llvm.org GIT mirror llvm / release_40 test / CodeGen / AArch64 / arm64-neon-2velem-high.ll
release_40

Tree @release_40 (Download .tar.gz)

arm64-neon-2velem-high.ll @release_40raw · history · blame

  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
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
; RUN: llc -mtriple=arm64-none-linux-gnu -mattr=+neon -fp-contract=fast \
; RUN:     < %s -verify-machineinstrs -asm-verbose=false | FileCheck %s

define <4 x i32> @test_vmull_high_n_s16(<8 x i16> %a, i16 %b) #0 {
; CHECK-LABEL: test_vmull_high_n_s16:
; CHECK-NEXT: dup [[REPLICATE:v[0-9]+]].8h, w0
; CHECK-NEXT: smull2 {{v[0-9]+}}.4s, {{v[0-9]+}}.8h, [[REPLICATE]].8h
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <8 x i16> %a, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %vecinit.i.i = insertelement <4 x i16> undef, i16 %b, i32 0
  %vecinit1.i.i = insertelement <4 x i16> %vecinit.i.i, i16 %b, i32 1
  %vecinit2.i.i = insertelement <4 x i16> %vecinit1.i.i, i16 %b, i32 2
  %vecinit3.i.i = insertelement <4 x i16> %vecinit2.i.i, i16 %b, i32 3
  %vmull15.i.i = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> %shuffle.i.i, <4 x i16> %vecinit3.i.i)
  ret <4 x i32> %vmull15.i.i
}

define <4 x i32> @test_vmull_high_n_s16_imm(<8 x i16> %a) #0 {
; CHECK-LABEL: test_vmull_high_n_s16_imm:
; CHECK-NEXT: movi [[REPLICATE:v[0-9]+]].8h, #29
; CHECK-NEXT: smull2 {{v[0-9]+}}.4s, {{v[0-9]+}}.8h, [[REPLICATE]].8h
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <8 x i16> %a, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %vmull15.i.i = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> %shuffle.i.i, <4 x i16> <i16 29, i16 29, i16 29, i16 29>)
  ret <4 x i32> %vmull15.i.i
}

define <2 x i64> @test_vmull_high_n_s32(<4 x i32> %a, i32 %b) #0 {
; CHECK-LABEL: test_vmull_high_n_s32:
; CHECK-NEXT: dup [[REPLICATE:v[0-9]+]].4s, w0
; CHECK-NEXT: smull2 {{v[0-9]+}}.2d, {{v[0-9]+}}.4s, [[REPLICATE]].4s
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <4 x i32> %a, <4 x i32> undef, <2 x i32> <i32 2, i32 3>
  %vecinit.i.i = insertelement <2 x i32> undef, i32 %b, i32 0
  %vecinit1.i.i = insertelement <2 x i32> %vecinit.i.i, i32 %b, i32 1
  %vmull9.i.i = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> %shuffle.i.i, <2 x i32> %vecinit1.i.i)
  ret <2 x i64> %vmull9.i.i
}

define <2 x i64> @test_vmull_high_n_s32_imm(<4 x i32> %a) #0 {
; CHECK-LABEL: test_vmull_high_n_s32_imm:
; CHECK-NEXT: movi [[REPLICATE:v[0-9]+]].4s, #1, msl #8
; CHECK-NEXT: smull2 {{v[0-9]+}}.2d, {{v[0-9]+}}.4s, [[REPLICATE]].4s
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <4 x i32> %a, <4 x i32> undef, <2 x i32> <i32 2, i32 3>
  %vmull9.i.i = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> %shuffle.i.i, <2 x i32> <i32 511, i32 511>)
  ret <2 x i64> %vmull9.i.i
}

define <4 x i32> @test_vmull_high_n_u16(<8 x i16> %a, i16 %b) #0 {
; CHECK-LABEL: test_vmull_high_n_u16:
; CHECK-NEXT: dup [[REPLICATE:v[0-9]+]].8h, w0
; CHECK-NEXT: umull2 {{v[0-9]+}}.4s, {{v[0-9]+}}.8h, [[REPLICATE]].8h
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <8 x i16> %a, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %vecinit.i.i = insertelement <4 x i16> undef, i16 %b, i32 0
  %vecinit1.i.i = insertelement <4 x i16> %vecinit.i.i, i16 %b, i32 1
  %vecinit2.i.i = insertelement <4 x i16> %vecinit1.i.i, i16 %b, i32 2
  %vecinit3.i.i = insertelement <4 x i16> %vecinit2.i.i, i16 %b, i32 3
  %vmull15.i.i = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> %shuffle.i.i, <4 x i16> %vecinit3.i.i)
  ret <4 x i32> %vmull15.i.i
}

define <4 x i32> @test_vmull_high_n_u16_imm(<8 x i16> %a) #0 {
; CHECK-LABEL: test_vmull_high_n_u16_imm:
; CHECK-NEXT: movi [[REPLICATE:v[0-9]+]].8h, #17, lsl #8
; CHECK-NEXT: umull2 {{v[0-9]+}}.4s, {{v[0-9]+}}.8h, [[REPLICATE]].8h
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <8 x i16> %a, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %vmull15.i.i = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> %shuffle.i.i, <4 x i16> <i16 4352, i16 4352, i16 4352, i16 4352>)
  ret <4 x i32> %vmull15.i.i
}

