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
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
|
//=- HexagonInstrInfoV5.td - Target Desc. for Hexagon Target -*- tablegen -*-=//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file describes the Hexagon V5 instructions in TableGen format.
//
//===----------------------------------------------------------------------===//
//===----------------------------------------------------------------------===//
// XTYPE/MPY
//===----------------------------------------------------------------------===//
let isCodeGenOnly = 0 in
def S2_asr_i_p_rnd : S_2OpInstImm<"asr", 0b110, 0b111, u6Imm,
[(set I64:$dst,
(sra (i64 (add (i64 (sra I64:$src1, u6ImmPred:$src2)), 1)),
(i32 1)))], 1>,
Requires<[HasV5T]> {
bits<6> src2;
let Inst{13-8} = src2;
}
let isCodeGenOnly = 0 in
def C4_fastcorner9 : T_LOGICAL_2OP<"fastcorner9", 0b000, 0, 0>,
Requires<[HasV5T]> {
let Inst{13,7,4} = 0b111;
}
let isCodeGenOnly = 0 in
def C4_fastcorner9_not : T_LOGICAL_2OP<"!fastcorner9", 0b000, 0, 0>,
Requires<[HasV5T]> {
let Inst{20,13,7,4} = 0b1111;
}
def SDTHexagonFCONST32 : SDTypeProfile<1, 1, [
SDTCisVT<0, f32>,
SDTCisPtrTy<1>]>;
def HexagonFCONST32 : SDNode<"HexagonISD::FCONST32", SDTHexagonFCONST32>;
let isReMaterializable = 1, isMoveImm = 1 in
def FCONST32_nsdata : LDInst<(outs IntRegs:$dst), (ins globaladdress:$global),
"$dst = CONST32(#$global)",
[(set (f32 IntRegs:$dst),
(HexagonFCONST32 tglobaladdr:$global))]>,
Requires<[HasV5T]>;
let isReMaterializable = 1, isMoveImm = 1 in
def CONST64_Float_Real : LDInst<(outs DoubleRegs:$dst), (ins f64imm:$src1),
"$dst = CONST64(#$src1)",
[(set DoubleRegs:$dst, fpimm:$src1)]>,
Requires<[HasV5T]>;
let isReMaterializable = 1, isMoveImm = 1 in
def CONST32_Float_Real : LDInst<(outs IntRegs:$dst), (ins f32imm:$src1),
"$dst = CONST32(#$src1)",
[(set IntRegs:$dst, fpimm:$src1)]>,
Requires<[HasV5T]>;
// Transfer immediate float.
// Only works with single precision fp value.
// For double precision, use CONST64_float_real, as 64bit transfer
// can only hold 40-bit values - 32 from const ext + 8 bit immediate.
// Make sure that complexity is more than the CONST32 pattern in
// HexagonInstrInfo.td patterns.
let isExtended = 1, opExtendable = 1, isMoveImm = 1, isReMaterializable = 1,
isPredicable = 1, AddedComplexity = 30, validSubTargets = HasV5SubT,
isCodeGenOnly = 1 in
def TFRI_f : ALU32_ri<(outs IntRegs:$dst), (ins f32Ext:$src1),
"$dst = #$src1",
[(set IntRegs:$dst, fpimm:$src1)]>,
Requires<[HasV5T]>;
let isExtended = 1, opExtendable = 2, isPredicated = 1,
hasSideEffects = 0, validSubTargets = HasV5SubT in
def TFRI_cPt_f : ALU32_ri<(outs IntRegs:$dst),
(ins PredRegs:$src1, f32Ext:$src2),
"if ($src1) $dst = #$src2",
[]>,
Requires<[HasV5T]>;
let isExtended = 1, opExtendable = 2, isPredicated = 1, isPredicatedFalse = 1,
hasSideEffects = 0, validSubTargets = HasV5SubT in
def TFRI_cNotPt_f : ALU32_ri<(outs IntRegs:$dst),
(ins PredRegs:$src1, f32Ext:$src2),
"if (!$src1) $dst =#$src2",
[]>,
Requires<[HasV5T]>;
def SDTHexagonI32I64: SDTypeProfile<1, 1, [SDTCisVT<0, i32>,
SDTCisVT<1, i64>]>;
