llvm.org GIT mirror llvm / release_80 test / CodeGen / X86 / fma-intrinsics-canonical.ll
release_80

Tree @release_80 (Download .tar.gz)

fma-intrinsics-canonical.ll @release_80raw · 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
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+fma,-fma4 -show-mc-encoding | FileCheck %s --check-prefix=CHECK --check-prefix=CHECK-FMA
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl,-fma4 -show-mc-encoding | FileCheck %s --check-prefix=CHECK --check-prefix=CHECK-AVX512VL
; RUN: llc < %s -mtriple=x86_64-pc-windows -mattr=+fma,-fma4 -show-mc-encoding | FileCheck %s --check-prefix=CHECK --check-prefix=CHECK-FMA-WIN

; NOTE: This should use IR equivalent to what is generated by clang/test/CodeGen/fma-builtins.c

define <4 x float> @test_mm_fmadd_ps(<4 x float> %a, <4 x float> %b, <4 x float> %c) {
; CHECK-FMA-LABEL: test_mm_fmadd_ps:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfmadd213ps %xmm2, %xmm1, %xmm0 # encoding: [0xc4,0xe2,0x71,0xa8,0xc2]
; CHECK-FMA-NEXT:    # xmm0 = (xmm1 * xmm0) + xmm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm_fmadd_ps:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfmadd213ps %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x71,0xa8,0xc2]
; CHECK-AVX512VL-NEXT:    # xmm0 = (xmm1 * xmm0) + xmm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm_fmadd_ps:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovaps (%rcx), %xmm1 # encoding: [0xc5,0xf8,0x28,0x09]
; CHECK-FMA-WIN-NEXT:    vmovaps (%rdx), %xmm0 # encoding: [0xc5,0xf8,0x28,0x02]
; CHECK-FMA-WIN-NEXT:    vfmadd213ps (%r8), %xmm1, %xmm0 # encoding: [0xc4,0xc2,0x71,0xa8,0x00]
; CHECK-FMA-WIN-NEXT:    # xmm0 = (xmm1 * xmm0) + mem
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %0 = tail call <4 x float> @llvm.fma.v4f32(<4 x float> %a, <4 x float> %b, <4 x float> %c) #2
  ret <4 x float> %0
}

define <2 x double> @test_mm_fmadd_pd(<2 x double> %a, <2 x double> %b, <2 x double> %c) {
; CHECK-FMA-LABEL: test_mm_fmadd_pd:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfmadd213pd %xmm2, %xmm1, %xmm0 # encoding: [0xc4,0xe2,0xf1,0xa8,0xc2]
; CHECK-FMA-NEXT:    # xmm0 = (xmm1 * xmm0) + xmm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm_fmadd_pd:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfmadd213pd %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf1,0xa8,0xc2]
; CHECK-AVX512VL-NEXT:    # xmm0 = (xmm1 * xmm0) + xmm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm_fmadd_pd:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovapd (%rcx), %xmm1 # encoding: [0xc5,0xf9,0x28,0x09]
; CHECK-FMA-WIN-NEXT:    vmovapd (%rdx), %xmm0 # encoding: [0xc5,0xf9,0x28,0x02]
; CHECK-FMA-WIN-NEXT:    vfmadd213pd (%r8), %xmm1, %xmm0 # encoding: [0xc4,0xc2,0xf1,0xa8,0x00]
; CHECK-FMA-WIN-NEXT:    # xmm0 = (xmm1 * xmm0) + mem
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %0 = tail call <2 x double> @llvm.fma.v2f64(<2 x double> %a, <2 x double> %b, <2 x double> %c) #2
  ret <2 x double> %0
}

define <4 x float> @test_mm_fmadd_ss(<4 x float> %a, <4 x float> %b, <4 x float> %c) {
; CHECK-FMA-LABEL: test_mm_fmadd_ss:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfmadd213ss %xmm2, %xmm1, %xmm0 # encoding: [0xc4,0xe2,0x71,0xa9,0xc2]
; CHECK-FMA-NEXT:    # xmm0 = (xmm1 * xmm0) + xmm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm_fmadd_ss:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfmadd213ss %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x71,0xa9,0xc2]
; CHECK-AVX512VL-NEXT:    # xmm0 = (xmm1 * xmm0) + xmm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm_fmadd_ss:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovaps (%rcx), %xmm0 # encoding: [0xc5,0xf8,0x28,0x01]
; CHECK-FMA-WIN-NEXT:    vmovss (%r8), %xmm1 # encoding: [0xc4,0xc1,0x7a,0x10,0x08]
; CHECK-FMA-WIN-NEXT:    # xmm1 = mem[0],zero,zero,zero
; CHECK-FMA-WIN-NEXT:    vfmadd132ss (%rdx), %xmm1, %xmm0 # encoding: [0xc4,0xe2,0x71,0x99,0x02]
; CHECK-FMA-WIN-NEXT:    # xmm0 = (xmm0 * mem) + xmm1
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %0 = extractelement <4 x float> %a, i64 0
  %1 = extractelement <4 x float> %b, i64 0
  %2 = extractelement <4 x float> %c, i64 0
  %3 = tail call float @llvm.fma.f32(float %0, float %1, float %2) #2
  %4 = insertelement <4 x float> %a, float %3, i64 0
  ret <4 x float> %4
}

define <2 x double> @test_mm_fmadd_sd(<2 x double> %a, <2 x double> %b, <2 x double> %c) {
; CHECK-FMA-LABEL: test_mm_fmadd_sd:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfmadd213sd %xmm2, %xmm1, %xmm0 # encoding: [0xc4,0xe2,0xf1,0xa9,0xc2]
; CHECK-FMA-NEXT:    # xmm0 = (xmm1 * xmm0) + xmm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm_fmadd_sd:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfmadd213sd %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf1,0xa9,0xc2]
; CHECK-AVX512VL-NEXT:    # xmm0 = (xmm1 * xmm0) + xmm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm_fmadd_sd:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovapd (%rcx), %xmm0 # encoding: [0xc5,0xf9,0x28,0x01]
; CHECK-FMA-WIN-NEXT:    vmovsd (%r8), %xmm1 # encoding: [0xc4,0xc1,0x7b,0x10,0x08]
; CHECK-FMA-WIN-NEXT:    # xmm1 = mem[0],zero
; CHECK-FMA-WIN-NEXT:    vfmadd132sd (%rdx), %xmm1, %xmm0 # encoding: [0xc4,0xe2,0xf1,0x99,0x02]
; CHECK-FMA-WIN-NEXT:    # xmm0 = (xmm0 * mem) + xmm1
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %0 = extractelement <2 x double> %a, i64 0
  %1 = extractelement <2 x double> %b, i64 0
  %2 = extractelement <2 x double> %c, i64 0
  %3 = tail call double @llvm.fma.f64(double %0, double %1, double %2) #2
  %4 = insertelement <2 x double> %a, double %3, i64 0
  ret <2 x double> %4
}