define <2 x i64> @test_vmull_high_n_u32(<4 x i32> %a, i32 %b) #0 {
; CHECK-LABEL: test_vmull_high_n_u32:
; CHECK-NEXT: dup [[REPLICATE:v[0-9]+]].4s, w0
; CHECK-NEXT: umull2 {{v[0-9]+}}.2d, {{v[0-9]+}}.4s, [[REPLICATE]].4s
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <4 x i32> %a, <4 x i32> undef, <2 x i32> <i32 2, i32 3>
  %vecinit.i.i = insertelement <2 x i32> undef, i32 %b, i32 0
  %vecinit1.i.i = insertelement <2 x i32> %vecinit.i.i, i32 %b, i32 1
  %vmull9.i.i = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> %shuffle.i.i, <2 x i32> %vecinit1.i.i)
  ret <2 x i64> %vmull9.i.i
}

define <2 x i64> @test_vmull_high_n_u32_imm(<4 x i32> %a) #0 {
; CHECK-LABEL: test_vmull_high_n_u32_imm:
; CHECK-NEXT: mvni [[REPLICATE:v[0-9]+]].4s, #1, msl #8
; CHECK-NEXT: umull2 {{v[0-9]+}}.2d, {{v[0-9]+}}.4s, [[REPLICATE]].4s
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <4 x i32> %a, <4 x i32> undef, <2 x i32> <i32 2, i32 3>
  %vmull9.i.i = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> %shuffle.i.i, <2 x i32> <i32 4294966784, i32 4294966784>)
  ret <2 x i64> %vmull9.i.i
}

define <4 x i32> @test_vqdmull_high_n_s16(<8 x i16> %a, i16 %b) #0 {
; CHECK-LABEL: test_vqdmull_high_n_s16:
; CHECK-NEXT: dup [[REPLICATE:v[0-9]+]].8h, w0
; CHECK-NEXT: sqdmull2 {{v[0-9]+}}.4s, {{v[0-9]+}}.8h, [[REPLICATE]].8h
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <8 x i16> %a, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %vecinit.i.i = insertelement <4 x i16> undef, i16 %b, i32 0
  %vecinit1.i.i = insertelement <4 x i16> %vecinit.i.i, i16 %b, i32 1
  %vecinit2.i.i = insertelement <4 x i16> %vecinit1.i.i, i16 %b, i32 2
  %vecinit3.i.i = insertelement <4 x i16> %vecinit2.i.i, i16 %b, i32 3
  %vqdmull15.i.i = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> %shuffle.i.i, <4 x i16> %vecinit3.i.i)
  ret <4 x i32> %vqdmull15.i.i
}

define <4 x i32> @test_vqdmull_high_n_s16_imm(<8 x i16> %a) #0 {
; CHECK-LABEL: test_vqdmull_high_n_s16_imm:
; CHECK-NEXT: mvni [[REPLICATE:v[0-9]+]].8h, #17, lsl #8
; CHECK-NEXT: sqdmull2 {{v[0-9]+}}.4s, {{v[0-9]+}}.8h, [[REPLICATE]].8h
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <8 x i16> %a, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %vqdmull15.i.i = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> %shuffle.i.i, <4 x i16> <i16 61183, i16 61183, i16 61183, i16 61183>)
  ret <4 x i32> %vqdmull15.i.i
}

define <2 x i64> @test_vqdmull_high_n_s32(<4 x i32> %a, i32 %b) #0 {
; CHECK-LABEL: test_vqdmull_high_n_s32:
; CHECK-NEXT: dup [[REPLICATE:v[0-9]+]].4s, w0
; CHECK-NEXT: sqdmull2 {{v[0-9]+}}.2d, {{v[0-9]+}}.4s, [[REPLICATE]].4s
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <4 x i32> %a, <4 x i32> undef, <2 x i32> <i32 2, i32 3>
  %vecinit.i.i = insertelement <2 x i32> undef, i32 %b, i32 0
  %vecinit1.i.i = insertelement <2 x i32> %vecinit.i.i, i32 %b, i32 1
  %vqdmull9.i.i = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> %shuffle.i.i, <2 x i32> %vecinit1.i.i)
  ret <2 x i64> %vqdmull9.i.i
}

define <2 x i64> @test_vqdmull_high_n_s32_imm(<4 x i32> %a) #0 {
; CHECK-LABEL: test_vqdmull_high_n_s32_imm:
; CHECK-NEXT: movi [[REPLICATE:v[0-9]+]].4s, #29
; CHECK-NEXT: sqdmull2 {{v[0-9]+}}.2d, {{v[0-9]+}}.4s, [[REPLICATE]].4s
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <4 x i32> %a, <4 x i32> undef, <2 x i32> <i32 2, i32 3>
  %vqdmull9.i.i = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> %shuffle.i.i, <2 x i32> <i32 29, i32 29>)
  ret <2 x i64> %vqdmull9.i.i
}