def HexagonPOPCOUNT: SDNode<"HexagonISD::POPCOUNT", SDTHexagonI32I64>;
let hasNewValue = 1, validSubTargets = HasV5SubT, isCodeGenOnly = 0 in
def S5_popcountp : ALU64_rr<(outs IntRegs:$Rd), (ins DoubleRegs:$Rss),
"$Rd = popcount($Rss)",
[(set I32:$Rd, (HexagonPOPCOUNT I64:$Rss))], "", S_2op_tc_2_SLOT23>,
Requires<[HasV5T]> {
bits<5> Rd;
bits<5> Rss;
let IClass = 0b1000;
let Inst{27-21} = 0b1000011;
let Inst{7-5} = 0b011;
let Inst{4-0} = Rd;
let Inst{20-16} = Rss;
}
let isFP = 1, hasNewValue = 1, opNewValue = 0 in
class T_MInstFloat <string mnemonic, bits<3> MajOp, bits<3> MinOp>
: MInst<(outs IntRegs:$Rd),
(ins IntRegs:$Rs, IntRegs:$Rt),
"$Rd = "#mnemonic#"($Rs, $Rt)", [],
"" , M_tc_3or4x_SLOT23 > ,
Requires<[HasV5T]> {
bits<5> Rd;
bits<5> Rs;
bits<5> Rt;
let IClass = 0b1110;
let Inst{27-24} = 0b1011;
let Inst{23-21} = MajOp;
let Inst{20-16} = Rs;
let Inst{13} = 0b0;
let Inst{12-8} = Rt;
let Inst{7-5} = MinOp;
let Inst{4-0} = Rd;
}
let isCommutable = 1, isCodeGenOnly = 0 in {
def F2_sfadd : T_MInstFloat < "sfadd", 0b000, 0b000>;
def F2_sfmpy : T_MInstFloat < "sfmpy", 0b010, 0b000>;
}
let isCodeGenOnly = 0 in
def F2_sfsub : T_MInstFloat < "sfsub", 0b000, 0b001>;
let Itinerary = M_tc_3x_SLOT23, isCodeGenOnly = 0 in {
def F2_sfmax : T_MInstFloat < "sfmax", 0b100, 0b000>;
def F2_sfmin : T_MInstFloat < "sfmin", 0b100, 0b001>;
}
let isCodeGenOnly = 0 in {
def F2_sffixupn : T_MInstFloat < "sffixupn", 0b110, 0b000>;
def F2_sffixupd : T_MInstFloat < "sffixupd", 0b110, 0b001>;
}
// F2_sfrecipa: Reciprocal approximation for division.
let isPredicateLate = 1, isFP = 1,
hasSideEffects = 0, hasNewValue = 1, isCodeGenOnly = 0 in
def F2_sfrecipa: MInst <
(outs IntRegs:$Rd, PredRegs:$Pe),
(ins IntRegs:$Rs, IntRegs:$Rt),
"$Rd, $Pe = sfrecipa($Rs, $Rt)">,
Requires<[HasV5T]> {
bits<5> Rd;
bits<2> Pe;
bits<5> Rs;
bits<5> Rt;
let IClass = 0b1110;
let Inst{27-21} = 0b1011111;
let Inst{20-16} = Rs;
let Inst{13} = 0b0;
let Inst{12-8} = Rt;
let Inst{7} = 0b1;
let Inst{6-5} = Pe;
let Inst{4-0} = Rd;
}
// F2_dfcmpeq: Floating point compare for equal.
let isCompare = 1, isFP = 1 in
class T_fcmp <string mnemonic, RegisterClass RC, bits<3> MinOp,
list<dag> pattern = [] >
: ALU64Inst <(outs PredRegs:$dst), (ins RC:$src1, RC:$src2),
"$dst = "#mnemonic#"($src1, $src2)", pattern,
"" , ALU64_tc_2early_SLOT23 > ,
Requires<[HasV5T]> {
bits<2> dst;
bits<5> src1;
bits<5> src2;
let IClass = 0b1101;
let Inst{27-21} = 0b0010111;
let Inst{20-16} = src1;
let Inst{12-8} = src2;
let Inst{7-5} = MinOp;
let Inst{1-0} = dst;
}
class T_fcmp64 <string mnemonic, PatFrag OpNode, bits<3> MinOp>
: T_fcmp <mnemonic, DoubleRegs, MinOp,
[(set I1:$dst, (OpNode F64:$src1, F64:$src2))]> {
let IClass = 0b1101;
let Inst{27-21} = 0b0010111;
}
class T_fcmp32 <string mnemonic, PatFrag OpNode, bits<3> MinOp>
: T_fcmp <mnemonic, IntRegs, MinOp,
[(set I1:$dst, (OpNode F32:$src1, F32:$src2))]> {
let IClass = 0b1100;
let Inst{27-21} = 0b0111111;
}
let isCodeGenOnly = 0 in {
def F2_dfcmpeq : T_fcmp64<"dfcmp.eq", setoeq, 0b000>;
def F2_dfcmpgt : T_fcmp64<"dfcmp.gt", setogt, 0b001>;
def F2_dfcmpge : T_fcmp64<"dfcmp.ge", setoge, 0b010>;