define <4 x float> @test_mm_fmsub_ps(<4 x float> %a, <4 x float> %b, <4 x float> %c) {
; CHECK-FMA-LABEL: test_mm_fmsub_ps:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfmsub213ps %xmm2, %xmm1, %xmm0 # encoding: [0xc4,0xe2,0x71,0xaa,0xc2]
; CHECK-FMA-NEXT:    # xmm0 = (xmm1 * xmm0) - xmm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm_fmsub_ps:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfmsub213ps %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x71,0xaa,0xc2]
; CHECK-AVX512VL-NEXT:    # xmm0 = (xmm1 * xmm0) - xmm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm_fmsub_ps:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovaps (%rcx), %xmm1 # encoding: [0xc5,0xf8,0x28,0x09]
; CHECK-FMA-WIN-NEXT:    vmovaps (%rdx), %xmm0 # encoding: [0xc5,0xf8,0x28,0x02]
; CHECK-FMA-WIN-NEXT:    vfmsub213ps (%r8), %xmm1, %xmm0 # encoding: [0xc4,0xc2,0x71,0xaa,0x00]
; CHECK-FMA-WIN-NEXT:    # xmm0 = (xmm1 * xmm0) - mem
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %sub.i = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %c
  %0 = tail call <4 x float> @llvm.fma.v4f32(<4 x float> %a, <4 x float> %b, <4 x float> %sub.i) #2
  ret <4 x float> %0
}

define <2 x double> @test_mm_fmsub_pd(<2 x double> %a, <2 x double> %b, <2 x double> %c) {
; CHECK-FMA-LABEL: test_mm_fmsub_pd:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfmsub213pd %xmm2, %xmm1, %xmm0 # encoding: [0xc4,0xe2,0xf1,0xaa,0xc2]
; CHECK-FMA-NEXT:    # xmm0 = (xmm1 * xmm0) - xmm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm_fmsub_pd:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfmsub213pd %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf1,0xaa,0xc2]
; CHECK-AVX512VL-NEXT:    # xmm0 = (xmm1 * xmm0) - xmm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm_fmsub_pd:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovapd (%rcx), %xmm1 # encoding: [0xc5,0xf9,0x28,0x09]
; CHECK-FMA-WIN-NEXT:    vmovapd (%rdx), %xmm0 # encoding: [0xc5,0xf9,0x28,0x02]
; CHECK-FMA-WIN-NEXT:    vfmsub213pd (%r8), %xmm1, %xmm0 # encoding: [0xc4,0xc2,0xf1,0xaa,0x00]
; CHECK-FMA-WIN-NEXT:    # xmm0 = (xmm1 * xmm0) - mem
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %sub.i = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %c
  %0 = tail call <2 x double> @llvm.fma.v2f64(<2 x double> %a, <2 x double> %b, <2 x double> %sub.i) #2
  ret <2 x double> %0
}

define <4 x float> @test_mm_fmsub_ss(<4 x float> %a, <4 x float> %b, <4 x float> %c) {
; CHECK-FMA-LABEL: test_mm_fmsub_ss:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfmsub213ss %xmm2, %xmm1, %xmm0 # encoding: [0xc4,0xe2,0x71,0xab,0xc2]
; CHECK-FMA-NEXT:    # xmm0 = (xmm1 * xmm0) - xmm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm_fmsub_ss:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfmsub213ss %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x71,0xab,0xc2]
; CHECK-AVX512VL-NEXT:    # xmm0 = (xmm1 * xmm0) - xmm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm_fmsub_ss:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovaps (%rcx), %xmm0 # encoding: [0xc5,0xf8,0x28,0x01]
; CHECK-FMA-WIN-NEXT:    vmovss (%r8), %xmm1 # encoding: [0xc4,0xc1,0x7a,0x10,0x08]
; CHECK-FMA-WIN-NEXT:    # xmm1 = mem[0],zero,zero,zero
; CHECK-FMA-WIN-NEXT:    vfmsub132ss (%rdx), %xmm1, %xmm0 # encoding: [0xc4,0xe2,0x71,0x9b,0x02]
; CHECK-FMA-WIN-NEXT:    # xmm0 = (xmm0 * mem) - xmm1
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %0 = extractelement <4 x float> %a, i64 0
  %1 = extractelement <4 x float> %b, i64 0
  %.rhs.i = extractelement <4 x float> %c, i64 0
  %2 = fsub float -0.000000e+00, %.rhs.i
  %3 = tail call float @llvm.fma.f32(float %0, float %1, float %2) #2
  %4 = insertelement <4 x float> %a, float %3, i64 0
  ret <4 x float> %4
}

define <2 x double> @test_mm_fmsub_sd(<2 x double> %a, <2 x double> %b, <2 x double> %c) {
; CHECK-FMA-LABEL: test_mm_fmsub_sd:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfmsub213sd %xmm2, %xmm1, %xmm0 # encoding: [0xc4,0xe2,0xf1,0xab,0xc2]
; CHECK-FMA-NEXT:    # xmm0 = (xmm1 * xmm0) - xmm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm_fmsub_sd:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfmsub213sd %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf1,0xab,0xc2]
; CHECK-AVX512VL-NEXT:    # xmm0 = (xmm1 * xmm0) - xmm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm_fmsub_sd:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovapd (%rcx), %xmm0 # encoding: [0xc5,0xf9,0x28,0x01]
; CHECK-FMA-WIN-NEXT:    vmovsd (%r8), %xmm1 # encoding: [0xc4,0xc1,0x7b,0x10,0x08]
; CHECK-FMA-WIN-NEXT:    # xmm1 = mem[0],zero
; CHECK-FMA-WIN-NEXT:    vfmsub132sd (%rdx), %xmm1, %xmm0 # encoding: [0xc4,0xe2,0xf1,0x9b,0x02]
; CHECK-FMA-WIN-NEXT:    # xmm0 = (xmm0 * mem) - xmm1
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %0 = extractelement <2 x double> %a, i64 0
  %1 = extractelement <2 x double> %b, i64 0
  %.rhs.i = extractelement <2 x double> %c, i64 0
  %2 = fsub double -0.000000e+00, %.rhs.i
  %3 = tail call double @llvm.fma.f64(double %0, double %1, double %2) #2
  %4 = insertelement <2 x double> %a, double %3, i64 0
  ret <2 x double> %4
}