define <4 x i32> @test_vmlal_high_n_s16(<4 x i32> %a, <8 x i16> %b, i16 %c) #0 {
; CHECK-LABEL: test_vmlal_high_n_s16:
; CHECK-NEXT: dup [[REPLICATE:v[0-9]+]].8h, w0
; CHECK-NEXT: smlal2 {{v[0-9]+}}.4s, {{v[0-9]+}}.8h, [[REPLICATE]].8h
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <8 x i16> %b, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %vecinit.i.i = insertelement <4 x i16> undef, i16 %c, i32 0
  %vecinit1.i.i = insertelement <4 x i16> %vecinit.i.i, i16 %c, i32 1
  %vecinit2.i.i = insertelement <4 x i16> %vecinit1.i.i, i16 %c, i32 2
  %vecinit3.i.i = insertelement <4 x i16> %vecinit2.i.i, i16 %c, i32 3
  %vmull2.i.i.i = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> %shuffle.i.i, <4 x i16> %vecinit3.i.i)
  %add.i.i = add <4 x i32> %vmull2.i.i.i, %a
  ret <4 x i32> %add.i.i
}

define <4 x i32> @test_vmlal_high_n_s16_imm(<4 x i32> %a, <8 x i16> %b) #0 {
; CHECK-LABEL: test_vmlal_high_n_s16_imm:
; CHECK-NEXT: movi [[REPLICATE:v[0-9]+]].8h, #29
; CHECK-NEXT: smlal2 {{v[0-9]+}}.4s, {{v[0-9]+}}.8h, [[REPLICATE]].8h
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <8 x i16> %b, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %vmull2.i.i.i = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> %shuffle.i.i, <4 x i16> <i16 29, i16 29, i16 29, i16 29>)
  %add.i.i = add <4 x i32> %vmull2.i.i.i, %a
  ret <4 x i32> %add.i.i
}

define <2 x i64> @test_vmlal_high_n_s32(<2 x i64> %a, <4 x i32> %b, i32 %c) #0 {
; CHECK-LABEL: test_vmlal_high_n_s32:
; CHECK-NEXT: dup [[REPLICATE:v[0-9]+]].4s, w0
; CHECK-NEXT: smlal2 {{v[0-9]+}}.2d, {{v[0-9]+}}.4s, [[REPLICATE]].4s
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <4 x i32> %b, <4 x i32> undef, <2 x i32> <i32 2, i32 3>
  %vecinit.i.i = insertelement <2 x i32> undef, i32 %c, i32 0
  %vecinit1.i.i = insertelement <2 x i32> %vecinit.i.i, i32 %c, i32 1
  %vmull2.i.i.i = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> %shuffle.i.i, <2 x i32> %vecinit1.i.i)
  %add.i.i = add <2 x i64> %vmull2.i.i.i, %a
  ret <2 x i64> %add.i.i
}

define <2 x i64> @test_vmlal_high_n_s32_imm(<2 x i64> %a, <4 x i32> %b) #0 {
; CHECK-LABEL: test_vmlal_high_n_s32_imm:
; CHECK-NEXT: movi [[REPLICATE:v[0-9]+]].4s, #29
; CHECK-NEXT: smlal2 {{v[0-9]+}}.2d, {{v[0-9]+}}.4s, [[REPLICATE]].4s
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <4 x i32> %b, <4 x i32> undef, <2 x i32> <i32 2, i32 3>
  %vmull2.i.i.i = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> %shuffle.i.i, <2 x i32> <i32 29, i32 29>)
  %add.i.i = add <2 x i64> %vmull2.i.i.i, %a
  ret <2 x i64> %add.i.i
}

define <4 x i32> @test_vmlal_high_n_u16(<4 x i32> %a, <8 x i16> %b, i16 %c) #0 {
; CHECK-LABEL: test_vmlal_high_n_u16:
; CHECK-NEXT: dup [[REPLICATE:v[0-9]+]].8h, w0
; CHECK-NEXT: umlal2 {{v[0-9]+}}.4s, {{v[0-9]+}}.8h, [[REPLICATE]].8h
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <8 x i16> %b, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %vecinit.i.i = insertelement <4 x i16> undef, i16 %c, i32 0
  %vecinit1.i.i = insertelement <4 x i16> %vecinit.i.i, i16 %c, i32 1
  %vecinit2.i.i = insertelement <4 x i16> %vecinit1.i.i, i16 %c, i32 2
  %vecinit3.i.i = insertelement <4 x i16> %vecinit2.i.i, i16 %c, i32 3
  %vmull2.i.i.i = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> %shuffle.i.i, <4 x i16> %vecinit3.i.i)
  %add.i.i = add <4 x i32> %vmull2.i.i.i, %a
  ret <4 x i32> %add.i.i
}