def F2_dfcmpuo : T_fcmp64<"dfcmp.uo", setuo, 0b011>;
def F2_sfcmpge : T_fcmp32<"sfcmp.ge", setoge, 0b000>;
def F2_sfcmpuo : T_fcmp32<"sfcmp.uo", setuo, 0b001>;
def F2_sfcmpeq : T_fcmp32<"sfcmp.eq", setoeq, 0b011>;
def F2_sfcmpgt : T_fcmp32<"sfcmp.gt", setogt, 0b100>;
}
// F2 convert template classes:
let isFP = 1 in
class F2_RDD_RSS_CONVERT<string mnemonic, bits<3> MinOp,
SDNode Op, PatLeaf RCOut, PatLeaf RCIn,
string chop ="">
: SInst <(outs DoubleRegs:$Rdd), (ins DoubleRegs:$Rss),
"$Rdd = "#mnemonic#"($Rss)"#chop,
[(set RCOut:$Rdd, (Op RCIn:$Rss))], "",
S_2op_tc_3or4x_SLOT23> {
bits<5> Rdd;
bits<5> Rss;
let IClass = 0b1000;
let Inst{27-21} = 0b0000111;
let Inst{20-16} = Rss;
let Inst{7-5} = MinOp;
let Inst{4-0} = Rdd;
}
let isFP = 1 in
class F2_RDD_RS_CONVERT<string mnemonic, bits<3> MinOp,
SDNode Op, PatLeaf RCOut, PatLeaf RCIn,
string chop ="">
: SInst <(outs DoubleRegs:$Rdd), (ins IntRegs:$Rs),
"$Rdd = "#mnemonic#"($Rs)"#chop,
[(set RCOut:$Rdd, (Op RCIn:$Rs))], "",
S_2op_tc_3or4x_SLOT23> {
bits<5> Rdd;
bits<5> Rs;
let IClass = 0b1000;
let Inst{27-21} = 0b0100100;
let Inst{20-16} = Rs;
let Inst{7-5} = MinOp;
let Inst{4-0} = Rdd;
}
let isFP = 1, hasNewValue = 1 in
class F2_RD_RSS_CONVERT<string mnemonic, bits<3> MinOp,
SDNode Op, PatLeaf RCOut, PatLeaf RCIn,
string chop ="">
: SInst <(outs IntRegs:$Rd), (ins DoubleRegs:$Rss),
"$Rd = "#mnemonic#"($Rss)"#chop,
[(set RCOut:$Rd, (Op RCIn:$Rss))], "",
S_2op_tc_3or4x_SLOT23> {
bits<5> Rd;
bits<5> Rss;
let IClass = 0b1000;
let Inst{27-24} = 0b1000;
let Inst{23-21} = MinOp;
let Inst{20-16} = Rss;
let Inst{7-5} = 0b001;
let Inst{4-0} = Rd;
}
let isFP = 1, hasNewValue = 1 in
class F2_RD_RS_CONVERT<string mnemonic, bits<3> MajOp, bits<3> MinOp,
SDNode Op, PatLeaf RCOut, PatLeaf RCIn,
string chop ="">
: SInst <(outs IntRegs:$Rd), (ins IntRegs:$Rs),
"$Rd = "#mnemonic#"($Rs)"#chop,
[(set RCOut:$Rd, (Op RCIn:$Rs))], "",
S_2op_tc_3or4x_SLOT23> {
bits<5> Rd;
bits<5> Rs;
let IClass = 0b1000;
let Inst{27-24} = 0b1011;
let Inst{23-21} = MajOp;
let Inst{20-16} = Rs;
let Inst{7-5} = MinOp;
let Inst{4-0} = Rd;
}
// Convert single precision to double precision and vice-versa.
let isCodeGenOnly = 0 in {
def F2_conv_sf2df : F2_RDD_RS_CONVERT <"convert_sf2df", 0b000,
fextend, F64, F32>;
def F2_conv_df2sf : F2_RD_RSS_CONVERT <"convert_df2sf", 0b000,
fround, F32, F64>;
// Convert Integer to Floating Point.
def F2_conv_d2sf : F2_RD_RSS_CONVERT <"convert_d2sf", 0b010,
sint_to_fp, F32, I64>;
def F2_conv_ud2sf : F2_RD_RSS_CONVERT <"convert_ud2sf", 0b001,
uint_to_fp, F32, I64>;
def F2_conv_uw2sf : F2_RD_RS_CONVERT <"convert_uw2sf", 0b001, 0b000,
uint_to_fp, F32, I32>;
def F2_conv_w2sf : F2_RD_RS_CONVERT <"convert_w2sf", 0b010, 0b000,
sint_to_fp, F32, I32>;
def F2_conv_d2df : F2_RDD_RSS_CONVERT <"convert_d2df", 0b011,
sint_to_fp, F64, I64>;
def F2_conv_ud2df : F2_RDD_RSS_CONVERT <"convert_ud2df", 0b010,
uint_to_fp, F64, I64>;
def F2_conv_uw2df : F2_RDD_RS_CONVERT <"convert_uw2df", 0b001,
uint_to_fp, F64, I32>;
def F2_conv_w2df : F2_RDD_RS_CONVERT <"convert_w2df", 0b010,
sint_to_fp, F64, I32>;
// Convert Floating Point to Integer - default.