define <4 x float> @test_mm_fnmadd_ps(<4 x float> %a, <4 x float> %b, <4 x float> %c) {
; CHECK-FMA-LABEL: test_mm_fnmadd_ps:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfnmadd213ps %xmm2, %xmm1, %xmm0 # encoding: [0xc4,0xe2,0x71,0xac,0xc2]
; CHECK-FMA-NEXT:    # xmm0 = -(xmm1 * xmm0) + xmm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm_fnmadd_ps:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfnmadd213ps %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x71,0xac,0xc2]
; CHECK-AVX512VL-NEXT:    # xmm0 = -(xmm1 * xmm0) + xmm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm_fnmadd_ps:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovaps (%rcx), %xmm1 # encoding: [0xc5,0xf8,0x28,0x09]
; CHECK-FMA-WIN-NEXT:    vmovaps (%rdx), %xmm0 # encoding: [0xc5,0xf8,0x28,0x02]
; CHECK-FMA-WIN-NEXT:    vfnmadd213ps (%r8), %xmm1, %xmm0 # encoding: [0xc4,0xc2,0x71,0xac,0x00]
; CHECK-FMA-WIN-NEXT:    # xmm0 = -(xmm1 * xmm0) + mem
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %sub.i = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %a
  %0 = tail call <4 x float> @llvm.fma.v4f32(<4 x float> %sub.i, <4 x float> %b, <4 x float> %c) #2
  ret <4 x float> %0
}

define <2 x double> @test_mm_fnmadd_pd(<2 x double> %a, <2 x double> %b, <2 x double> %c) {
; CHECK-FMA-LABEL: test_mm_fnmadd_pd:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfnmadd213pd %xmm2, %xmm1, %xmm0 # encoding: [0xc4,0xe2,0xf1,0xac,0xc2]
; CHECK-FMA-NEXT:    # xmm0 = -(xmm1 * xmm0) + xmm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm_fnmadd_pd:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfnmadd213pd %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf1,0xac,0xc2]
; CHECK-AVX512VL-NEXT:    # xmm0 = -(xmm1 * xmm0) + xmm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm_fnmadd_pd:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovapd (%rcx), %xmm1 # encoding: [0xc5,0xf9,0x28,0x09]
; CHECK-FMA-WIN-NEXT:    vmovapd (%rdx), %xmm0 # encoding: [0xc5,0xf9,0x28,0x02]
; CHECK-FMA-WIN-NEXT:    vfnmadd213pd (%r8), %xmm1, %xmm0 # encoding: [0xc4,0xc2,0xf1,0xac,0x00]
; CHECK-FMA-WIN-NEXT:    # xmm0 = -(xmm1 * xmm0) + mem
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %sub.i = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %a
  %0 = tail call <2 x double> @llvm.fma.v2f64(<2 x double> %sub.i, <2 x double> %b, <2 x double> %c) #2
  ret <2 x double> %0
}

define <4 x float> @test_mm_fnmadd_ss(<4 x float> %a, <4 x float> %b, <4 x float> %c) {
; CHECK-FMA-LABEL: test_mm_fnmadd_ss:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfnmadd213ss %xmm2, %xmm1, %xmm0 # encoding: [0xc4,0xe2,0x71,0xad,0xc2]
; CHECK-FMA-NEXT:    # xmm0 = -(xmm1 * xmm0) + xmm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm_fnmadd_ss:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfnmadd213ss %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x71,0xad,0xc2]
; CHECK-AVX512VL-NEXT:    # xmm0 = -(xmm1 * xmm0) + xmm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm_fnmadd_ss:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovaps (%rcx), %xmm0 # encoding: [0xc5,0xf8,0x28,0x01]
; CHECK-FMA-WIN-NEXT:    vmovss (%r8), %xmm1 # encoding: [0xc4,0xc1,0x7a,0x10,0x08]
; CHECK-FMA-WIN-NEXT:    # xmm1 = mem[0],zero,zero,zero
; CHECK-FMA-WIN-NEXT:    vfnmadd132ss (%rdx), %xmm1, %xmm0 # encoding: [0xc4,0xe2,0x71,0x9d,0x02]
; CHECK-FMA-WIN-NEXT:    # xmm0 = -(xmm0 * mem) + xmm1
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %0 = extractelement <4 x float> %a, i64 0
  %.rhs.i = extractelement <4 x float> %b, i64 0
  %1 = fsub float -0.000000e+00, %.rhs.i
  %2 = extractelement <4 x float> %c, i64 0
  %3 = tail call float @llvm.fma.f32(float %0, float %1, float %2) #2
  %4 = insertelement <4 x float> %a, float %3, i64 0
  ret <4 x float> %4
}

define <2 x double> @test_mm_fnmadd_sd(<2 x double> %a, <2 x double> %b, <2 x double> %c) {
; CHECK-FMA-LABEL: test_mm_fnmadd_sd:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfnmadd213sd %xmm2, %xmm1, %xmm0 # encoding: [0xc4,0xe2,0xf1,0xad,0xc2]
; CHECK-FMA-NEXT:    # xmm0 = -(xmm1 * xmm0) + xmm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm_fnmadd_sd:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfnmadd213sd %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf1,0xad,0xc2]
; CHECK-AVX512VL-NEXT:    # xmm0 = -(xmm1 * xmm0) + xmm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm_fnmadd_sd:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovapd (%rcx), %xmm0 # encoding: [0xc5,0xf9,0x28,0x01]
; CHECK-FMA-WIN-NEXT:    vmovsd (%r8), %xmm1 # encoding: [0xc4,0xc1,0x7b,0x10,0x08]
; CHECK-FMA-WIN-NEXT:    # xmm1 = mem[0],zero
; CHECK-FMA-WIN-NEXT:    vfnmadd132sd (%rdx), %xmm1, %xmm0 # encoding: [0xc4,0xe2,0xf1,0x9d,0x02]
; CHECK-FMA-WIN-NEXT:    # xmm0 = -(xmm0 * mem) + xmm1
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %0 = extractelement <2 x double> %a, i64 0
  %.rhs.i = extractelement <2 x double> %b, i64 0
  %1 = fsub double -0.000000e+00, %.rhs.i
  %2 = extractelement <2 x double> %c, i64 0
  %3 = tail call double @llvm.fma.f64(double %0, double %1, double %2) #2
  %4 = insertelement <2 x double> %a, double %3, i64 0
  ret <2 x double> %4
}