define <4 x i32> @test_vmlal_high_n_u16_imm(<4 x i32> %a, <8 x i16> %b) #0 {
; CHECK-LABEL: test_vmlal_high_n_u16_imm:
; CHECK-NEXT: movi [[REPLICATE:v[0-9]+]].8h, #29
; CHECK-NEXT: umlal2 {{v[0-9]+}}.4s, {{v[0-9]+}}.8h, [[REPLICATE]].8h
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <8 x i16> %b, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %vmull2.i.i.i = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> %shuffle.i.i, <4 x i16> <i16 29, i16 29, i16 29, i16 29>)
  %add.i.i = add <4 x i32> %vmull2.i.i.i, %a
  ret <4 x i32> %add.i.i
}

define <2 x i64> @test_vmlal_high_n_u32(<2 x i64> %a, <4 x i32> %b, i32 %c) #0 {
; CHECK-LABEL: test_vmlal_high_n_u32:
; CHECK-NEXT: dup [[REPLICATE:v[0-9]+]].4s, w0
; CHECK-NEXT: umlal2 {{v[0-9]+}}.2d, {{v[0-9]+}}.4s, [[REPLICATE]].4s
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <4 x i32> %b, <4 x i32> undef, <2 x i32> <i32 2, i32 3>
  %vecinit.i.i = insertelement <2 x i32> undef, i32 %c, i32 0
  %vecinit1.i.i = insertelement <2 x i32> %vecinit.i.i, i32 %c, i32 1
  %vmull2.i.i.i = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> %shuffle.i.i, <2 x i32> %vecinit1.i.i)
  %add.i.i = add <2 x i64> %vmull2.i.i.i, %a
  ret <2 x i64> %add.i.i
}

define <2 x i64> @test_vmlal_high_n_u32_imm(<2 x i64> %a, <4 x i32> %b) #0 {
; CHECK-LABEL: test_vmlal_high_n_u32_imm:
; CHECK-NEXT: movi [[REPLICATE:v[0-9]+]].4s, #29
; CHECK-NEXT: umlal2 {{v[0-9]+}}.2d, {{v[0-9]+}}.4s, [[REPLICATE]].4s
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <4 x i32> %b, <4 x i32> undef, <2 x i32> <i32 2, i32 3>
  %vmull2.i.i.i = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> %shuffle.i.i, <2 x i32> <i32 29, i32 29>)
  %add.i.i = add <2 x i64> %vmull2.i.i.i, %a
  ret <2 x i64> %add.i.i
}

define <4 x i32> @test_vqdmlal_high_n_s16(<4 x i32> %a, <8 x i16> %b, i16 %c) #0 {
; CHECK-LABEL: test_vqdmlal_high_n_s16:
; CHECK-NEXT: dup [[REPLICATE:v[0-9]+]].8h, w0
; CHECK-NEXT: sqdmlal2 {{v[0-9]+}}.4s, {{v[0-9]+}}.8h, [[REPLICATE]].8h
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <8 x i16> %b, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %vecinit.i.i = insertelement <4 x i16> undef, i16 %c, i32 0
  %vecinit1.i.i = insertelement <4 x i16> %vecinit.i.i, i16 %c, i32 1
  %vecinit2.i.i = insertelement <4 x i16> %vecinit1.i.i, i16 %c, i32 2
  %vecinit3.i.i = insertelement <4 x i16> %vecinit2.i.i, i16 %c, i32 3
  %vqdmlal15.i.i = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> %shuffle.i.i, <4 x i16> %vecinit3.i.i)
  %vqdmlal17.i.i = call <4 x i32> @llvm.aarch64.neon.sqadd.v4i32(<4 x i32> %a, <4 x i32> %vqdmlal15.i.i)
  ret <4 x i32> %vqdmlal17.i.i
}

define <4 x i32> @test_vqdmlal_high_n_s16_imm(<4 x i32> %a, <8 x i16> %b) #0 {
; CHECK-LABEL: test_vqdmlal_high_n_s16_imm:
; CHECK-NEXT: movi [[REPLICATE:v[0-9]+]].8h, #29
; CHECK-NEXT: sqdmlal2 {{v[0-9]+}}.4s, {{v[0-9]+}}.8h, [[REPLICATE]].8h
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <8 x i16> %b, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %vqdmlal15.i.i = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> %shuffle.i.i, <4 x i16> <i16 29, i16 29, i16 29, i16 29>)
  %vqdmlal17.i.i = call <4 x i32> @llvm.aarch64.neon.sqadd.v4i32(<4 x i32> %a, <4 x i32> %vqdmlal15.i.i)
  ret <4 x i32> %vqdmlal17.i.i
}

define <2 x i64> @test_vqdmlal_high_n_s32(<2 x i64> %a, <4 x i32> %b, i32 %c) #0 {
; CHECK-LABEL: test_vqdmlal_high_n_s32:
; CHECK-NEXT: dup [[REPLICATE:v[0-9]+]].4s, w0
; CHECK-NEXT: sqdmlal2 {{v[0-9]+}}.2d, {{v[0-9]+}}.4s, [[REPLICATE]].4s
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <4 x i32> %b, <4 x i32> undef, <2 x i32> <i32 2, i32 3>
  %vecinit.i.i = insertelement <2 x i32> undef, i32 %c, i32 0
  %vecinit1.i.i = insertelement <2 x i32> %vecinit.i.i, i32 %c, i32 1
  %vqdmlal9.i.i = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> %shuffle.i.i, <2 x i32> %vecinit1.i.i)
  %vqdmlal11.i.i = call <2 x i64> @llvm.aarch64.neon.sqadd.v2i64(<2 x i64> %a, <2 x i64> %vqdmlal9.i.i)
  ret <2 x i64> %vqdmlal11.i.i
}

