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
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
|
// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
// RUN: %clang_cc1 \
// RUN: -triple aarch64-arm-none-eabi -target-feature +neon -target-feature +bf16 \
// RUN: -disable-O0-optnone -emit-llvm -fno-legacy-pass-manager -o - %s \
// RUN: | opt -S -mem2reg \
// RUN: | FileCheck --check-prefixes=CHECK,CHECK-A64 %s
// RUN: %clang_cc1 \
// RUN: -triple armv8.6a-arm-none-eabi -target-feature +neon \
// RUN: -target-feature +bf16 -mfloat-abi hard \
// RUN: -disable-O0-optnone -emit-llvm -fno-legacy-pass-manager -o - %s \
// RUN: | opt -S -mem2reg \
// RUN: | FileCheck --check-prefixes=CHECK,CHECK-A32-HARDFP %s
// RUN: %clang_cc1 \
// RUN: -triple armv8.6a-arm-none-eabi -target-feature +neon \
// RUN: -target-feature +bf16 -mfloat-abi softfp \
// RUN: -disable-O0-optnone -emit-llvm -fno-legacy-pass-manager -o - %s \
// RUN: | opt -S -mem2reg \
// RUN: | FileCheck --check-prefixes=CHECK,CHECK-A32-SOFTFP %s
// REQUIRES: arm-registered-target
// REQUIRES: aarch64-registered-target
#include <arm_neon.h>
// CHECK-A64-LABEL: @test_vcvt_f32_bf16(
// CHECK-A64-NEXT: entry:
// CHECK-A64-NEXT: [[__REINT_150_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A64-NEXT: [[__REINT1_150_I:%.*]] = alloca <4 x i32>, align 16
// CHECK-A64-NEXT: store <4 x bfloat> [[A:%.*]], <4 x bfloat>* [[__REINT_150_I]], align 8
// CHECK-A64-NEXT: [[TMP0:%.*]] = bitcast <4 x bfloat>* [[__REINT_150_I]] to <4 x i16>*
// CHECK-A64-NEXT: [[TMP1:%.*]] = load <4 x i16>, <4 x i16>* [[TMP0]], align 8
// CHECK-A64-NEXT: [[TMP2:%.*]] = bitcast <4 x i16> [[TMP1]] to <8 x i8>
// CHECK-A64-NEXT: [[TMP3:%.*]] = sext <4 x i16> [[TMP1]] to <4 x i32>
// CHECK-A64-NEXT: [[VSHLL_N_I:%.*]] = shl <4 x i32> [[TMP3]], <i32 16, i32 16, i32 16, i32 16>
// CHECK-A64-NEXT: store <4 x i32> [[VSHLL_N_I]], <4 x i32>* [[__REINT1_150_I]], align 16
// CHECK-A64-NEXT: [[TMP4:%.*]] = bitcast <4 x i32>* [[__REINT1_150_I]] to <4 x float>*
// CHECK-A64-NEXT: [[TMP5:%.*]] = load <4 x float>, <4 x float>* [[TMP4]], align 16
// CHECK-A64-NEXT: ret <4 x float> [[TMP5]]
//
// CHECK-A32-HARDFP-LABEL: @test_vcvt_f32_bf16(
// CHECK-A32-HARDFP-NEXT: entry:
// CHECK-A32-HARDFP-NEXT: [[__REINT_150_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-HARDFP-NEXT: [[__REINT1_150_I:%.*]] = alloca <4 x i32>, align 8
// CHECK-A32-HARDFP-NEXT: store <4 x bfloat> [[A:%.*]], <4 x bfloat>* [[__REINT_150_I]], align 8
// CHECK-A32-HARDFP-NEXT: [[TMP0:%.*]] = bitcast <4 x bfloat>* [[__REINT_150_I]] to <4 x i16>*
// CHECK-A32-HARDFP-NEXT: [[TMP1:%.*]] = load <4 x i16>, <4 x i16>* [[TMP0]], align 8
// CHECK-A32-HARDFP-NEXT: [[TMP2:%.*]] = bitcast <4 x i16> [[TMP1]] to <8 x i8>
// CHECK-A32-HARDFP-NEXT: [[TMP3:%.*]] = sext <4 x i16> [[TMP1]] to <4 x i32>
// CHECK-A32-HARDFP-NEXT: [[VSHLL_N_I:%.*]] = shl <4 x i32> [[TMP3]], <i32 16, i32 16, i32 16, i32 16>
// CHECK-A32-HARDFP-NEXT: store <4 x i32> [[VSHLL_N_I]], <4 x i32>* [[__REINT1_150_I]], align 8
// CHECK-A32-HARDFP-NEXT: [[TMP4:%.*]] = bitcast <4 x i32>* [[__REINT1_150_I]] to <4 x float>*
// CHECK-A32-HARDFP-NEXT: [[TMP5:%.*]] = load <4 x float>, <4 x float>* [[TMP4]], align 8
// CHECK-A32-HARDFP-NEXT: ret <4 x float> [[TMP5]]
//
// CHECK-A32-SOFTFP-LABEL: @test_vcvt_f32_bf16(
// CHECK-A32-SOFTFP-NEXT: entry:
// CHECK-A32-SOFTFP-NEXT: [[__P0_150_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[__REINT_150_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[__REINT1_150_I:%.*]] = alloca <4 x i32>, align 8
// CHECK-A32-SOFTFP-NEXT: [[A:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[COERCE:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP0:%.*]] = bitcast <4 x bfloat>* [[A]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: store <2 x i32> [[A_COERCE:%.*]], <2 x i32>* [[TMP0]], align 8
// CHECK-A32-SOFTFP-NEXT: [[A1:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[A]], align 8
// CHECK-A32-SOFTFP-NEXT: store <4 x bfloat> [[A1]], <4 x bfloat>* [[COERCE]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP1:%.*]] = bitcast <4 x bfloat>* [[COERCE]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: [[TMP2:%.*]] = load <2 x i32>, <2 x i32>* [[TMP1]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP3:%.*]] = bitcast <4 x bfloat>* [[__P0_150_I]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: store <2 x i32> [[TMP2]], <2 x i32>* [[TMP3]], align 8
// CHECK-A32-SOFTFP-NEXT: [[__P0_1501_I:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[__P0_150_I]], align 8
// CHECK-A32-SOFTFP-NEXT: store <4 x bfloat> [[__P0_1501_I]], <4 x bfloat>* [[__REINT_150_I]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP4:%.*]] = bitcast <4 x bfloat>* [[__REINT_150_I]] to <4 x i16>*
// CHECK-A32-SOFTFP-NEXT: [[TMP5:%.*]] = load <4 x i16>, <4 x i16>* [[TMP4]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP6:%.*]] = bitcast <4 x i16> [[TMP5]] to <8 x i8>