define <4 x float> @test_mm_fnmsub_ps(<4 x float> %a, <4 x float> %b, <4 x float> %c) {
; CHECK-FMA-LABEL: test_mm_fnmsub_ps:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfnmsub213ps %xmm2, %xmm1, %xmm0 # encoding: [0xc4,0xe2,0x71,0xae,0xc2]
; CHECK-FMA-NEXT:    # xmm0 = -(xmm1 * xmm0) - xmm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm_fnmsub_ps:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfnmsub213ps %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x71,0xae,0xc2]
; CHECK-AVX512VL-NEXT:    # xmm0 = -(xmm1 * xmm0) - xmm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm_fnmsub_ps:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovaps (%rcx), %xmm1 # encoding: [0xc5,0xf8,0x28,0x09]
; CHECK-FMA-WIN-NEXT:    vmovaps (%rdx), %xmm0 # encoding: [0xc5,0xf8,0x28,0x02]
; CHECK-FMA-WIN-NEXT:    vfnmsub213ps (%r8), %xmm1, %xmm0 # encoding: [0xc4,0xc2,0x71,0xae,0x00]
; CHECK-FMA-WIN-NEXT:    # xmm0 = -(xmm1 * xmm0) - mem
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %sub.i = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %a
  %sub1.i = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %c
  %0 = tail call <4 x float> @llvm.fma.v4f32(<4 x float> %sub.i, <4 x float> %b, <4 x float> %sub1.i) #2
  ret <4 x float> %0
}

define <2 x double> @test_mm_fnmsub_pd(<2 x double> %a, <2 x double> %b, <2 x double> %c) {
; CHECK-FMA-LABEL: test_mm_fnmsub_pd:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfnmsub213pd %xmm2, %xmm1, %xmm0 # encoding: [0xc4,0xe2,0xf1,0xae,0xc2]
; CHECK-FMA-NEXT:    # xmm0 = -(xmm1 * xmm0) - xmm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm_fnmsub_pd:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfnmsub213pd %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf1,0xae,0xc2]
; CHECK-AVX512VL-NEXT:    # xmm0 = -(xmm1 * xmm0) - xmm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm_fnmsub_pd:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovapd (%rcx), %xmm1 # encoding: [0xc5,0xf9,0x28,0x09]
; CHECK-FMA-WIN-NEXT:    vmovapd (%rdx), %xmm0 # encoding: [0xc5,0xf9,0x28,0x02]
; CHECK-FMA-WIN-NEXT:    vfnmsub213pd (%r8), %xmm1, %xmm0 # encoding: [0xc4,0xc2,0xf1,0xae,0x00]
; CHECK-FMA-WIN-NEXT:    # xmm0 = -(xmm1 * xmm0) - mem
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %sub.i = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %a
  %sub1.i = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %c
  %0 = tail call <2 x double> @llvm.fma.v2f64(<2 x double> %sub.i, <2 x double> %b, <2 x double> %sub1.i) #2
  ret <2 x double> %0
}

define <4 x float> @test_mm_fnmsub_ss(<4 x float> %a, <4 x float> %b, <4 x float> %c) {
; CHECK-FMA-LABEL: test_mm_fnmsub_ss:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfnmsub213ss %xmm2, %xmm1, %xmm0 # encoding: [0xc4,0xe2,0x71,0xaf,0xc2]
; CHECK-FMA-NEXT:    # xmm0 = -(xmm1 * xmm0) - xmm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm_fnmsub_ss:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfnmsub213ss %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x71,0xaf,0xc2]
; CHECK-AVX512VL-NEXT:    # xmm0 = -(xmm1 * xmm0) - xmm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm_fnmsub_ss:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovaps (%rcx), %xmm0 # encoding: [0xc5,0xf8,0x28,0x01]
; CHECK-FMA-WIN-NEXT:    vmovss (%r8), %xmm1 # encoding: [0xc4,0xc1,0x7a,0x10,0x08]
; CHECK-FMA-WIN-NEXT:    # xmm1 = mem[0],zero,zero,zero
; CHECK-FMA-WIN-NEXT:    vfnmsub132ss (%rdx), %xmm1, %xmm0 # encoding: [0xc4,0xe2,0x71,0x9f,0x02]
; CHECK-FMA-WIN-NEXT:    # xmm0 = -(xmm0 * mem) - xmm1
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %0 = extractelement <4 x float> %a, i64 0
  %.rhs.i = extractelement <4 x float> %b, i64 0
  %1 = fsub float -0.000000e+00, %.rhs.i
  %.rhs2.i = extractelement <4 x float> %c, i64 0
  %2 = fsub float -0.000000e+00, %.rhs2.i
  %3 = tail call float @llvm.fma.f32(float %0, float %1, float %2) #2
  %4 = insertelement <4 x float> %a, float %3, i64 0
  ret <4 x float> %4
}

define <2 x double> @test_mm_fnmsub_sd(<2 x double> %a, <2 x double> %b, <2 x double> %c) {
; CHECK-FMA-LABEL: test_mm_fnmsub_sd:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfnmsub213sd %xmm2, %xmm1, %xmm0 # encoding: [0xc4,0xe2,0xf1,0xaf,0xc2]
; CHECK-FMA-NEXT:    # xmm0 = -(xmm1 * xmm0) - xmm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm_fnmsub_sd:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfnmsub213sd %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf1,0xaf,0xc2]
; CHECK-AVX512VL-NEXT:    # xmm0 = -(xmm1 * xmm0) - xmm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm_fnmsub_sd:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovapd (%rcx), %xmm0 # encoding: [0xc5,0xf9,0x28,0x01]
; CHECK-FMA-WIN-NEXT:    vmovsd (%r8), %xmm1 # encoding: [0xc4,0xc1,0x7b,0x10,0x08]
; CHECK-FMA-WIN-NEXT:    # xmm1 = mem[0],zero
; CHECK-FMA-WIN-NEXT:    vfnmsub132sd (%rdx), %xmm1, %xmm0 # encoding: [0xc4,0xe2,0xf1,0x9f,0x02]
; CHECK-FMA-WIN-NEXT:    # xmm0 = -(xmm0 * mem) - xmm1
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %0 = extractelement <2 x double> %a, i64 0
  %.rhs.i = extractelement <2 x double> %b, i64 0
  %1 = fsub double -0.000000e+00, %.rhs.i
  %.rhs2.i = extractelement <2 x double> %c, i64 0
  %2 = fsub double -0.000000e+00, %.rhs2.i
  %3 = tail call double @llvm.fma.f64(double %0, double %1, double %2) #2
  %4 = insertelement <2 x double> %a, double %3, i64 0
  ret <2 x double> %4
}