define <2 x i64> @test_vqdmlal_high_n_s32_imm(<2 x i64> %a, <4 x i32> %b) #0 {
; CHECK-LABEL: test_vqdmlal_high_n_s32_imm:
; CHECK-NEXT: movi [[REPLICATE:v[0-9]+]].4s, #29
; CHECK-NEXT: sqdmlal2 {{v[0-9]+}}.2d, {{v[0-9]+}}.4s, [[REPLICATE]].4s
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <4 x i32> %b, <4 x i32> undef, <2 x i32> <i32 2, i32 3>
  %vqdmlal9.i.i = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> %shuffle.i.i, <2 x i32> <i32 29, i32 29>)
  %vqdmlal11.i.i = call <2 x i64> @llvm.aarch64.neon.sqadd.v2i64(<2 x i64> %a, <2 x i64> %vqdmlal9.i.i)
  ret <2 x i64> %vqdmlal11.i.i
}

define <4 x i32> @test_vmlsl_high_n_s16(<4 x i32> %a, <8 x i16> %b, i16 %c) #0 {
; CHECK-LABEL: test_vmlsl_high_n_s16:
; CHECK-NEXT: dup [[REPLICATE:v[0-9]+]].8h, w0
; CHECK-NEXT: smlsl2 {{v[0-9]+}}.4s, {{v[0-9]+}}.8h, [[REPLICATE]].8h
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <8 x i16> %b, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %vecinit.i.i = insertelement <4 x i16> undef, i16 %c, i32 0
  %vecinit1.i.i = insertelement <4 x i16> %vecinit.i.i, i16 %c, i32 1
  %vecinit2.i.i = insertelement <4 x i16> %vecinit1.i.i, i16 %c, i32 2
  %vecinit3.i.i = insertelement <4 x i16> %vecinit2.i.i, i16 %c, i32 3
  %vmull2.i.i.i = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> %shuffle.i.i, <4 x i16> %vecinit3.i.i)
  %sub.i.i = sub <4 x i32> %a, %vmull2.i.i.i
  ret <4 x i32> %sub.i.i
}

define <4 x i32> @test_vmlsl_high_n_s16_imm(<4 x i32> %a, <8 x i16> %b) #0 {
; CHECK-LABEL: test_vmlsl_high_n_s16_imm:
; CHECK-NEXT: movi [[REPLICATE:v[0-9]+]].8h, #29
; CHECK-NEXT: smlsl2 {{v[0-9]+}}.4s, {{v[0-9]+}}.8h, [[REPLICATE]].8h
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <8 x i16> %b, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %vmull2.i.i.i = call <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16> %shuffle.i.i, <4 x i16> <i16 29, i16 29, i16 29, i16 29>)
  %sub.i.i = sub <4 x i32> %a, %vmull2.i.i.i
  ret <4 x i32> %sub.i.i
}

define <2 x i64> @test_vmlsl_high_n_s32(<2 x i64> %a, <4 x i32> %b, i32 %c) #0 {
; CHECK-LABEL: test_vmlsl_high_n_s32:
; CHECK-NEXT: dup [[REPLICATE:v[0-9]+]].4s, w0
; CHECK-NEXT: smlsl2 {{v[0-9]+}}.2d, {{v[0-9]+}}.4s, [[REPLICATE]].4s
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <4 x i32> %b, <4 x i32> undef, <2 x i32> <i32 2, i32 3>
  %vecinit.i.i = insertelement <2 x i32> undef, i32 %c, i32 0
  %vecinit1.i.i = insertelement <2 x i32> %vecinit.i.i, i32 %c, i32 1
  %vmull2.i.i.i = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> %shuffle.i.i, <2 x i32> %vecinit1.i.i)
  %sub.i.i = sub <2 x i64> %a, %vmull2.i.i.i
  ret <2 x i64> %sub.i.i
}

define <2 x i64> @test_vmlsl_high_n_s32_imm(<2 x i64> %a, <4 x i32> %b) #0 {
; CHECK-LABEL: test_vmlsl_high_n_s32_imm:
; CHECK-NEXT: movi [[REPLICATE:v[0-9]+]].4s, #29
; CHECK-NEXT: smlsl2 {{v[0-9]+}}.2d, {{v[0-9]+}}.4s, [[REPLICATE]].4s
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <4 x i32> %b, <4 x i32> undef, <2 x i32> <i32 2, i32 3>
  %vmull2.i.i.i = call <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32> %shuffle.i.i, <2 x i32> <i32 29, i32 29>)
  %sub.i.i = sub <2 x i64> %a, %vmull2.i.i.i
  ret <2 x i64> %sub.i.i
}