// CHECK-A32-SOFTFP-NEXT: [[TMP7:%.*]] = sext <4 x i16> [[TMP5]] to <4 x i32>
// CHECK-A32-SOFTFP-NEXT: [[VSHLL_N_I:%.*]] = shl <4 x i32> [[TMP7]], <i32 16, i32 16, i32 16, i32 16>
// CHECK-A32-SOFTFP-NEXT: store <4 x i32> [[VSHLL_N_I]], <4 x i32>* [[__REINT1_150_I]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP8:%.*]] = bitcast <4 x i32>* [[__REINT1_150_I]] to <4 x float>*
// CHECK-A32-SOFTFP-NEXT: [[TMP9:%.*]] = load <4 x float>, <4 x float>* [[TMP8]], align 8
// CHECK-A32-SOFTFP-NEXT: ret <4 x float> [[TMP9]]
//
float32x4_t test_vcvt_f32_bf16(bfloat16x4_t a) {
return vcvt_f32_bf16(a);
}
// CHECK-A64-LABEL: @test_vcvtq_low_f32_bf16(
// CHECK-A64-NEXT: entry:
// CHECK-A64-NEXT: [[__REINT_150_I_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A64-NEXT: [[__REINT1_150_I_I:%.*]] = alloca <4 x i32>, align 16
// CHECK-A64-NEXT: [[SHUFFLE_I:%.*]] = shufflevector <8 x bfloat> [[A:%.*]], <8 x bfloat> [[A]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
// CHECK-A64-NEXT: store <4 x bfloat> [[SHUFFLE_I]], <4 x bfloat>* [[__REINT_150_I_I]], align 8
// CHECK-A64-NEXT: [[TMP0:%.*]] = bitcast <4 x bfloat>* [[__REINT_150_I_I]] to <4 x i16>*
// CHECK-A64-NEXT: [[TMP1:%.*]] = load <4 x i16>, <4 x i16>* [[TMP0]], align 8
// CHECK-A64-NEXT: [[TMP2:%.*]] = bitcast <4 x i16> [[TMP1]] to <8 x i8>
// CHECK-A64-NEXT: [[TMP3:%.*]] = sext <4 x i16> [[TMP1]] to <4 x i32>
// CHECK-A64-NEXT: [[VSHLL_N_I_I:%.*]] = shl <4 x i32> [[TMP3]], <i32 16, i32 16, i32 16, i32 16>
// CHECK-A64-NEXT: store <4 x i32> [[VSHLL_N_I_I]], <4 x i32>* [[__REINT1_150_I_I]], align 16
// CHECK-A64-NEXT: [[TMP4:%.*]] = bitcast <4 x i32>* [[__REINT1_150_I_I]] to <4 x float>*
// CHECK-A64-NEXT: [[TMP5:%.*]] = load <4 x float>, <4 x float>* [[TMP4]], align 16
// CHECK-A64-NEXT: ret <4 x float> [[TMP5]]
//
// CHECK-A32-HARDFP-LABEL: @test_vcvtq_low_f32_bf16(
// CHECK-A32-HARDFP-NEXT: entry:
// CHECK-A32-HARDFP-NEXT: [[__REINT_150_I_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-HARDFP-NEXT: [[__REINT1_150_I_I:%.*]] = alloca <4 x i32>, align 8
// CHECK-A32-HARDFP-NEXT: [[SHUFFLE_I:%.*]] = shufflevector <8 x bfloat> [[A:%.*]], <8 x bfloat> [[A]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
// CHECK-A32-HARDFP-NEXT: store <4 x bfloat> [[SHUFFLE_I]], <4 x bfloat>* [[__REINT_150_I_I]], align 8
// CHECK-A32-HARDFP-NEXT: [[TMP0:%.*]] = bitcast <4 x bfloat>* [[__REINT_150_I_I]] to <4 x i16>*
// CHECK-A32-HARDFP-NEXT: [[TMP1:%.*]] = load <4 x i16>, <4 x i16>* [[TMP0]], align 8
// CHECK-A32-HARDFP-NEXT: [[TMP2:%.*]] = bitcast <4 x i16> [[TMP1]] to <8 x i8>
// CHECK-A32-HARDFP-NEXT: [[TMP3:%.*]] = sext <4 x i16> [[TMP1]] to <4 x i32>
// CHECK-A32-HARDFP-NEXT: [[VSHLL_N_I_I:%.*]] = shl <4 x i32> [[TMP3]], <i32 16, i32 16, i32 16, i32 16>
// CHECK-A32-HARDFP-NEXT: store <4 x i32> [[VSHLL_N_I_I]], <4 x i32>* [[__REINT1_150_I_I]], align 8
// CHECK-A32-HARDFP-NEXT: [[TMP4:%.*]] = bitcast <4 x i32>* [[__REINT1_150_I_I]] to <4 x float>*
// CHECK-A32-HARDFP-NEXT: [[TMP5:%.*]] = load <4 x float>, <4 x float>* [[TMP4]], align 8
// CHECK-A32-HARDFP-NEXT: ret <4 x float> [[TMP5]]
//
// CHECK-A32-SOFTFP-LABEL: @test_vcvtq_low_f32_bf16(
// CHECK-A32-SOFTFP-NEXT: entry:
// CHECK-A32-SOFTFP-NEXT: [[RETVAL_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[__P0_I2:%.*]] = alloca <8 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[__P0_150_I_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[__REINT_150_I_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[__REINT1_150_I_I:%.*]] = alloca <4 x i32>, align 8
// CHECK-A32-SOFTFP-NEXT: [[__P0_I:%.*]] = alloca <8 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[COERCE_I:%.*]] = alloca <8 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[COERCE2_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[COERCE3_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[A:%.*]] = alloca <8 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[COERCE:%.*]] = alloca <8 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP0:%.*]] = bitcast <8 x bfloat>* [[A]] to <4 x i32>*
// CHECK-A32-SOFTFP-NEXT: store <4 x i32> [[A_COERCE:%.*]], <4 x i32>* [[TMP0]], align 8
// CHECK-A32-SOFTFP-NEXT: [[A1:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[A]], align 8
// CHECK-A32-SOFTFP-NEXT: store <8 x bfloat> [[A1]], <8 x bfloat>* [[COERCE]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP1:%.*]] = bitcast <8 x bfloat>* [[COERCE]] to <4 x i32>*
// CHECK-A32-SOFTFP-NEXT: [[TMP2:%.*]] = load <4 x i32>, <4 x i32>* [[TMP1]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP3:%.*]] = bitcast <8 x bfloat>* [[__P0_I]] to <4 x i32>*
// CHECK-A32-SOFTFP-NEXT: store <4 x i32> [[TMP2]], <4 x i32>* [[TMP3]], align 8
// CHECK-A32-SOFTFP-NEXT: [[__P01_I:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[__P0_I]], align 8
// CHECK-A32-SOFTFP-NEXT: store <8 x bfloat> [[__P01_I]], <8 x bfloat>* [[COERCE_I]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP4:%.*]] = bitcast <8 x bfloat>* [[COERCE_I]] to <4 x i32>*