define <4 x float> @test_mm_fmaddsub_ps(<4 x float> %a, <4 x float> %b, <4 x float> %c) {
; CHECK-FMA-LABEL: test_mm_fmaddsub_ps:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfmaddsub213ps %xmm2, %xmm1, %xmm0 # encoding: [0xc4,0xe2,0x71,0xa6,0xc2]
; CHECK-FMA-NEXT:    # xmm0 = (xmm1 * xmm0) +/- xmm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm_fmaddsub_ps:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfmaddsub213ps %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x71,0xa6,0xc2]
; CHECK-AVX512VL-NEXT:    # xmm0 = (xmm1 * xmm0) +/- xmm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm_fmaddsub_ps:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovaps (%rcx), %xmm1 # encoding: [0xc5,0xf8,0x28,0x09]
; CHECK-FMA-WIN-NEXT:    vmovaps (%rdx), %xmm0 # encoding: [0xc5,0xf8,0x28,0x02]
; CHECK-FMA-WIN-NEXT:    vfmaddsub213ps (%r8), %xmm1, %xmm0 # encoding: [0xc4,0xc2,0x71,0xa6,0x00]
; CHECK-FMA-WIN-NEXT:    # xmm0 = (xmm1 * xmm0) +/- mem
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %0 = tail call <4 x float> @llvm.fma.v4f32(<4 x float> %a, <4 x float> %b, <4 x float> %c) #2
  %1 = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %c
  %2 = tail call <4 x float> @llvm.fma.v4f32(<4 x float> %a, <4 x float> %b, <4 x float> %1) #2
  %3 = shufflevector <4 x float> %2, <4 x float> %0, <4 x i32> <i32 0, i32 5, i32 2, i32 7>
  ret <4 x float> %3
}

define <2 x double> @test_mm_fmaddsub_pd(<2 x double> %a, <2 x double> %b, <2 x double> %c) {
; CHECK-FMA-LABEL: test_mm_fmaddsub_pd:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfmaddsub213pd %xmm2, %xmm1, %xmm0 # encoding: [0xc4,0xe2,0xf1,0xa6,0xc2]
; CHECK-FMA-NEXT:    # xmm0 = (xmm1 * xmm0) +/- xmm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm_fmaddsub_pd:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfmaddsub213pd %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf1,0xa6,0xc2]
; CHECK-AVX512VL-NEXT:    # xmm0 = (xmm1 * xmm0) +/- xmm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm_fmaddsub_pd:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovapd (%rcx), %xmm1 # encoding: [0xc5,0xf9,0x28,0x09]
; CHECK-FMA-WIN-NEXT:    vmovapd (%rdx), %xmm0 # encoding: [0xc5,0xf9,0x28,0x02]
; CHECK-FMA-WIN-NEXT:    vfmaddsub213pd (%r8), %xmm1, %xmm0 # encoding: [0xc4,0xc2,0xf1,0xa6,0x00]
; CHECK-FMA-WIN-NEXT:    # xmm0 = (xmm1 * xmm0) +/- mem
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %0 = tail call <2 x double> @llvm.fma.v2f64(<2 x double> %a, <2 x double> %b, <2 x double> %c) #2
  %1 = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %c
  %2 = tail call <2 x double> @llvm.fma.v2f64(<2 x double> %a, <2 x double> %b, <2 x double> %1) #2
  %3 = shufflevector <2 x double> %2, <2 x double> %0, <2 x i32> <i32 0, i32 3>
  ret <2 x double> %3
}

define <4 x float> @test_mm_fmsubadd_ps(<4 x float> %a, <4 x float> %b, <4 x float> %c) {
; CHECK-FMA-LABEL: test_mm_fmsubadd_ps:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfmsubadd213ps %xmm2, %xmm1, %xmm0 # encoding: [0xc4,0xe2,0x71,0xa7,0xc2]
; CHECK-FMA-NEXT:    # xmm0 = (xmm1 * xmm0) -/+ xmm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm_fmsubadd_ps:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfmsubadd213ps %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x71,0xa7,0xc2]
; CHECK-AVX512VL-NEXT:    # xmm0 = (xmm1 * xmm0) -/+ xmm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm_fmsubadd_ps:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovaps (%rcx), %xmm1 # encoding: [0xc5,0xf8,0x28,0x09]
; CHECK-FMA-WIN-NEXT:    vmovaps (%rdx), %xmm0 # encoding: [0xc5,0xf8,0x28,0x02]
; CHECK-FMA-WIN-NEXT:    vfmsubadd213ps (%r8), %xmm1, %xmm0 # encoding: [0xc4,0xc2,0x71,0xa7,0x00]
; CHECK-FMA-WIN-NEXT:    # xmm0 = (xmm1 * xmm0) -/+ mem
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %sub.i = fsub <4 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %c
  %0 = tail call <4 x float> @llvm.fma.v4f32(<4 x float> %a, <4 x float> %b, <4 x float> %sub.i) #2
  %1 = tail call <4 x float> @llvm.fma.v4f32(<4 x float> %a, <4 x float> %b, <4 x float> %c) #2
  %2 = shufflevector <4 x float> %1, <4 x float> %0, <4 x i32> <i32 0, i32 5, i32 2, i32 7>
  ret <4 x float> %2
}

define <2 x double> @test_mm_fmsubadd_pd(<2 x double> %a, <2 x double> %b, <2 x double> %c) {
; CHECK-FMA-LABEL: test_mm_fmsubadd_pd:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfmsubadd213pd %xmm2, %xmm1, %xmm0 # encoding: [0xc4,0xe2,0xf1,0xa7,0xc2]
; CHECK-FMA-NEXT:    # xmm0 = (xmm1 * xmm0) -/+ xmm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm_fmsubadd_pd:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfmsubadd213pd %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf1,0xa7,0xc2]
; CHECK-AVX512VL-NEXT:    # xmm0 = (xmm1 * xmm0) -/+ xmm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm_fmsubadd_pd:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovapd (%rcx), %xmm1 # encoding: [0xc5,0xf9,0x28,0x09]
; CHECK-FMA-WIN-NEXT:    vmovapd (%rdx), %xmm0 # encoding: [0xc5,0xf9,0x28,0x02]
; CHECK-FMA-WIN-NEXT:    vfmsubadd213pd (%r8), %xmm1, %xmm0 # encoding: [0xc4,0xc2,0xf1,0xa7,0x00]
; CHECK-FMA-WIN-NEXT:    # xmm0 = (xmm1 * xmm0) -/+ mem
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %sub.i = fsub <2 x double> <double -0.000000e+00, double -0.000000e+00>, %c
  %0 = tail call <2 x double> @llvm.fma.v2f64(<2 x double> %a, <2 x double> %b, <2 x double> %sub.i) #2
  %1 = tail call <2 x double> @llvm.fma.v2f64(<2 x double> %a, <2 x double> %b, <2 x double> %c) #2
  %2 = shufflevector <2 x double> %1, <2 x double> %0, <2 x i32> <i32 0, i32 3>
  ret <2 x double> %2
}