def F2_conv_df2uw_chop : F2_RD_RSS_CONVERT <"convert_df2uw", 0b101,
fp_to_uint, I32, F64, ":chop">;
def F2_conv_df2w_chop : F2_RD_RSS_CONVERT <"convert_df2w", 0b111,
fp_to_sint, I32, F64, ":chop">;
def F2_conv_sf2uw_chop : F2_RD_RS_CONVERT <"convert_sf2uw", 0b011, 0b001,
fp_to_uint, I32, F32, ":chop">;
def F2_conv_sf2w_chop : F2_RD_RS_CONVERT <"convert_sf2w", 0b100, 0b001,
fp_to_sint, I32, F32, ":chop">;
def F2_conv_df2d_chop : F2_RDD_RSS_CONVERT <"convert_df2d", 0b110,
fp_to_sint, I64, F64, ":chop">;
def F2_conv_df2ud_chop : F2_RDD_RSS_CONVERT <"convert_df2ud", 0b111,
fp_to_uint, I64, F64, ":chop">;
def F2_conv_sf2d_chop : F2_RDD_RS_CONVERT <"convert_sf2d", 0b110,
fp_to_sint, I64, F32, ":chop">;
def F2_conv_sf2ud_chop : F2_RDD_RS_CONVERT <"convert_sf2ud", 0b101,
fp_to_uint, I64, F32, ":chop">;
// Convert Floating Point to Integer: non-chopped.
let AddedComplexity = 20, Predicates = [HasV5T, IEEERndNearV5T] in {
def F2_conv_df2d : F2_RDD_RSS_CONVERT <"convert_df2d", 0b000,
fp_to_sint, I64, F64>;
def F2_conv_df2ud : F2_RDD_RSS_CONVERT <"convert_df2ud", 0b001,
fp_to_uint, I64, F64>;
def F2_conv_sf2ud : F2_RDD_RS_CONVERT <"convert_sf2ud", 0b011,
fp_to_uint, I64, F32>;
def F2_conv_sf2d : F2_RDD_RS_CONVERT <"convert_sf2d", 0b100,
fp_to_sint, I64, F32>;
def F2_conv_df2uw : F2_RD_RSS_CONVERT <"convert_df2uw", 0b011,
fp_to_uint, I32, F64>;
def F2_conv_df2w : F2_RD_RSS_CONVERT <"convert_df2w", 0b100,
fp_to_sint, I32, F64>;
def F2_conv_sf2uw : F2_RD_RS_CONVERT <"convert_sf2uw", 0b011, 0b000,
fp_to_uint, I32, F32>;
def F2_conv_sf2w : F2_RD_RS_CONVERT <"convert_sf2w", 0b100, 0b000,
fp_to_sint, I32, F32>;
}
}
// Fix up radicand.
let isFP = 1, hasNewValue = 1, isCodeGenOnly = 0 in
def F2_sffixupr: SInst<(outs IntRegs:$Rd), (ins IntRegs:$Rs),
"$Rd = sffixupr($Rs)",
[], "" , S_2op_tc_3or4x_SLOT23>, Requires<[HasV5T]> {
bits<5> Rd;
bits<5> Rs;
let IClass = 0b1000;
let Inst{27-21} = 0b1011101;
let Inst{20-16} = Rs;
let Inst{7-5} = 0b000;
let Inst{4-0} = Rd;
}
// F2_sffma: Floating-point fused multiply add.
let isFP = 1, hasNewValue = 1 in
class T_sfmpy_acc <bit isSub, bit isLib>
: MInst<(outs IntRegs:$Rx),
(ins IntRegs:$dst2, IntRegs:$Rs, IntRegs:$Rt),
"$Rx "#!if(isSub, "-=","+=")#" sfmpy($Rs, $Rt)"#!if(isLib, ":lib",""),
[], "$dst2 = $Rx" , M_tc_3_SLOT23 > ,
Requires<[HasV5T]> {
bits<5> Rx;
bits<5> Rs;
bits<5> Rt;
let IClass = 0b1110;
let Inst{27-21} = 0b1111000;
let Inst{20-16} = Rs;
let Inst{13} = 0b0;
let Inst{12-8} = Rt;
let Inst{7} = 0b1;
let Inst{6} = isLib;
let Inst{5} = isSub;
let Inst{4-0} = Rx;
}
let isCodeGenOnly = 0 in {
def F2_sffma: T_sfmpy_acc <0, 0>;
def F2_sffms: T_sfmpy_acc <1, 0>;
def F2_sffma_lib: T_sfmpy_acc <0, 1>;
def F2_sffms_lib: T_sfmpy_acc <1, 1>;
}
// Floating-point fused multiply add w/ additional scaling (2**pu).