// CHECK-A32-SOFTFP-NEXT: [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* [[TMP4]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP6:%.*]] = bitcast <8 x bfloat>* [[__P0_I2]] to <4 x i32>*
// CHECK-A32-SOFTFP-NEXT: store <4 x i32> [[TMP5]], <4 x i32>* [[TMP6]], align 8
// CHECK-A32-SOFTFP-NEXT: [[__P01_I5:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[__P0_I2]], align 8
// CHECK-A32-SOFTFP-NEXT: [[SHUFFLE_I:%.*]] = shufflevector <8 x bfloat> [[__P01_I5]], <8 x bfloat> [[__P01_I5]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
// CHECK-A32-SOFTFP-NEXT: store <4 x bfloat> [[SHUFFLE_I]], <4 x bfloat>* [[RETVAL_I]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP7:%.*]] = bitcast <4 x bfloat>* [[RETVAL_I]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: [[TMP8:%.*]] = load <2 x i32>, <2 x i32>* [[TMP7]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP9:%.*]] = bitcast <4 x bfloat>* [[COERCE2_I]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: store <2 x i32> [[TMP8]], <2 x i32>* [[TMP9]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP10:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[COERCE2_I]], align 8
// CHECK-A32-SOFTFP-NEXT: store <4 x bfloat> [[TMP10]], <4 x bfloat>* [[COERCE3_I]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP11:%.*]] = bitcast <4 x bfloat>* [[COERCE3_I]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: [[TMP12:%.*]] = load <2 x i32>, <2 x i32>* [[TMP11]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP13:%.*]] = bitcast <4 x bfloat>* [[__P0_150_I_I]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: store <2 x i32> [[TMP12]], <2 x i32>* [[TMP13]], align 8
// CHECK-A32-SOFTFP-NEXT: [[__P0_1501_I_I:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[__P0_150_I_I]], align 8
// CHECK-A32-SOFTFP-NEXT: store <4 x bfloat> [[__P0_1501_I_I]], <4 x bfloat>* [[__REINT_150_I_I]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP14:%.*]] = bitcast <4 x bfloat>* [[__REINT_150_I_I]] to <4 x i16>*
// CHECK-A32-SOFTFP-NEXT: [[TMP15:%.*]] = load <4 x i16>, <4 x i16>* [[TMP14]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP16:%.*]] = bitcast <4 x i16> [[TMP15]] to <8 x i8>
// CHECK-A32-SOFTFP-NEXT: [[TMP17:%.*]] = sext <4 x i16> [[TMP15]] to <4 x i32>
// CHECK-A32-SOFTFP-NEXT: [[VSHLL_N_I_I:%.*]] = shl <4 x i32> [[TMP17]], <i32 16, i32 16, i32 16, i32 16>
// CHECK-A32-SOFTFP-NEXT: store <4 x i32> [[VSHLL_N_I_I]], <4 x i32>* [[__REINT1_150_I_I]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP18:%.*]] = bitcast <4 x i32>* [[__REINT1_150_I_I]] to <4 x float>*
// CHECK-A32-SOFTFP-NEXT: [[TMP19:%.*]] = load <4 x float>, <4 x float>* [[TMP18]], align 8
// CHECK-A32-SOFTFP-NEXT: ret <4 x float> [[TMP19]]
//
float32x4_t test_vcvtq_low_f32_bf16(bfloat16x8_t a) {
return vcvtq_low_f32_bf16(a);
}
// CHECK-A64-LABEL: @test_vcvtq_high_f32_bf16(
// CHECK-A64-NEXT: entry:
// CHECK-A64-NEXT: [[__REINT_150_I_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A64-NEXT: [[__REINT1_150_I_I:%.*]] = alloca <4 x i32>, align 16
// CHECK-A64-NEXT: [[SHUFFLE_I:%.*]] = shufflevector <8 x bfloat> [[A:%.*]], <8 x bfloat> [[A]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
// CHECK-A64-NEXT: store <4 x bfloat> [[SHUFFLE_I]], <4 x bfloat>* [[__REINT_150_I_I]], align 8
// CHECK-A64-NEXT: [[TMP0:%.*]] = bitcast <4 x bfloat>* [[__REINT_150_I_I]] to <4 x i16>*
// CHECK-A64-NEXT: [[TMP1:%.*]] = load <4 x i16>, <4 x i16>* [[TMP0]], align 8
// CHECK-A64-NEXT: [[TMP2:%.*]] = bitcast <4 x i16> [[TMP1]] to <8 x i8>
// CHECK-A64-NEXT: [[TMP3:%.*]] = sext <4 x i16> [[TMP1]] to <4 x i32>
// CHECK-A64-NEXT: [[VSHLL_N_I_I:%.*]] = shl <4 x i32> [[TMP3]], <i32 16, i32 16, i32 16, i32 16>
// CHECK-A64-NEXT: store <4 x i32> [[VSHLL_N_I_I]], <4 x i32>* [[__REINT1_150_I_I]], align 16
// CHECK-A64-NEXT: [[TMP4:%.*]] = bitcast <4 x i32>* [[__REINT1_150_I_I]] to <4 x float>*
// CHECK-A64-NEXT: [[TMP5:%.*]] = load <4 x float>, <4 x float>* [[TMP4]], align 16
// CHECK-A64-NEXT: ret <4 x float> [[TMP5]]
//
// CHECK-A32-HARDFP-LABEL: @test_vcvtq_high_f32_bf16(
// CHECK-A32-HARDFP-NEXT: entry:
// CHECK-A32-HARDFP-NEXT: [[__REINT_150_I_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-HARDFP-NEXT: [[__REINT1_150_I_I:%.*]] = alloca <4 x i32>, align 8
// CHECK-A32-HARDFP-NEXT: [[SHUFFLE_I:%.*]] = shufflevector <8 x bfloat> [[A:%.*]], <8 x bfloat> [[A]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
// CHECK-A32-HARDFP-NEXT: store <4 x bfloat> [[SHUFFLE_I]], <4 x bfloat>* [[__REINT_150_I_I]], align 8
// CHECK-A32-HARDFP-NEXT: [[TMP0:%.*]] = bitcast <4 x bfloat>* [[__REINT_150_I_I]] to <4 x i16>*
// CHECK-A32-HARDFP-NEXT: [[TMP1:%.*]] = load <4 x i16>, <4 x i16>* [[TMP0]], align 8
// CHECK-A32-HARDFP-NEXT: [[TMP2:%.*]] = bitcast <4 x i16> [[TMP1]] to <8 x i8>
// CHECK-A32-HARDFP-NEXT: [[TMP3:%.*]] = sext <4 x i16> [[TMP1]] to <4 x i32>