define <8 x float> @test_mm256_fmadd_ps(<8 x float> %a, <8 x float> %b, <8 x float> %c) {
; CHECK-FMA-LABEL: test_mm256_fmadd_ps:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfmadd213ps %ymm2, %ymm1, %ymm0 # encoding: [0xc4,0xe2,0x75,0xa8,0xc2]
; CHECK-FMA-NEXT:    # ymm0 = (ymm1 * ymm0) + ymm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm256_fmadd_ps:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfmadd213ps %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x75,0xa8,0xc2]
; CHECK-AVX512VL-NEXT:    # ymm0 = (ymm1 * ymm0) + ymm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm256_fmadd_ps:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovaps (%rcx), %ymm1 # encoding: [0xc5,0xfc,0x28,0x09]
; CHECK-FMA-WIN-NEXT:    vmovaps (%rdx), %ymm0 # encoding: [0xc5,0xfc,0x28,0x02]
; CHECK-FMA-WIN-NEXT:    vfmadd213ps (%r8), %ymm1, %ymm0 # encoding: [0xc4,0xc2,0x75,0xa8,0x00]
; CHECK-FMA-WIN-NEXT:    # ymm0 = (ymm1 * ymm0) + mem
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %0 = tail call <8 x float> @llvm.fma.v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %c) #2
  ret <8 x float> %0
}

define <4 x double> @test_mm256_fmadd_pd(<4 x double> %a, <4 x double> %b, <4 x double> %c) {
; CHECK-FMA-LABEL: test_mm256_fmadd_pd:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfmadd213pd %ymm2, %ymm1, %ymm0 # encoding: [0xc4,0xe2,0xf5,0xa8,0xc2]
; CHECK-FMA-NEXT:    # ymm0 = (ymm1 * ymm0) + ymm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm256_fmadd_pd:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfmadd213pd %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf5,0xa8,0xc2]
; CHECK-AVX512VL-NEXT:    # ymm0 = (ymm1 * ymm0) + ymm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm256_fmadd_pd:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovapd (%rcx), %ymm1 # encoding: [0xc5,0xfd,0x28,0x09]
; CHECK-FMA-WIN-NEXT:    vmovapd (%rdx), %ymm0 # encoding: [0xc5,0xfd,0x28,0x02]
; CHECK-FMA-WIN-NEXT:    vfmadd213pd (%r8), %ymm1, %ymm0 # encoding: [0xc4,0xc2,0xf5,0xa8,0x00]
; CHECK-FMA-WIN-NEXT:    # ymm0 = (ymm1 * ymm0) + mem
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %0 = tail call <4 x double> @llvm.fma.v4f64(<4 x double> %a, <4 x double> %b, <4 x double> %c) #2
  ret <4 x double> %0
}

define <8 x float> @test_mm256_fmsub_ps(<8 x float> %a, <8 x float> %b, <8 x float> %c) {
; CHECK-FMA-LABEL: test_mm256_fmsub_ps:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfmsub213ps %ymm2, %ymm1, %ymm0 # encoding: [0xc4,0xe2,0x75,0xaa,0xc2]
; CHECK-FMA-NEXT:    # ymm0 = (ymm1 * ymm0) - ymm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm256_fmsub_ps:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfmsub213ps %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x75,0xaa,0xc2]
; CHECK-AVX512VL-NEXT:    # ymm0 = (ymm1 * ymm0) - ymm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm256_fmsub_ps:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovaps (%rcx), %ymm1 # encoding: [0xc5,0xfc,0x28,0x09]
; CHECK-FMA-WIN-NEXT:    vmovaps (%rdx), %ymm0 # encoding: [0xc5,0xfc,0x28,0x02]
; CHECK-FMA-WIN-NEXT:    vfmsub213ps (%r8), %ymm1, %ymm0 # encoding: [0xc4,0xc2,0x75,0xaa,0x00]
; CHECK-FMA-WIN-NEXT:    # ymm0 = (ymm1 * ymm0) - mem
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %sub.i = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %c
  %0 = tail call <8 x float> @llvm.fma.v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %sub.i) #2
  ret <8 x float> %0
}

define <4 x double> @test_mm256_fmsub_pd(<4 x double> %a, <4 x double> %b, <4 x double> %c) {
; CHECK-FMA-LABEL: test_mm256_fmsub_pd:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfmsub213pd %ymm2, %ymm1, %ymm0 # encoding: [0xc4,0xe2,0xf5,0xaa,0xc2]
; CHECK-FMA-NEXT:    # ymm0 = (ymm1 * ymm0) - ymm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm256_fmsub_pd:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfmsub213pd %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf5,0xaa,0xc2]
; CHECK-AVX512VL-NEXT:    # ymm0 = (ymm1 * ymm0) - ymm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm256_fmsub_pd:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovapd (%rcx), %ymm1 # encoding: [0xc5,0xfd,0x28,0x09]
; CHECK-FMA-WIN-NEXT:    vmovapd (%rdx), %ymm0 # encoding: [0xc5,0xfd,0x28,0x02]
; CHECK-FMA-WIN-NEXT:    vfmsub213pd (%r8), %ymm1, %ymm0 # encoding: [0xc4,0xc2,0xf5,0xaa,0x00]
; CHECK-FMA-WIN-NEXT:    # ymm0 = (ymm1 * ymm0) - mem
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %sub.i = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, %c
  %0 = tail call <4 x double> @llvm.fma.v4f64(<4 x double> %a, <4 x double> %b, <4 x double> %sub.i) #2
  ret <4 x double> %0
}

define <8 x float> @test_mm256_fnmadd_ps(<8 x float> %a, <8 x float> %b, <8 x float> %c) {
; CHECK-FMA-LABEL: test_mm256_fnmadd_ps:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfnmadd213ps %ymm2, %ymm1, %ymm0 # encoding: [0xc4,0xe2,0x75,0xac,0xc2]
; CHECK-FMA-NEXT:    # ymm0 = -(ymm1 * ymm0) + ymm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm256_fnmadd_ps:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfnmadd213ps %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x75,0xac,0xc2]
; CHECK-AVX512VL-NEXT:    # ymm0 = -(ymm1 * ymm0) + ymm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm256_fnmadd_ps:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovaps (%rcx), %ymm1 # encoding: [0xc5,0xfc,0x28,0x09]
; CHECK-FMA-WIN-NEXT:    vmovaps (%rdx), %ymm0 # encoding: [0xc5,0xfc,0x28,0x02]
; CHECK-FMA-WIN-NEXT:    vfnmadd213ps (%r8), %ymm1, %ymm0 # encoding: [0xc4,0xc2,0x75,0xac,0x00]
; CHECK-FMA-WIN-NEXT:    # ymm0 = -(ymm1 * ymm0) + mem
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %sub.i = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %a
  %0 = tail call <8 x float> @llvm.fma.v8f32(<8 x float> %sub.i, <8 x float> %b, <8 x float> %c) #2
  ret <8 x float> %0
}