let isFP = 1, hasNewValue = 1, isCodeGenOnly = 0 in
def F2_sffma_sc: MInst <
(outs IntRegs:$Rx),
(ins IntRegs:$dst2, IntRegs:$Rs, IntRegs:$Rt, PredRegs:$Pu),
"$Rx += sfmpy($Rs, $Rt, $Pu):scale" ,
[], "$dst2 = $Rx" , M_tc_3_SLOT23 > ,
Requires<[HasV5T]> {
bits<5> Rx;
bits<5> Rs;
bits<5> Rt;
bits<2> Pu;
let IClass = 0b1110;
let Inst{27-21} = 0b1111011;
let Inst{20-16} = Rs;
let Inst{13} = 0b0;
let Inst{12-8} = Rt;
let Inst{7} = 0b1;
let Inst{6-5} = Pu;
let Inst{4-0} = Rx;
}
// Classify floating-point value
let isFP = 1, isCodeGenOnly = 0 in
def F2_sfclass : T_TEST_BIT_IMM<"sfclass", 0b111>;
let isFP = 1, isCodeGenOnly = 0 in
def F2_dfclass: ALU64Inst<(outs PredRegs:$Pd), (ins DoubleRegs:$Rss, u5Imm:$u5),
"$Pd = dfclass($Rss, #$u5)",
[], "" , ALU64_tc_2early_SLOT23 > , Requires<[HasV5T]> {
bits<2> Pd;
bits<5> Rss;
bits<5> u5;
let IClass = 0b1101;
let Inst{27-21} = 0b1100100;
let Inst{20-16} = Rss;
let Inst{12-10} = 0b000;
let Inst{9-5} = u5;
let Inst{4-3} = 0b10;
let Inst{1-0} = Pd;
}
// Instructions to create floating point constant
let hasNewValue = 1, opNewValue = 0 in
class T_fimm <string mnemonic, RegisterClass RC, bits<4> RegType, bit isNeg>
: ALU64Inst<(outs RC:$dst), (ins u10Imm:$src),
"$dst = "#mnemonic#"(#$src)"#!if(isNeg, ":neg", ":pos"),
[], "", ALU64_tc_3x_SLOT23>, Requires<[HasV5T]> {
bits<5> dst;
bits<10> src;
let IClass = 0b1101;
let Inst{27-24} = RegType;
let Inst{23} = 0b0;
let Inst{22} = isNeg;
let Inst{21} = src{9};
let Inst{13-5} = src{8-0};
let Inst{4-0} = dst;
}
let isCodeGenOnly = 0 in {
def F2_sfimm_p : T_fimm <"sfmake", IntRegs, 0b0110, 0>;
def F2_sfimm_n : T_fimm <"sfmake", IntRegs, 0b0110, 1>;
def F2_dfimm_p : T_fimm <"dfmake", DoubleRegs, 0b1001, 0>;
def F2_dfimm_n : T_fimm <"dfmake", DoubleRegs, 0b1001, 1>;
}
// Convert single precision to double precision and vice-versa.
def CONVERT_sf2df : ALU64_rr<(outs DoubleRegs:$dst), (ins IntRegs:$src),
"$dst = convert_sf2df($src)",
[(set DoubleRegs:$dst, (fextend IntRegs:$src))]>,
Requires<[HasV5T]>;
def CONVERT_df2sf : ALU64_rr<(outs IntRegs:$dst), (ins DoubleRegs:$src),
"$dst = convert_df2sf($src)",
[(set IntRegs:$dst, (fround DoubleRegs:$src))]>,
Requires<[HasV5T]>;
// Load.
def LDrid_f : LDInst<(outs DoubleRegs:$dst),
(ins MEMri:$addr),
"$dst = memd($addr)",
[(set DoubleRegs:$dst, (f64 (load ADDRriS11_3:$addr)))]>,
Requires<[HasV5T]>;
let AddedComplexity = 20 in
def LDrid_indexed_f : LDInst<(outs DoubleRegs:$dst),
(ins IntRegs:$src1, s11_3Imm:$offset),
"$dst = memd($src1+#$offset)",
[(set DoubleRegs:$dst, (f64 (load (add IntRegs:$src1,
s11_3ImmPred:$offset))))]>,
Requires<[HasV5T]>;
def LDriw_f : LDInst<(outs IntRegs:$dst),
(ins MEMri:$addr), "$dst = memw($addr)",
[(set IntRegs:$dst, (f32 (load ADDRriS11_2:$addr)))]>,
Requires<[HasV5T]>;
let AddedComplexity = 20 in
def LDriw_indexed_f : LDInst<(outs IntRegs:$dst),
(ins IntRegs:$src1, s11_2Imm:$offset),
"$dst = memw($src1+#$offset)",
[(set IntRegs:$dst, (f32 (load (add IntRegs:$src1,
s11_2ImmPred:$offset))))]>,
Requires<[HasV5T]>;
// Store.