// CHECK-A32-HARDFP-NEXT: [[VSHLL_N_I_I:%.*]] = shl <4 x i32> [[TMP3]], <i32 16, i32 16, i32 16, i32 16>
// CHECK-A32-HARDFP-NEXT: store <4 x i32> [[VSHLL_N_I_I]], <4 x i32>* [[__REINT1_150_I_I]], align 8
// CHECK-A32-HARDFP-NEXT: [[TMP4:%.*]] = bitcast <4 x i32>* [[__REINT1_150_I_I]] to <4 x float>*
// CHECK-A32-HARDFP-NEXT: [[TMP5:%.*]] = load <4 x float>, <4 x float>* [[TMP4]], align 8
// CHECK-A32-HARDFP-NEXT: ret <4 x float> [[TMP5]]
//
// CHECK-A32-SOFTFP-LABEL: @test_vcvtq_high_f32_bf16(
// CHECK-A32-SOFTFP-NEXT: entry:
// CHECK-A32-SOFTFP-NEXT: [[RETVAL_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[__P0_I2:%.*]] = alloca <8 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[__P0_150_I_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[__REINT_150_I_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[__REINT1_150_I_I:%.*]] = alloca <4 x i32>, align 8
// CHECK-A32-SOFTFP-NEXT: [[__P0_I:%.*]] = alloca <8 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[COERCE_I:%.*]] = alloca <8 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[COERCE2_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[COERCE3_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[A:%.*]] = alloca <8 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[COERCE:%.*]] = alloca <8 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP0:%.*]] = bitcast <8 x bfloat>* [[A]] to <4 x i32>*
// CHECK-A32-SOFTFP-NEXT: store <4 x i32> [[A_COERCE:%.*]], <4 x i32>* [[TMP0]], align 8
// CHECK-A32-SOFTFP-NEXT: [[A1:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[A]], align 8
// CHECK-A32-SOFTFP-NEXT: store <8 x bfloat> [[A1]], <8 x bfloat>* [[COERCE]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP1:%.*]] = bitcast <8 x bfloat>* [[COERCE]] to <4 x i32>*
// CHECK-A32-SOFTFP-NEXT: [[TMP2:%.*]] = load <4 x i32>, <4 x i32>* [[TMP1]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP3:%.*]] = bitcast <8 x bfloat>* [[__P0_I]] to <4 x i32>*
// CHECK-A32-SOFTFP-NEXT: store <4 x i32> [[TMP2]], <4 x i32>* [[TMP3]], align 8
// CHECK-A32-SOFTFP-NEXT: [[__P01_I:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[__P0_I]], align 8
// CHECK-A32-SOFTFP-NEXT: store <8 x bfloat> [[__P01_I]], <8 x bfloat>* [[COERCE_I]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP4:%.*]] = bitcast <8 x bfloat>* [[COERCE_I]] to <4 x i32>*
// CHECK-A32-SOFTFP-NEXT: [[TMP5:%.*]] = load <4 x i32>, <4 x i32>* [[TMP4]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP6:%.*]] = bitcast <8 x bfloat>* [[__P0_I2]] to <4 x i32>*
// CHECK-A32-SOFTFP-NEXT: store <4 x i32> [[TMP5]], <4 x i32>* [[TMP6]], align 8
// CHECK-A32-SOFTFP-NEXT: [[__P01_I5:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[__P0_I2]], align 8
// CHECK-A32-SOFTFP-NEXT: [[SHUFFLE_I:%.*]] = shufflevector <8 x bfloat> [[__P01_I5]], <8 x bfloat> [[__P01_I5]], <4 x i32> <i32 4, i32 5, i32 6, i32 7>
// CHECK-A32-SOFTFP-NEXT: store <4 x bfloat> [[SHUFFLE_I]], <4 x bfloat>* [[RETVAL_I]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP7:%.*]] = bitcast <4 x bfloat>* [[RETVAL_I]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: [[TMP8:%.*]] = load <2 x i32>, <2 x i32>* [[TMP7]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP9:%.*]] = bitcast <4 x bfloat>* [[COERCE2_I]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: store <2 x i32> [[TMP8]], <2 x i32>* [[TMP9]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP10:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[COERCE2_I]], align 8
// CHECK-A32-SOFTFP-NEXT: store <4 x bfloat> [[TMP10]], <4 x bfloat>* [[COERCE3_I]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP11:%.*]] = bitcast <4 x bfloat>* [[COERCE3_I]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: [[TMP12:%.*]] = load <2 x i32>, <2 x i32>* [[TMP11]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP13:%.*]] = bitcast <4 x bfloat>* [[__P0_150_I_I]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: store <2 x i32> [[TMP12]], <2 x i32>* [[TMP13]], align 8
// CHECK-A32-SOFTFP-NEXT: [[__P0_1501_I_I:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[__P0_150_I_I]], align 8
// CHECK-A32-SOFTFP-NEXT: store <4 x bfloat> [[__P0_1501_I_I]], <4 x bfloat>* [[__REINT_150_I_I]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP14:%.*]] = bitcast <4 x bfloat>* [[__REINT_150_I_I]] to <4 x i16>*
// CHECK-A32-SOFTFP-NEXT: [[TMP15:%.*]] = load <4 x i16>, <4 x i16>* [[TMP14]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP16:%.*]] = bitcast <4 x i16> [[TMP15]] to <8 x i8>
// CHECK-A32-SOFTFP-NEXT: [[TMP17:%.*]] = sext <4 x i16> [[TMP15]] to <4 x i32>
// CHECK-A32-SOFTFP-NEXT: [[VSHLL_N_I_I:%.*]] = shl <4 x i32> [[TMP17]], <i32 16, i32 16, i32 16, i32 16>
// CHECK-A32-SOFTFP-NEXT: store <4 x i32> [[VSHLL_N_I_I]], <4 x i32>* [[__REINT1_150_I_I]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP18:%.*]] = bitcast <4 x i32>* [[__REINT1_150_I_I]] to <4 x float>*