define <4 x double> @test_mm256_fnmadd_pd(<4 x double> %a, <4 x double> %b, <4 x double> %c) {
; CHECK-FMA-LABEL: test_mm256_fnmadd_pd:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfnmadd213pd %ymm2, %ymm1, %ymm0 # encoding: [0xc4,0xe2,0xf5,0xac,0xc2]
; CHECK-FMA-NEXT:    # ymm0 = -(ymm1 * ymm0) + ymm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm256_fnmadd_pd:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfnmadd213pd %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf5,0xac,0xc2]
; CHECK-AVX512VL-NEXT:    # ymm0 = -(ymm1 * ymm0) + ymm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm256_fnmadd_pd:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovapd (%rcx), %ymm1 # encoding: [0xc5,0xfd,0x28,0x09]
; CHECK-FMA-WIN-NEXT:    vmovapd (%rdx), %ymm0 # encoding: [0xc5,0xfd,0x28,0x02]
; CHECK-FMA-WIN-NEXT:    vfnmadd213pd (%r8), %ymm1, %ymm0 # encoding: [0xc4,0xc2,0xf5,0xac,0x00]
; CHECK-FMA-WIN-NEXT:    # ymm0 = -(ymm1 * ymm0) + mem
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %sub.i = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, %a
  %0 = tail call <4 x double> @llvm.fma.v4f64(<4 x double> %sub.i, <4 x double> %b, <4 x double> %c) #2
  ret <4 x double> %0
}

define <8 x float> @test_mm256_fnmsub_ps(<8 x float> %a, <8 x float> %b, <8 x float> %c) {
; CHECK-FMA-LABEL: test_mm256_fnmsub_ps:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfnmsub213ps %ymm2, %ymm1, %ymm0 # encoding: [0xc4,0xe2,0x75,0xae,0xc2]
; CHECK-FMA-NEXT:    # ymm0 = -(ymm1 * ymm0) - ymm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm256_fnmsub_ps:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfnmsub213ps %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x75,0xae,0xc2]
; CHECK-AVX512VL-NEXT:    # ymm0 = -(ymm1 * ymm0) - ymm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm256_fnmsub_ps:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovaps (%rcx), %ymm1 # encoding: [0xc5,0xfc,0x28,0x09]
; CHECK-FMA-WIN-NEXT:    vmovaps (%rdx), %ymm0 # encoding: [0xc5,0xfc,0x28,0x02]
; CHECK-FMA-WIN-NEXT:    vfnmsub213ps (%r8), %ymm1, %ymm0 # encoding: [0xc4,0xc2,0x75,0xae,0x00]
; CHECK-FMA-WIN-NEXT:    # ymm0 = -(ymm1 * ymm0) - mem
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %sub.i = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %a
  %sub1.i = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %c
  %0 = tail call <8 x float> @llvm.fma.v8f32(<8 x float> %sub.i, <8 x float> %b, <8 x float> %sub1.i) #2
  ret <8 x float> %0
}

define <4 x double> @test_mm256_fnmsub_pd(<4 x double> %a, <4 x double> %b, <4 x double> %c) {
; CHECK-FMA-LABEL: test_mm256_fnmsub_pd:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfnmsub213pd %ymm2, %ymm1, %ymm0 # encoding: [0xc4,0xe2,0xf5,0xae,0xc2]
; CHECK-FMA-NEXT:    # ymm0 = -(ymm1 * ymm0) - ymm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm256_fnmsub_pd:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfnmsub213pd %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf5,0xae,0xc2]
; CHECK-AVX512VL-NEXT:    # ymm0 = -(ymm1 * ymm0) - ymm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm256_fnmsub_pd:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovapd (%rcx), %ymm1 # encoding: [0xc5,0xfd,0x28,0x09]
; CHECK-FMA-WIN-NEXT:    vmovapd (%rdx), %ymm0 # encoding: [0xc5,0xfd,0x28,0x02]
; CHECK-FMA-WIN-NEXT:    vfnmsub213pd (%r8), %ymm1, %ymm0 # encoding: [0xc4,0xc2,0xf5,0xae,0x00]
; CHECK-FMA-WIN-NEXT:    # ymm0 = -(ymm1 * ymm0) - mem
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %sub.i = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, %a
  %sub1.i = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, %c
  %0 = tail call <4 x double> @llvm.fma.v4f64(<4 x double> %sub.i, <4 x double> %b, <4 x double> %sub1.i) #2
  ret <4 x double> %0
}

define <8 x float> @test_mm256_fmaddsub_ps(<8 x float> %a, <8 x float> %b, <8 x float> %c) {
; CHECK-FMA-LABEL: test_mm256_fmaddsub_ps:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfmaddsub213ps %ymm2, %ymm1, %ymm0 # encoding: [0xc4,0xe2,0x75,0xa6,0xc2]
; CHECK-FMA-NEXT:    # ymm0 = (ymm1 * ymm0) +/- ymm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm256_fmaddsub_ps:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfmaddsub213ps %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x75,0xa6,0xc2]
; CHECK-AVX512VL-NEXT:    # ymm0 = (ymm1 * ymm0) +/- ymm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm256_fmaddsub_ps:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovaps (%rcx), %ymm1 # encoding: [0xc5,0xfc,0x28,0x09]
; CHECK-FMA-WIN-NEXT:    vmovaps (%rdx), %ymm0 # encoding: [0xc5,0xfc,0x28,0x02]
; CHECK-FMA-WIN-NEXT:    vfmaddsub213ps (%r8), %ymm1, %ymm0 # encoding: [0xc4,0xc2,0x75,0xa6,0x00]
; CHECK-FMA-WIN-NEXT:    # ymm0 = (ymm1 * ymm0) +/- mem
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %0 = tail call <8 x float> @llvm.fma.v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %c) #2
  %1 = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %c
  %2 = tail call <8 x float> @llvm.fma.v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %1) #2
  %3 = shufflevector <8 x float> %2, <8 x float> %0, <8 x i32> <i32 0, i32 9, i32 2, i32 11, i32 4, i32 13, i32 6, i32 15>
  ret <8 x float> %3
}