def STriw_f : STInst<(outs),
(ins MEMri:$addr, IntRegs:$src1),
"memw($addr) = $src1",
[(store (f32 IntRegs:$src1), ADDRriS11_2:$addr)]>,
Requires<[HasV5T]>;
let AddedComplexity = 10 in
def STriw_indexed_f : STInst<(outs),
(ins IntRegs:$src1, s11_2Imm:$src2, IntRegs:$src3),
"memw($src1+#$src2) = $src3",
[(store (f32 IntRegs:$src3),
(add IntRegs:$src1, s11_2ImmPred:$src2))]>,
Requires<[HasV5T]>;
def STrid_f : STInst<(outs),
(ins MEMri:$addr, DoubleRegs:$src1),
"memd($addr) = $src1",
[(store (f64 DoubleRegs:$src1), ADDRriS11_2:$addr)]>,
Requires<[HasV5T]>;
// Indexed store double word.
let AddedComplexity = 10 in
def STrid_indexed_f : STInst<(outs),
(ins IntRegs:$src1, s11_3Imm:$src2, DoubleRegs:$src3),
"memd($src1+#$src2) = $src3",
[(store (f64 DoubleRegs:$src3),
(add IntRegs:$src1, s11_3ImmPred:$src2))]>,
Requires<[HasV5T]>;
// Add
let isCommutable = 1 in
def fADD_rr : ALU64_rr<(outs IntRegs:$dst),
(ins IntRegs:$src1, IntRegs:$src2),
"$dst = sfadd($src1, $src2)",
[(set IntRegs:$dst, (fadd IntRegs:$src1, IntRegs:$src2))]>,
Requires<[HasV5T]>;
let isCommutable = 1 in
def fADD64_rr : ALU64_rr<(outs DoubleRegs:$dst), (ins DoubleRegs:$src1,
DoubleRegs:$src2),
"$dst = dfadd($src1, $src2)",
[(set DoubleRegs:$dst, (fadd DoubleRegs:$src1,
DoubleRegs:$src2))]>,
Requires<[HasV5T]>;
def fSUB_rr : ALU64_rr<(outs IntRegs:$dst),
(ins IntRegs:$src1, IntRegs:$src2),
"$dst = sfsub($src1, $src2)",
[(set IntRegs:$dst, (fsub IntRegs:$src1, IntRegs:$src2))]>,
Requires<[HasV5T]>;
def fSUB64_rr : ALU64_rr<(outs DoubleRegs:$dst), (ins DoubleRegs:$src1,
DoubleRegs:$src2),
"$dst = dfsub($src1, $src2)",
[(set DoubleRegs:$dst, (fsub DoubleRegs:$src1,
DoubleRegs:$src2))]>,
Requires<[HasV5T]>;
let isCommutable = 1 in
def fMUL_rr : ALU64_rr<(outs IntRegs:$dst),
(ins IntRegs:$src1, IntRegs:$src2),
"$dst = sfmpy($src1, $src2)",
[(set IntRegs:$dst, (fmul IntRegs:$src1, IntRegs:$src2))]>,
Requires<[HasV5T]>;
let isCommutable = 1 in
def fMUL64_rr : ALU64_rr<(outs DoubleRegs:$dst), (ins DoubleRegs:$src1,
DoubleRegs:$src2),
"$dst = dfmpy($src1, $src2)",
[(set DoubleRegs:$dst, (fmul DoubleRegs:$src1,
DoubleRegs:$src2))]>,
Requires<[HasV5T]>;
// Compare.
let isCompare = 1 in {
multiclass FCMP64_rr<string OpcStr, PatFrag OpNode> {
def _rr : ALU64_rr<(outs PredRegs:$dst), (ins DoubleRegs:$b, DoubleRegs:$c),
!strconcat("$dst = ", !strconcat(OpcStr, "($b, $c)")),
[(set PredRegs:$dst,
(OpNode (f64 DoubleRegs:$b), (f64 DoubleRegs:$c)))]>,
Requires<[HasV5T]>;
}
multiclass FCMP32_rr<string OpcStr, PatFrag OpNode> {
def _rr : ALU64_rr<(outs PredRegs:$dst), (ins IntRegs:$b, IntRegs:$c),
!strconcat("$dst = ", !strconcat(OpcStr, "($b, $c)")),
[(set PredRegs:$dst,
(OpNode (f32 IntRegs:$b), (f32 IntRegs:$c)))]>,
Requires<[HasV5T]>;
}
}
defm FCMPOEQ64 : FCMP64_rr<"dfcmp.eq", setoeq>;
defm FCMPUEQ64 : FCMP64_rr<"dfcmp.eq", setueq>;
defm FCMPOGT64 : FCMP64_rr<"dfcmp.gt", setogt>;
defm FCMPUGT64 : FCMP64_rr<"dfcmp.gt", setugt>;
defm FCMPOGE64 : FCMP64_rr<"dfcmp.ge", setoge>;
defm FCMPUGE64 : FCMP64_rr<"dfcmp.ge", setuge>;
defm FCMPOEQ32 : FCMP32_rr<"sfcmp.eq", setoeq>;
defm FCMPUEQ32 : FCMP32_rr<"sfcmp.eq", setueq>;
defm FCMPOGT32 : FCMP32_rr<"sfcmp.gt", setogt>;
defm FCMPUGT32 : FCMP32_rr<"sfcmp.gt", setugt>;
defm FCMPOGE32 : FCMP32_rr<"sfcmp.ge", setoge>;
defm FCMPUGE32 : FCMP32_rr<"sfcmp.ge", setuge>;
// olt.