// CHECK-A32-SOFTFP-NEXT: [[TMP19:%.*]] = load <4 x float>, <4 x float>* [[TMP18]], align 8
// CHECK-A32-SOFTFP-NEXT: ret <4 x float> [[TMP19]]
//
float32x4_t test_vcvtq_high_f32_bf16(bfloat16x8_t a) {
return vcvtq_high_f32_bf16(a);
}
// CHECK-A64-LABEL: @test_vcvt_bf16_f32(
// CHECK-A64-NEXT: entry:
// CHECK-A64-NEXT: [[TMP0:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
// CHECK-A64-NEXT: [[__A64_VCVTQ_LOW_BF16_V1_I:%.*]] = call <8 x bfloat> @llvm.aarch64.neon.bfcvtn(<4 x float> [[A]]) #[[ATTR3:[0-9]+]]
// CHECK-A64-NEXT: [[__A64_VCVTQ_LOW_BF16_V2_I:%.*]] = bitcast <8 x bfloat> [[__A64_VCVTQ_LOW_BF16_V1_I]] to <16 x i8>
// CHECK-A64-NEXT: [[SHUFFLE_I:%.*]] = shufflevector <8 x bfloat> [[__A64_VCVTQ_LOW_BF16_V1_I]], <8 x bfloat> [[__A64_VCVTQ_LOW_BF16_V1_I]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
// CHECK-A64-NEXT: ret <4 x bfloat> [[SHUFFLE_I]]
//
// CHECK-A32-HARDFP-LABEL: @test_vcvt_bf16_f32(
// CHECK-A32-HARDFP-NEXT: entry:
// CHECK-A32-HARDFP-NEXT: [[TMP0:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
// CHECK-A32-HARDFP-NEXT: [[VCVTFP2BF1_I:%.*]] = call <4 x bfloat> @llvm.arm.neon.vcvtfp2bf.v4bf16(<4 x float> [[A]]) #[[ATTR3:[0-9]+]]
// CHECK-A32-HARDFP-NEXT: ret <4 x bfloat> [[VCVTFP2BF1_I]]
//
// CHECK-A32-SOFTFP-LABEL: @test_vcvt_bf16_f32(
// CHECK-A32-SOFTFP-NEXT: entry:
// CHECK-A32-SOFTFP-NEXT: [[RETVAL_I1:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[RETVAL_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[COERCE_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[RETVAL:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[COERCE:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP0:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
// CHECK-A32-SOFTFP-NEXT: [[VCVTFP2BF1_I:%.*]] = call <4 x i16> @llvm.arm.neon.vcvtfp2bf.v4i16(<4 x float> [[A]]) #[[ATTR3:[0-9]+]]
// CHECK-A32-SOFTFP-NEXT: [[TMP1:%.*]] = bitcast <4 x i16> [[VCVTFP2BF1_I]] to <4 x bfloat>
// CHECK-A32-SOFTFP-NEXT: store <4 x bfloat> [[TMP1]], <4 x bfloat>* [[RETVAL_I1]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP2:%.*]] = bitcast <4 x bfloat>* [[RETVAL_I1]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: [[TMP3:%.*]] = load <2 x i32>, <2 x i32>* [[TMP2]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP4:%.*]] = bitcast <4 x bfloat>* [[COERCE_I]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: store <2 x i32> [[TMP3]], <2 x i32>* [[TMP4]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP5:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[COERCE_I]], align 8
// CHECK-A32-SOFTFP-NEXT: store <4 x bfloat> [[TMP5]], <4 x bfloat>* [[RETVAL_I]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP6:%.*]] = bitcast <4 x bfloat>* [[RETVAL_I]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: [[TMP7:%.*]] = load <2 x i32>, <2 x i32>* [[TMP6]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP8:%.*]] = bitcast <4 x bfloat>* [[COERCE]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: store <2 x i32> [[TMP7]], <2 x i32>* [[TMP8]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP9:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[COERCE]], align 8
// CHECK-A32-SOFTFP-NEXT: store <4 x bfloat> [[TMP9]], <4 x bfloat>* [[RETVAL]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP10:%.*]] = bitcast <4 x bfloat>* [[RETVAL]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: [[TMP11:%.*]] = load <2 x i32>, <2 x i32>* [[TMP10]], align 8
// CHECK-A32-SOFTFP-NEXT: ret <2 x i32> [[TMP11]]
//
bfloat16x4_t test_vcvt_bf16_f32(float32x4_t a) {
return vcvt_bf16_f32(a);
}
// CHECK-A64-LABEL: @test_vcvtq_low_bf16_f32(
// CHECK-A64-NEXT: entry:
// CHECK-A64-NEXT: [[TMP0:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
// CHECK-A64-NEXT: [[__A64_VCVTQ_LOW_BF16_V1_I:%.*]] = call <8 x bfloat> @llvm.aarch64.neon.bfcvtn(<4 x float> [[A]]) #[[ATTR3]]
// CHECK-A64-NEXT: [[__A64_VCVTQ_LOW_BF16_V2_I:%.*]] = bitcast <8 x bfloat> [[__A64_VCVTQ_LOW_BF16_V1_I]] to <16 x i8>
// CHECK-A64-NEXT: ret <8 x bfloat> [[__A64_VCVTQ_LOW_BF16_V1_I]]
//
// CHECK-A32-HARDFP-LABEL: @test_vcvtq_low_bf16_f32(
// CHECK-A32-HARDFP-NEXT: entry:
// CHECK-A32-HARDFP-NEXT: [[TMP0:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
// CHECK-A32-HARDFP-NEXT: [[VCVTFP2BF1_I:%.*]] = call <4 x bfloat> @llvm.arm.neon.vcvtfp2bf.v4bf16(<4 x float> [[A]]) #[[ATTR3]]
// CHECK-A32-HARDFP-NEXT: [[SHUFFLE_I:%.*]] = shufflevector <4 x bfloat> zeroinitializer, <4 x bfloat> [[VCVTFP2BF1_I]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
// CHECK-A32-HARDFP-NEXT: ret <8 x bfloat> [[SHUFFLE_I]]
//
// CHECK-A32-SOFTFP-LABEL: @test_vcvtq_low_bf16_f32(
// CHECK-A32-SOFTFP-NEXT: entry:
// CHECK-A32-SOFTFP-NEXT: [[RETVAL_I4:%.*]] = alloca <8 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[__P0_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[__P1_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[RETVAL_I1:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[RETVAL_I:%.*]] = alloca <8 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[COERCE_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[COERCE1_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[COERCE2_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[COERCE4_I:%.*]] = alloca <8 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[RETVAL:%.*]] = alloca <8 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[COERCE:%.*]] = alloca <8 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP0:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
// CHECK-A32-SOFTFP-NEXT: [[VCVTFP2BF1_I:%.*]] = call <4 x i16> @llvm.arm.neon.vcvtfp2bf.v4i16(<4 x float> [[A]]) #[[ATTR3]]
// CHECK-A32-SOFTFP-NEXT: [[TMP1:%.*]] = bitcast <4 x i16> [[VCVTFP2BF1_I]] to <4 x bfloat>
// CHECK-A32-SOFTFP-NEXT: store <4 x bfloat> [[TMP1]], <4 x bfloat>* [[RETVAL_I1]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP2:%.*]] = bitcast <4 x bfloat>* [[RETVAL_I1]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: [[TMP3:%.*]] = load <2 x i32>, <2 x i32>* [[TMP2]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP4:%.*]] = bitcast <4 x bfloat>* [[COERCE_I]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: store <2 x i32> [[TMP3]], <2 x i32>* [[TMP4]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP5:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[COERCE_I]], align 8
// CHECK-A32-SOFTFP-NEXT: store <4 x bfloat> zeroinitializer, <4 x bfloat>* [[COERCE1_I]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP6:%.*]] = bitcast <4 x bfloat>* [[COERCE1_I]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: [[TMP7:%.*]] = load <2 x i32>, <2 x i32>* [[TMP6]], align 8
// CHECK-A32-SOFTFP-NEXT: store <4 x bfloat> [[TMP5]], <4 x bfloat>* [[COERCE2_I]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP8:%.*]] = bitcast <4 x bfloat>* [[COERCE2_I]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: [[TMP9:%.*]] = load <2 x i32>, <2 x i32>* [[TMP8]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP10:%.*]] = bitcast <4 x bfloat>* [[__P0_I]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: store <2 x i32> [[TMP7]], <2 x i32>* [[TMP10]], align 8
// CHECK-A32-SOFTFP-NEXT: [[__P01_I:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[__P0_I]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP11:%.*]] = bitcast <4 x bfloat>* [[__P1_I]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: store <2 x i32> [[TMP9]], <2 x i32>* [[TMP11]], align 8
// CHECK-A32-SOFTFP-NEXT: [[__P12_I:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[__P1_I]], align 8
// CHECK-A32-SOFTFP-NEXT: [[SHUFFLE_I:%.*]] = shufflevector <4 x bfloat> [[__P01_I]], <4 x bfloat> [[__P12_I]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
// CHECK-A32-SOFTFP-NEXT: store <8 x bfloat> [[SHUFFLE_I]], <8 x bfloat>* [[RETVAL_I4]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP12:%.*]] = bitcast <8 x bfloat>* [[RETVAL_I4]] to <4 x i32>*
// CHECK-A32-SOFTFP-NEXT: [[TMP13:%.*]] = load <4 x i32>, <4 x i32>* [[TMP12]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP14:%.*]] = bitcast <8 x bfloat>* [[COERCE4_I]] to <4 x i32>*
// CHECK-A32-SOFTFP-NEXT: store <4 x i32> [[TMP13]], <4 x i32>* [[TMP14]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP15:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[COERCE4_I]], align 8
// CHECK-A32-SOFTFP-NEXT: store <8 x bfloat> [[TMP15]], <8 x bfloat>* [[RETVAL_I]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP16:%.*]] = bitcast <8 x bfloat>* [[RETVAL_I]] to <4 x i32>*
// CHECK-A32-SOFTFP-NEXT: [[TMP17:%.*]] = load <4 x i32>, <4 x i32>* [[TMP16]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP18:%.*]] = bitcast <8 x bfloat>* [[COERCE]] to <4 x i32>*
// CHECK-A32-SOFTFP-NEXT: store <4 x i32> [[TMP17]], <4 x i32>* [[TMP18]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP19:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[COERCE]], align 8
// CHECK-A32-SOFTFP-NEXT: store <8 x bfloat> [[TMP19]], <8 x bfloat>* [[RETVAL]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP20:%.*]] = bitcast <8 x bfloat>* [[RETVAL]] to <4 x i32>*
// CHECK-A32-SOFTFP-NEXT: [[TMP21:%.*]] = load <4 x i32>, <4 x i32>* [[TMP20]], align 8
// CHECK-A32-SOFTFP-NEXT: ret <4 x i32> [[TMP21]]
//
bfloat16x8_t test_vcvtq_low_bf16_f32(float32x4_t a) {
return vcvtq_low_bf16_f32(a);
}
// CHECK-A64-LABEL: @test_vcvtq_high_bf16_f32(
// CHECK-A64-NEXT: entry:
// CHECK-A64-NEXT: [[TMP0:%.*]] = bitcast <8 x bfloat> [[INACTIVE:%.*]] to <16 x i8>
// CHECK-A64-NEXT: [[TMP1:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
// CHECK-A64-NEXT: [[VCVTQ_HIGH_BF16_V2_I:%.*]] = call <8 x bfloat> @llvm.aarch64.neon.bfcvtn2(<8 x bfloat> [[INACTIVE]], <4 x float> [[A]]) #[[ATTR3]]
// CHECK-A64-NEXT: [[VCVTQ_HIGH_BF16_V3_I:%.*]] = bitcast <8 x bfloat> [[VCVTQ_HIGH_BF16_V2_I]] to <16 x i8>
// CHECK-A64-NEXT: ret <8 x bfloat> [[VCVTQ_HIGH_BF16_V2_I]]