define <4 x i32> @test_vmlsl_high_n_u16(<4 x i32> %a, <8 x i16> %b, i16 %c) #0 {
; CHECK-LABEL: test_vmlsl_high_n_u16:
; CHECK-NEXT: dup [[REPLICATE:v[0-9]+]].8h, w0
; CHECK-NEXT: umlsl2 {{v[0-9]+}}.4s, {{v[0-9]+}}.8h, [[REPLICATE]].8h
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <8 x i16> %b, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %vecinit.i.i = insertelement <4 x i16> undef, i16 %c, i32 0
  %vecinit1.i.i = insertelement <4 x i16> %vecinit.i.i, i16 %c, i32 1
  %vecinit2.i.i = insertelement <4 x i16> %vecinit1.i.i, i16 %c, i32 2
  %vecinit3.i.i = insertelement <4 x i16> %vecinit2.i.i, i16 %c, i32 3
  %vmull2.i.i.i = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> %shuffle.i.i, <4 x i16> %vecinit3.i.i)
  %sub.i.i = sub <4 x i32> %a, %vmull2.i.i.i
  ret <4 x i32> %sub.i.i
}

define <4 x i32> @test_vmlsl_high_n_u16_imm(<4 x i32> %a, <8 x i16> %b) #0 {
; CHECK-LABEL: test_vmlsl_high_n_u16_imm:
; CHECK-NEXT: movi [[REPLICATE:v[0-9]+]].8h, #29
; CHECK-NEXT: umlsl2 {{v[0-9]+}}.4s, {{v[0-9]+}}.8h, [[REPLICATE]].8h
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <8 x i16> %b, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %vmull2.i.i.i = call <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16> %shuffle.i.i, <4 x i16> <i16 29, i16 29, i16 29, i16 29>)
  %sub.i.i = sub <4 x i32> %a, %vmull2.i.i.i
  ret <4 x i32> %sub.i.i
}

define <2 x i64> @test_vmlsl_high_n_u32(<2 x i64> %a, <4 x i32> %b, i32 %c) #0 {
; CHECK-LABEL: test_vmlsl_high_n_u32:
; CHECK-NEXT: dup [[REPLICATE:v[0-9]+]].4s, w0
; CHECK-NEXT: umlsl2 {{v[0-9]+}}.2d, {{v[0-9]+}}.4s, [[REPLICATE]].4s
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <4 x i32> %b, <4 x i32> undef, <2 x i32> <i32 2, i32 3>
  %vecinit.i.i = insertelement <2 x i32> undef, i32 %c, i32 0
  %vecinit1.i.i = insertelement <2 x i32> %vecinit.i.i, i32 %c, i32 1
  %vmull2.i.i.i = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> %shuffle.i.i, <2 x i32> %vecinit1.i.i)
  %sub.i.i = sub <2 x i64> %a, %vmull2.i.i.i
  ret <2 x i64> %sub.i.i
}

define <2 x i64> @test_vmlsl_high_n_u32_imm(<2 x i64> %a, <4 x i32> %b) #0 {
; CHECK-LABEL: test_vmlsl_high_n_u32_imm:
; CHECK-NEXT: movi [[REPLICATE:v[0-9]+]].4s, #29
; CHECK-NEXT: umlsl2 {{v[0-9]+}}.2d, {{v[0-9]+}}.4s, [[REPLICATE]].4s
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <4 x i32> %b, <4 x i32> undef, <2 x i32> <i32 2, i32 3>
  %vmull2.i.i.i = call <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32> %shuffle.i.i, <2 x i32> <i32 29, i32 29>)
  %sub.i.i = sub <2 x i64> %a, %vmull2.i.i.i
  ret <2 x i64> %sub.i.i
}

define <4 x i32> @test_vqdmlsl_high_n_s16(<4 x i32> %a, <8 x i16> %b, i16 %c) #0 {
; CHECK-LABEL: test_vqdmlsl_high_n_s16:
; CHECK-NEXT: dup [[REPLICATE:v[0-9]+]].8h, w0
; CHECK-NEXT: sqdmlsl2 {{v[0-9]+}}.4s, {{v[0-9]+}}.8h, [[REPLICATE]].8h
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <8 x i16> %b, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %vecinit.i.i = insertelement <4 x i16> undef, i16 %c, i32 0
  %vecinit1.i.i = insertelement <4 x i16> %vecinit.i.i, i16 %c, i32 1
  %vecinit2.i.i = insertelement <4 x i16> %vecinit1.i.i, i16 %c, i32 2
  %vecinit3.i.i = insertelement <4 x i16> %vecinit2.i.i, i16 %c, i32 3
  %vqdmlsl15.i.i = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> %shuffle.i.i, <4 x i16> %vecinit3.i.i)
  %vqdmlsl17.i.i = call <4 x i32> @llvm.aarch64.neon.sqsub.v4i32(<4 x i32> %a, <4 x i32> %vqdmlsl15.i.i)
  ret <4 x i32> %vqdmlsl17.i.i
}