def : Pat <(i1 (setolt (f32 IntRegs:$src1), (f32 IntRegs:$src2))),
(i1 (FCMPOGT32_rr IntRegs:$src2, IntRegs:$src1))>,
Requires<[HasV5T]>;
def : Pat <(i1 (setolt (f32 IntRegs:$src1), (fpimm:$src2))),
(i1 (FCMPOGT32_rr (f32 (TFRI_f fpimm:$src2)), (f32 IntRegs:$src1)))>,
Requires<[HasV5T]>;
def : Pat <(i1 (setolt (f64 DoubleRegs:$src1), (f64 DoubleRegs:$src2))),
(i1 (FCMPOGT64_rr DoubleRegs:$src2, DoubleRegs:$src1))>,
Requires<[HasV5T]>;
def : Pat <(i1 (setolt (f64 DoubleRegs:$src1), (fpimm:$src2))),
(i1 (FCMPOGT64_rr (f64 (CONST64_Float_Real fpimm:$src2)),
(f64 DoubleRegs:$src1)))>,
Requires<[HasV5T]>;
// gt.
def : Pat <(i1 (setugt (f64 DoubleRegs:$src1), (fpimm:$src2))),
(i1 (FCMPUGT64_rr (f64 DoubleRegs:$src1),
(f64 (CONST64_Float_Real fpimm:$src2))))>,
Requires<[HasV5T]>;
def : Pat <(i1 (setugt (f32 IntRegs:$src1), (fpimm:$src2))),
(i1 (FCMPUGT32_rr (f32 IntRegs:$src1), (f32 (TFRI_f fpimm:$src2))))>,
Requires<[HasV5T]>;
// ult.
def : Pat <(i1 (setult (f32 IntRegs:$src1), (f32 IntRegs:$src2))),
(i1 (FCMPUGT32_rr IntRegs:$src2, IntRegs:$src1))>,
Requires<[HasV5T]>;
def : Pat <(i1 (setult (f32 IntRegs:$src1), (fpimm:$src2))),
(i1 (FCMPUGT32_rr (f32 (TFRI_f fpimm:$src2)), (f32 IntRegs:$src1)))>,
Requires<[HasV5T]>;
def : Pat <(i1 (setult (f64 DoubleRegs:$src1), (f64 DoubleRegs:$src2))),
(i1 (FCMPUGT64_rr DoubleRegs:$src2, DoubleRegs:$src1))>,
Requires<[HasV5T]>;
def : Pat <(i1 (setult (f64 DoubleRegs:$src1), (fpimm:$src2))),
(i1 (FCMPUGT64_rr (f64 (CONST64_Float_Real fpimm:$src2)),
(f64 DoubleRegs:$src1)))>,
Requires<[HasV5T]>;
// le.
// rs <= rt -> rt >= rs.
def : Pat<(i1 (setole (f32 IntRegs:$src1), (f32 IntRegs:$src2))),
(i1 (FCMPOGE32_rr IntRegs:$src2, IntRegs:$src1))>,
Requires<[HasV5T]>;
def : Pat<(i1 (setole (f32 IntRegs:$src1), (fpimm:$src2))),
(i1 (FCMPOGE32_rr (f32 (TFRI_f fpimm:$src2)), IntRegs:$src1))>,
Requires<[HasV5T]>;
// Rss <= Rtt -> Rtt >= Rss.
def : Pat<(i1 (setole (f64 DoubleRegs:$src1), (f64 DoubleRegs:$src2))),
(i1 (FCMPOGE64_rr DoubleRegs:$src2, DoubleRegs:$src1))>,
Requires<[HasV5T]>;
def : Pat<(i1 (setole (f64 DoubleRegs:$src1), (fpimm:$src2))),
(i1 (FCMPOGE64_rr (f64 (CONST64_Float_Real fpimm:$src2)),
DoubleRegs:$src1))>,
Requires<[HasV5T]>;
// rs <= rt -> rt >= rs.
def : Pat<(i1 (setule (f32 IntRegs:$src1), (f32 IntRegs:$src2))),
(i1 (FCMPUGE32_rr IntRegs:$src2, IntRegs:$src1))>,
Requires<[HasV5T]>;
def : Pat<(i1 (setule (f32 IntRegs:$src1), (fpimm:$src2))),
(i1 (FCMPUGE32_rr (f32 (TFRI_f fpimm:$src2)), IntRegs:$src1))>,
Requires<[HasV5T]>;
// Rss <= Rtt -> Rtt >= Rss.
