECK-NEXT: vmovaps %xmm0, (%rdi)
; CHECK-NEXT: retq
;
; CHECK_O0-LABEL: add4i64a16:
; CHECK_O0: # %bb.0:
; CHECK_O0-NEXT: vmovdqa (%rsi), %xmm2
; CHECK_O0-NEXT: vmovdqa 16(%rsi), %xmm1
; CHECK_O0-NEXT: # implicit-def: $ymm0
; CHECK_O0-NEXT: vmovaps %xmm2, %xmm0
; CHECK_O0-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
; CHECK_O0-NEXT: vmovdqu %ymm0, (%rdi)
; CHECK_O0-NEXT: vzeroupper
; CHECK_O0-NEXT: retq
%b = load <4 x i64>, <4 x i64>* %bp, align 16
%x = add <4 x i64> zeroinitializer, %b
store <4 x i64> %x, <4 x i64>* %ret, align 16
ret void
}
; This used to crash.
; v2i128 may not be a "simple" (MVT) type, but we can split that.
; This example gets split further in legalization.
define void @PR43916(<2 x i128> %y, <2 x i128>* %z) {
; CHECK-LABEL: PR43916:
; CHECK: # %bb.0:
; CHECK-NEXT: movq %rcx, 24(%r8)
; CHECK-NEXT: movq %rdx, 16(%r8)
; CHECK-NEXT: movq %rsi, 8(%r8)
; CHECK-NEXT: movq %rdi, (%r8)
; CHECK-NEXT: retq
;
; CHECK_O0-LABEL: PR43916:
; CHECK_O0: # %bb.0:
; CHECK_O0-NEXT: movq %rdi, (%r8)
; CHECK_O0-NEXT: movq %rsi, 8(%r8)
; CHECK_O0-NEXT: movq %rdx, 16(%r8)
; CHECK_O0-NEXT: movq %rcx, 24(%r8)
; CHECK_O0-NEXT: retq
store <2 x i128> %y, <2 x i128>* %z, align 16
ret void
}
|