//
// CHECK-A32-HARDFP-LABEL: @test_vcvtq_high_bf16_f32(
// CHECK-A32-HARDFP-NEXT: entry:
// CHECK-A32-HARDFP-NEXT: [[TMP0:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
// CHECK-A32-HARDFP-NEXT: [[VCVTFP2BF1_I:%.*]] = call <4 x bfloat> @llvm.arm.neon.vcvtfp2bf.v4bf16(<4 x float> [[A]]) #[[ATTR3]]
// CHECK-A32-HARDFP-NEXT: [[SHUFFLE_I:%.*]] = shufflevector <8 x bfloat> [[INACTIVE:%.*]], <8 x bfloat> [[INACTIVE]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
// CHECK-A32-HARDFP-NEXT: [[SHUFFLE_I8:%.*]] = shufflevector <4 x bfloat> [[VCVTFP2BF1_I]], <4 x bfloat> [[SHUFFLE_I]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
// CHECK-A32-HARDFP-NEXT: ret <8 x bfloat> [[SHUFFLE_I8]]
//
// CHECK-A32-SOFTFP-LABEL: @test_vcvtq_high_bf16_f32(
// CHECK-A32-SOFTFP-NEXT: entry:
// CHECK-A32-SOFTFP-NEXT: [[RETVAL_I11:%.*]] = alloca <8 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[__P0_I12:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[__P1_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[RETVAL_I8:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[RETVAL_I3:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[__P0_I4:%.*]] = alloca <8 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[RETVAL_I:%.*]] = alloca <8 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[__P0_I:%.*]] = alloca <8 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[COERCE_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[COERCE2_I:%.*]] = alloca <8 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[COERCE4_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[COERCE5_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[COERCE6_I:%.*]] = alloca <4 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[COERCE8_I:%.*]] = alloca <8 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[RETVAL:%.*]] = alloca <8 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[INACTIVE:%.*]] = alloca <8 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[COERCE:%.*]] = alloca <8 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[COERCE2:%.*]] = alloca <8 x bfloat>, align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP0:%.*]] = bitcast <8 x bfloat>* [[INACTIVE]] to <4 x i32>*
// CHECK-A32-SOFTFP-NEXT: store <4 x i32> [[INACTIVE_COERCE:%.*]], <4 x i32>* [[TMP0]], align 8
// CHECK-A32-SOFTFP-NEXT: [[INACTIVE1:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[INACTIVE]], align 8
// CHECK-A32-SOFTFP-NEXT: store <8 x bfloat> [[INACTIVE1]], <8 x bfloat>* [[COERCE]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP1:%.*]] = bitcast <8 x bfloat>* [[COERCE]] to <4 x i32>*
// CHECK-A32-SOFTFP-NEXT: [[TMP2:%.*]] = load <4 x i32>, <4 x i32>* [[TMP1]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP3:%.*]] = bitcast <8 x bfloat>* [[__P0_I]] to <4 x i32>*
// CHECK-A32-SOFTFP-NEXT: store <4 x i32> [[TMP2]], <4 x i32>* [[TMP3]], align 8
// CHECK-A32-SOFTFP-NEXT: [[__P01_I:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[__P0_I]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP4:%.*]] = bitcast <4 x float> [[A:%.*]] to <16 x i8>
// CHECK-A32-SOFTFP-NEXT: [[VCVTFP2BF1_I:%.*]] = call <4 x i16> @llvm.arm.neon.vcvtfp2bf.v4i16(<4 x float> [[A]]) #[[ATTR3]]
// CHECK-A32-SOFTFP-NEXT: [[TMP5:%.*]] = bitcast <4 x i16> [[VCVTFP2BF1_I]] to <4 x bfloat>
// CHECK-A32-SOFTFP-NEXT: store <4 x bfloat> [[TMP5]], <4 x bfloat>* [[RETVAL_I8]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP6:%.*]] = bitcast <4 x bfloat>* [[RETVAL_I8]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: [[TMP7:%.*]] = load <2 x i32>, <2 x i32>* [[TMP6]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP8:%.*]] = bitcast <4 x bfloat>* [[COERCE_I]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: store <2 x i32> [[TMP7]], <2 x i32>* [[TMP8]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP9:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[COERCE_I]], align 8
// CHECK-A32-SOFTFP-NEXT: store <8 x bfloat> [[__P01_I]], <8 x bfloat>* [[COERCE2_I]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP10:%.*]] = bitcast <8 x bfloat>* [[COERCE2_I]] to <4 x i32>*
// CHECK-A32-SOFTFP-NEXT: [[TMP11:%.*]] = load <4 x i32>, <4 x i32>* [[TMP10]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP12:%.*]] = bitcast <8 x bfloat>* [[__P0_I4]] to <4 x i32>*
// CHECK-A32-SOFTFP-NEXT: store <4 x i32> [[TMP11]], <4 x i32>* [[TMP12]], align 8
// CHECK-A32-SOFTFP-NEXT: [[__P01_I7:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[__P0_I4]], align 8
// CHECK-A32-SOFTFP-NEXT: [[SHUFFLE_I:%.*]] = shufflevector <8 x bfloat> [[__P01_I7]], <8 x bfloat> [[__P01_I7]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