define <4 x i32> @test_vqdmlsl_high_n_s16_imm(<4 x i32> %a, <8 x i16> %b) #0 {
; CHECK-LABEL: test_vqdmlsl_high_n_s16_imm:
; CHECK-NEXT: movi [[REPLICATE:v[0-9]+]].8h, #29
; CHECK-NEXT: sqdmlsl2 {{v[0-9]+}}.4s, {{v[0-9]+}}.8h, [[REPLICATE]].8h
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <8 x i16> %b, <8 x i16> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
  %vqdmlsl15.i.i = call <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16> %shuffle.i.i, <4 x i16> <i16 29, i16 29, i16 29, i16 29>)
  %vqdmlsl17.i.i = call <4 x i32> @llvm.aarch64.neon.sqsub.v4i32(<4 x i32> %a, <4 x i32> %vqdmlsl15.i.i)
  ret <4 x i32> %vqdmlsl17.i.i
}

define <2 x i64> @test_vqdmlsl_high_n_s32(<2 x i64> %a, <4 x i32> %b, i32 %c) #0 {
; CHECK-LABEL: test_vqdmlsl_high_n_s32:
; CHECK-NEXT: dup [[REPLICATE:v[0-9]+]].4s, w0
; CHECK-NEXT: sqdmlsl2 {{v[0-9]+}}.2d, {{v[0-9]+}}.4s, [[REPLICATE]].4s
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <4 x i32> %b, <4 x i32> undef, <2 x i32> <i32 2, i32 3>
  %vecinit.i.i = insertelement <2 x i32> undef, i32 %c, i32 0
  %vecinit1.i.i = insertelement <2 x i32> %vecinit.i.i, i32 %c, i32 1
  %vqdmlsl9.i.i = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> %shuffle.i.i, <2 x i32> %vecinit1.i.i)
  %vqdmlsl11.i.i = call <2 x i64> @llvm.aarch64.neon.sqsub.v2i64(<2 x i64> %a, <2 x i64> %vqdmlsl9.i.i)
  ret <2 x i64> %vqdmlsl11.i.i
}

define <2 x i64> @test_vqdmlsl_high_n_s32_imm(<2 x i64> %a, <4 x i32> %b) #0 {
; CHECK-LABEL: test_vqdmlsl_high_n_s32_imm:
; CHECK-NEXT: movi [[REPLICATE:v[0-9]+]].4s, #29
; CHECK-NEXT: sqdmlsl2 {{v[0-9]+}}.2d, {{v[0-9]+}}.4s, [[REPLICATE]].4s
; CHECK-NEXT: ret
entry:
  %shuffle.i.i = shufflevector <4 x i32> %b, <4 x i32> undef, <2 x i32> <i32 2, i32 3>
  %vqdmlsl9.i.i = call <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32> %shuffle.i.i, <2 x i32> <i32 29, i32 29>)
  %vqdmlsl11.i.i = call <2 x i64> @llvm.aarch64.neon.sqsub.v2i64(<2 x i64> %a, <2 x i64> %vqdmlsl9.i.i)
  ret <2 x i64> %vqdmlsl11.i.i
}

define <2 x float> @test_vmul_n_f32(<2 x float> %a, float %b) #0 {
; CHECK-LABEL: test_vmul_n_f32:
; CHECK-NEXT: fmul {{v[0-9]+}}.2s, {{v[0-9]+}}.2s, {{v[0-9]+}}.s[0]
; CHECK-NEXT: ret
entry:
  %vecinit.i = insertelement <2 x float> undef, float %b, i32 0
  %vecinit1.i = insertelement <2 x float> %vecinit.i, float %b, i32 1
  %mul.i = fmul <2 x float> %vecinit1.i, %a
  ret <2 x float> %mul.i
}

define <4 x float> @test_vmulq_n_f32(<4 x float> %a, float %b) #0 {
; CHECK-LABEL: test_vmulq_n_f32:
; CHECK-NEXT: fmul {{v[0-9]+}}.4s, {{v[0-9]+}}.4s, {{v[0-9]+}}.s[0]
; CHECK-NEXT: ret
entry:
  %vecinit.i = insertelement <4 x float> undef, float %b, i32 0
  %vecinit1.i = insertelement <4 x float> %vecinit.i, float %b, i32 1
  %vecinit2.i = insertelement <4 x float> %vecinit1.i, float %b, i32 2
  %vecinit3.i = insertelement <4 x float> %vecinit2.i, float %b, i32 3
  %mul.i = fmul <4 x float> %vecinit3.i, %a
  ret <4 x float> %mul.i
}

define <2 x double> @test_vmulq_n_f64(<2 x double> %a, double %b) #0 {
; CHECK-LABEL: test_vmulq_n_f64:
; CHECK-NEXT: fmul {{v[0-9]+}}.2d, {{v[0-9]+}}.2d, {{v[0-9]+}}.d[0]
; CHECK-NEXT: ret
entry:
  %vecinit.i = insertelement <2 x double> undef, double %b, i32 0
  %vecinit1.i = insertelement <2 x double> %vecinit.i, double %b, i32 1
  %mul.i = fmul <2 x double> %vecinit1.i, %a
  ret <2 x double> %mul.i
}