define <4 x double> @test_mm256_fmaddsub_pd(<4 x double> %a, <4 x double> %b, <4 x double> %c) {
; CHECK-FMA-LABEL: test_mm256_fmaddsub_pd:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfmaddsub213pd %ymm2, %ymm1, %ymm0 # encoding: [0xc4,0xe2,0xf5,0xa6,0xc2]
; CHECK-FMA-NEXT:    # ymm0 = (ymm1 * ymm0) +/- ymm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm256_fmaddsub_pd:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfmaddsub213pd %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf5,0xa6,0xc2]
; CHECK-AVX512VL-NEXT:    # ymm0 = (ymm1 * ymm0) +/- ymm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm256_fmaddsub_pd:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovapd (%rcx), %ymm1 # encoding: [0xc5,0xfd,0x28,0x09]
; CHECK-FMA-WIN-NEXT:    vmovapd (%rdx), %ymm0 # encoding: [0xc5,0xfd,0x28,0x02]
; CHECK-FMA-WIN-NEXT:    vfmaddsub213pd (%r8), %ymm1, %ymm0 # encoding: [0xc4,0xc2,0xf5,0xa6,0x00]
; CHECK-FMA-WIN-NEXT:    # ymm0 = (ymm1 * ymm0) +/- mem
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %0 = tail call <4 x double> @llvm.fma.v4f64(<4 x double> %a, <4 x double> %b, <4 x double> %c) #2
  %1 = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, %c
  %2 = tail call <4 x double> @llvm.fma.v4f64(<4 x double> %a, <4 x double> %b, <4 x double> %1) #2
  %3 = shufflevector <4 x double> %2, <4 x double> %0, <4 x i32> <i32 0, i32 5, i32 2, i32 7>
  ret <4 x double> %3
}

define <8 x float> @test_mm256_fmsubadd_ps(<8 x float> %a, <8 x float> %b, <8 x float> %c) {
; CHECK-FMA-LABEL: test_mm256_fmsubadd_ps:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfmsubadd213ps %ymm2, %ymm1, %ymm0 # encoding: [0xc4,0xe2,0x75,0xa7,0xc2]
; CHECK-FMA-NEXT:    # ymm0 = (ymm1 * ymm0) -/+ ymm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm256_fmsubadd_ps:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfmsubadd213ps %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x75,0xa7,0xc2]
; CHECK-AVX512VL-NEXT:    # ymm0 = (ymm1 * ymm0) -/+ ymm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm256_fmsubadd_ps:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovaps (%rcx), %ymm1 # encoding: [0xc5,0xfc,0x28,0x09]
; CHECK-FMA-WIN-NEXT:    vmovaps (%rdx), %ymm0 # encoding: [0xc5,0xfc,0x28,0x02]
; CHECK-FMA-WIN-NEXT:    vfmsubadd213ps (%r8), %ymm1, %ymm0 # encoding: [0xc4,0xc2,0x75,0xa7,0x00]
; CHECK-FMA-WIN-NEXT:    # ymm0 = (ymm1 * ymm0) -/+ mem
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %sub.i = fsub <8 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %c
  %0 = tail call <8 x float> @llvm.fma.v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %sub.i) #2
  %1 = tail call <8 x float> @llvm.fma.v8f32(<8 x float> %a, <8 x float> %b, <8 x float> %c) #2
  %2 = shufflevector <8 x float> %1, <8 x float> %0, <8 x i32> <i32 0, i32 9, i32 2, i32 11, i32 4, i32 13, i32 6, i32 15>
  ret <8 x float> %2
}

define <4 x double> @test_mm256_fmsubadd_pd(<4 x double> %a, <4 x double> %b, <4 x double> %c) {
; CHECK-FMA-LABEL: test_mm256_fmsubadd_pd:
; CHECK-FMA:       # %bb.0: # %entry
; CHECK-FMA-NEXT:    vfmsubadd213pd %ymm2, %ymm1, %ymm0 # encoding: [0xc4,0xe2,0xf5,0xa7,0xc2]
; CHECK-FMA-NEXT:    # ymm0 = (ymm1 * ymm0) -/+ ymm2
; CHECK-FMA-NEXT:    retq # encoding: [0xc3]
;
; CHECK-AVX512VL-LABEL: test_mm256_fmsubadd_pd:
; CHECK-AVX512VL:       # %bb.0: # %entry
; CHECK-AVX512VL-NEXT:    vfmsubadd213pd %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0xf5,0xa7,0xc2]
; CHECK-AVX512VL-NEXT:    # ymm0 = (ymm1 * ymm0) -/+ ymm2
; CHECK-AVX512VL-NEXT:    retq # encoding: [0xc3]
;
; CHECK-FMA-WIN-LABEL: test_mm256_fmsubadd_pd:
; CHECK-FMA-WIN:       # %bb.0: # %entry
; CHECK-FMA-WIN-NEXT:    vmovapd (%rcx), %ymm1 # encoding: [0xc5,0xfd,0x28,0x09]
; CHECK-FMA-WIN-NEXT:    vmovapd (%rdx), %ymm0 # encoding: [0xc5,0xfd,0x28,0x02]
; CHECK-FMA-WIN-NEXT:    vfmsubadd213pd (%r8), %ymm1, %ymm0 # encoding: [0xc4,0xc2,0xf5,0xa7,0x00]
; CHECK-FMA-WIN-NEXT:    # ymm0 = (ymm1 * ymm0) -/+ mem
; CHECK-FMA-WIN-NEXT:    retq # encoding: [0xc3]
entry:
  %sub.i = fsub <4 x double> <double -0.000000e+00, double -0.000000e+00, double -0.000000e+00, double -0.000000e+00>, %c
  %0 = tail call <4 x double> @llvm.fma.v4f64(<4 x double> %a, <4 x double> %b, <4 x double> %sub.i) #2
  %1 = tail call <4 x double> @llvm.fma.v4f64(<4 x double> %a, <4 x double> %b, <4 x double> %c) #2
  %2 = shufflevector <4 x double> %1, <4 x double> %0, <4 x i32> <i32 0, i32 5, i32 2, i32 7>
  ret <4 x double> %2
}

declare <4 x float> @llvm.fma.v4f32(<4 x float>, <4 x float>, <4 x float>) #1
declare <2 x double> @llvm.fma.v2f64(<2 x double>, <2 x double>, <2 x double>) #1
declare float @llvm.fma.f32(float, float, float) #1
declare double @llvm.fma.f64(double, double, double) #1
declare <8 x float> @llvm.fma.v8f32(<8 x float>, <8 x float>, <8 x float>) #1
declare <4 x double> @llvm.fma.v4f64(<4 x double>, <4 x double>, <4 x double>) #1