// CHECK-A32-SOFTFP-NEXT: store <4 x bfloat> [[SHUFFLE_I]], <4 x bfloat>* [[RETVAL_I3]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP13:%.*]] = bitcast <4 x bfloat>* [[RETVAL_I3]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: [[TMP14:%.*]] = load <2 x i32>, <2 x i32>* [[TMP13]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP15:%.*]] = bitcast <4 x bfloat>* [[COERCE4_I]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: store <2 x i32> [[TMP14]], <2 x i32>* [[TMP15]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP16:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[COERCE4_I]], align 8
// CHECK-A32-SOFTFP-NEXT: store <4 x bfloat> [[TMP9]], <4 x bfloat>* [[COERCE5_I]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP17:%.*]] = bitcast <4 x bfloat>* [[COERCE5_I]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: [[TMP18:%.*]] = load <2 x i32>, <2 x i32>* [[TMP17]], align 8
// CHECK-A32-SOFTFP-NEXT: store <4 x bfloat> [[TMP16]], <4 x bfloat>* [[COERCE6_I]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP19:%.*]] = bitcast <4 x bfloat>* [[COERCE6_I]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: [[TMP20:%.*]] = load <2 x i32>, <2 x i32>* [[TMP19]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP21:%.*]] = bitcast <4 x bfloat>* [[__P0_I12]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: store <2 x i32> [[TMP18]], <2 x i32>* [[TMP21]], align 8
// CHECK-A32-SOFTFP-NEXT: [[__P01_I16:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[__P0_I12]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP22:%.*]] = bitcast <4 x bfloat>* [[__P1_I]] to <2 x i32>*
// CHECK-A32-SOFTFP-NEXT: store <2 x i32> [[TMP20]], <2 x i32>* [[TMP22]], align 8
// CHECK-A32-SOFTFP-NEXT: [[__P12_I:%.*]] = load <4 x bfloat>, <4 x bfloat>* [[__P1_I]], align 8
// CHECK-A32-SOFTFP-NEXT: [[SHUFFLE_I17:%.*]] = shufflevector <4 x bfloat> [[__P01_I16]], <4 x bfloat> [[__P12_I]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
// CHECK-A32-SOFTFP-NEXT: store <8 x bfloat> [[SHUFFLE_I17]], <8 x bfloat>* [[RETVAL_I11]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP23:%.*]] = bitcast <8 x bfloat>* [[RETVAL_I11]] to <4 x i32>*
// CHECK-A32-SOFTFP-NEXT: [[TMP24:%.*]] = load <4 x i32>, <4 x i32>* [[TMP23]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP25:%.*]] = bitcast <8 x bfloat>* [[COERCE8_I]] to <4 x i32>*
// CHECK-A32-SOFTFP-NEXT: store <4 x i32> [[TMP24]], <4 x i32>* [[TMP25]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP26:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[COERCE8_I]], align 8
// CHECK-A32-SOFTFP-NEXT: store <8 x bfloat> [[TMP26]], <8 x bfloat>* [[RETVAL_I]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP27:%.*]] = bitcast <8 x bfloat>* [[RETVAL_I]] to <4 x i32>*
// CHECK-A32-SOFTFP-NEXT: [[TMP28:%.*]] = load <4 x i32>, <4 x i32>* [[TMP27]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP29:%.*]] = bitcast <8 x bfloat>* [[COERCE2]] to <4 x i32>*
// CHECK-A32-SOFTFP-NEXT: store <4 x i32> [[TMP28]], <4 x i32>* [[TMP29]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP30:%.*]] = load <8 x bfloat>, <8 x bfloat>* [[COERCE2]], align 8
// CHECK-A32-SOFTFP-NEXT: store <8 x bfloat> [[TMP30]], <8 x bfloat>* [[RETVAL]], align 8
// CHECK-A32-SOFTFP-NEXT: [[TMP31:%.*]] = bitcast <8 x bfloat>* [[RETVAL]] to <4 x i32>*
// CHECK-A32-SOFTFP-NEXT: [[TMP32:%.*]] = load <4 x i32>, <4 x i32>* [[TMP31]], align 8
// CHECK-A32-SOFTFP-NEXT: ret <4 x i32> [[TMP32]]
//
bfloat16x8_t test_vcvtq_high_bf16_f32(bfloat16x8_t inactive, float32x4_t a) {
return vcvtq_high_bf16_f32(inactive, a);
}
// CHECK-A64-LABEL: @test_vcvth_bf16_f32(
// CHECK-A64-NEXT: entry:
// CHECK-A64-NEXT: [[VCVTH_BF16_F32_I:%.*]] = call bfloat @llvm.aarch64.neon.bfcvt(float [[A:%.*]]) #[[ATTR3]]
// CHECK-A64-NEXT: ret bfloat [[VCVTH_BF16_F32_I]]
//
// CHECK-A32-HARDFP-LABEL: @test_vcvth_bf16_f32(
// CHECK-A32-HARDFP-NEXT: entry:
// CHECK-A32-HARDFP-NEXT: [[VCVTBFP2BF_I:%.*]] = call bfloat @llvm.arm.neon.vcvtbfp2bf(float [[A:%.*]]) #[[ATTR3]]
// CHECK-A32-HARDFP-NEXT: ret bfloat [[VCVTBFP2BF_I]]
//
// CHECK-A32-SOFTFP-LABEL: @test_vcvth_bf16_f32(
// CHECK-A32-SOFTFP-NEXT: entry:
// CHECK-A32-SOFTFP-NEXT: [[VCVTBFP2BF_I:%.*]] = call bfloat @llvm.arm.neon.vcvtbfp2bf(float [[A:%.*]]) #[[ATTR3]]
// CHECK-A32-SOFTFP-NEXT: ret bfloat [[VCVTBFP2BF_I]]
//
bfloat16_t test_vcvth_bf16_f32(float32_t a) {
return vcvth_bf16_f32(a);
}
// CHECK-LABEL: @test_vcvtah_f32_bf16(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[__REINT_I:%.*]] = alloca bfloat, align 2
// CHECK-NEXT: [[__REINT1_I:%.*]] = alloca i32, align 4
// CHECK-NEXT: store bfloat [[A:%.*]], bfloat* [[__REINT_I]], align 2
// CHECK-NEXT: [[TMP0:%.*]] = bitcast bfloat* [[__REINT_I]] to i32*
// CHECK-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 2
// CHECK-NEXT: [[SHL_I:%.*]] = shl i32 [[TMP1]], 16
// CHECK-NEXT: store i32 [[SHL_I]], i32* [[__REINT1_I]], align 4
// CHECK-NEXT: [[TMP2:%.*]] = bitcast i32* [[__REINT1_I]] to float*
// CHECK-NEXT: [[TMP3:%.*]] = load float, float* [[TMP2]], align 4
// CHECK-NEXT: ret float [[TMP3]]
//
float32_t test_vcvtah_f32_bf16(bfloat16_t a) {
return vcvtah_f32_bf16(a);
}
|