define <2 x float> @test_vfma_n_f32(<2 x float> %a, <2 x float> %b, float %n) #0 {
; CHECK-LABEL: test_vfma_n_f32:
; CHECK-NEXT: fmla {{v[0-9]+}}.2s, {{v[0-9]+}}.2s, {{v[0-9]+}}.s[{{[0-9]+}}]
; CHECK-NEXT: ret
entry:
  %vecinit.i = insertelement <2 x float> undef, float %n, i32 0
  %vecinit1.i = insertelement <2 x float> %vecinit.i, float %n, i32 1
  %0 = call <2 x float> @llvm.fma.v2f32(<2 x float> %b, <2 x float> %vecinit1.i, <2 x float> %a)
  ret <2 x float> %0
}

define <4 x float> @test_vfmaq_n_f32(<4 x float> %a, <4 x float> %b, float %n) #0 {
; CHECK-LABEL: test_vfmaq_n_f32:
; CHECK-NEXT: fmla {{v[0-9]+}}.4s, {{v[0-9]+}}.4s, {{v[0-9]+}}.s[{{[0-9]+}}]
; CHECK-NEXT: ret
entry:
  %vecinit.i = insertelement <4 x float> undef, float %n, i32 0
  %vecinit1.i = insertelement <4 x float> %vecinit.i, float %n, i32 1
  %vecinit2.i = insertelement <4 x float> %vecinit1.i, float %n, i32 2
  %vecinit3.i = insertelement <4 x float> %vecinit2.i, float %n, i32 3
  %0 = call <4 x float> @llvm.fma.v4f32(<4 x float> %b, <4 x float> %vecinit3.i, <4 x float> %a)
  ret <4 x float> %0
}

define <2 x float> @test_vfms_n_f32(<2 x float> %a, <2 x float> %b, float %n) #0 {
; CHECK-LABEL: test_vfms_n_f32:
; CHECK-NEXT: fmls {{v[0-9]+}}.2s, {{v[0-9]+}}.2s, {{v[0-9]+}}.s[{{[0-9]+}}]
; CHECK-NEXT: ret
entry:
  %vecinit.i = insertelement <2 x float> undef, float %n, i32 0
  %vecinit1.i = insertelement <2 x float> %vecinit.i, float %n, i32 1
  %0 = fsub <2 x float> <float -0.000000e+00, float -0.000000e+00>, %b
  %1 = call <2 x float> @llvm.fma.v2f32(<2 x float> %0, <2 x float> %vecinit1.i, <2 x float> %a)
  ret <2 x float> %1
}

define <4 x float> @test_vfmsq_n_f32(<4 x float> %a, <4 x float> %b, float %n) #0 {
; CHECK-LABEL: test_vfmsq_n_f32:
; CHECK-NEXT: fmls {{v[0-9]+}}.4s, {{v[0-9]+}}.4s, {{v[0-9]+}}.s[{{[0-9]+}}]
; CHECK-NEXT: ret
entry:
  %vecinit.i = insertelement <4 x float> undef, float %n, i32 0
  %vecinit1.i = insertelement <4 x float> %vecinit.i, float %n, i32 1
  %vecinit2.i = insertelement <4 x float> %vecinit1.i, float %n, i32 2
  %vecinit3.i = insertelement <4 x float> %vecinit2.i, float %n, i32 3
  %0 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %b
  %1 = call <4 x float> @llvm.fma.v4f32(<4 x float> %0, <4 x float> %vecinit3.i, <4 x float> %a)
  ret <4 x float> %1
}

attributes #0 = { nounwind }

declare <4 x i32> @llvm.aarch64.neon.smull.v4i32(<4 x i16>, <4 x i16>)
declare <2 x i64> @llvm.aarch64.neon.smull.v2i64(<2 x i32>, <2 x i32>)
declare <4 x i32> @llvm.aarch64.neon.umull.v4i32(<4 x i16>, <4 x i16>)
declare <2 x i64> @llvm.aarch64.neon.umull.v2i64(<2 x i32>, <2 x i32>)
declare <4 x i32> @llvm.aarch64.neon.sqdmull.v4i32(<4 x i16>, <4 x i16>)
declare <2 x i64> @llvm.aarch64.neon.sqdmull.v2i64(<2 x i32>, <2 x i32>)
declare <4 x i32> @llvm.aarch64.neon.sqsub.v4i32(<4 x i32>, <4 x i32>)
declare <2 x i64> @llvm.aarch64.neon.sqsub.v2i64(<2 x i64>, <2 x i64>)
declare <4 x i32> @llvm.aarch64.neon.sqadd.v4i32(<4 x i32>, <4 x i32>)
declare <2 x i64> @llvm.aarch64.neon.sqadd.v2i64(<2 x i64>, <2 x i64>)
declare <4 x float> @llvm.fma.v4f32(<4 x float>, <4 x float>, <4 x float>)
declare <2 x float> @llvm.fma.v2f32(<2 x float>, <2 x float>, <2 x float>)