def : Pat<(i1 (setule (f64 DoubleRegs:$src1), (f64 DoubleRegs:$src2))),
(i1 (FCMPUGE64_rr DoubleRegs:$src2, DoubleRegs:$src1))>,
Requires<[HasV5T]>;
def : Pat<(i1 (setule (f64 DoubleRegs:$src1), (fpimm:$src2))),
(i1 (FCMPUGE64_rr (f64 (CONST64_Float_Real fpimm:$src2)),
DoubleRegs:$src1))>,
Requires<[HasV5T]>;
// ne.
def : Pat<(i1 (setone (f32 IntRegs:$src1), (f32 IntRegs:$src2))),
(i1 (C2_not (FCMPOEQ32_rr IntRegs:$src1, IntRegs:$src2)))>,
Requires<[HasV5T]>;
def : Pat<(i1 (setone (f64 DoubleRegs:$src1), (f64 DoubleRegs:$src2))),
(i1 (C2_not (FCMPOEQ64_rr DoubleRegs:$src1, DoubleRegs:$src2)))>,
Requires<[HasV5T]>;
def : Pat<(i1 (setune (f32 IntRegs:$src1), (f32 IntRegs:$src2))),
(i1 (C2_not (FCMPUEQ32_rr IntRegs:$src1, IntRegs:$src2)))>,
Requires<[HasV5T]>;
def : Pat<(i1 (setune (f64 DoubleRegs:$src1), (f64 DoubleRegs:$src2))),
(i1 (C2_not (FCMPUEQ64_rr DoubleRegs:$src1, DoubleRegs:$src2)))>,
Requires<[HasV5T]>;
def : Pat<(i1 (setone (f32 IntRegs:$src1), (fpimm:$src2))),
(i1 (C2_not (FCMPOEQ32_rr IntRegs:$src1, (f32 (TFRI_f fpimm:$src2)))))>,
Requires<[HasV5T]>;
def : Pat<(i1 (setone (f64 DoubleRegs:$src1), (fpimm:$src2))),
(i1 (C2_not (FCMPOEQ64_rr DoubleRegs:$src1,
(f64 (CONST64_Float_Real fpimm:$src2)))))>,
Requires<[HasV5T]>;
def : Pat<(i1 (setune (f32 IntRegs:$src1), (fpimm:$src2))),
(i1 (C2_not (FCMPUEQ32_rr IntRegs:$src1, (f32 (TFRI_f fpimm:$src2)))))>,
Requires<[HasV5T]>;
def : Pat<(i1 (setune (f64 DoubleRegs:$src1), (fpimm:$src2))),
(i1 (C2_not (FCMPUEQ64_rr DoubleRegs:$src1,
(f64 (CONST64_Float_Real fpimm:$src2)))))>,
Requires<[HasV5T]>;
// Convert Integer to Floating Point.
def CONVERT_d2sf : ALU64_rr<(outs IntRegs:$dst), (ins DoubleRegs:$src),
"$dst = convert_d2sf($src)",
[(set (f32 IntRegs:$dst), (sint_to_fp (i64 DoubleRegs:$src)))]>,
Requires<[HasV5T]>;
def CONVERT_ud2sf : ALU64_rr<(outs IntRegs:$dst), (ins DoubleRegs:$src),
"$dst = convert_ud2sf($src)",
[(set (f32 IntRegs:$dst), (uint_to_fp (i64 DoubleRegs:$src)))]>,
Requires<[HasV5T]>;
def CONVERT_uw2sf : ALU64_rr<(outs IntRegs:$dst), (ins IntRegs:$src),
"$dst = convert_uw2sf($src)",
[(set (f32 IntRegs:$dst), (uint_to_fp (i32 IntRegs:$src)))]>,
Requires<[HasV5T]>;
def CONVERT_w2sf : ALU64_rr<(outs IntRegs:$dst), (ins IntRegs:$src),
"$dst = convert_w2sf($src)",
[(set (f32 IntRegs:$dst), (sint_to_fp (i32 IntRegs:$src)))]>,
Requires<[HasV5T]>;
def CONVERT_d2df : ALU64_rr<(outs DoubleRegs:$dst), (ins DoubleRegs:$src),
"$dst = convert_d2df($src)",
[(set (f64 DoubleRegs:$dst), (sint_to_fp (i64 DoubleRegs:$src)))]>,
Requires<[HasV5T]>;
def CONVERT_ud2df : ALU64_rr<(outs DoubleRegs:$dst), (ins DoubleRegs:$src),
"$dst = convert_ud2df($src)",
[(set (f64 DoubleRegs:$dst), (uint_to_fp (i64 DoubleRegs:$src)))]>,
Requires<[HasV5T]>;
def CONVERT_uw2df : ALU64_rr<(outs DoubleRegs:$dst), (ins IntRegs:$src),
"$dst = convert_uw2df($src)",
[(set (f64 DoubleRegs:$dst), (uint_to_fp (i32 IntRegs:$src)))]>,
Requires<[HasV5T]>;
def CONVERT_w2df : ALU64_rr<(outs DoubleRegs:$dst), (ins IntRegs:$src),
"$dst = convert_w2df($src)",
[(set (f64 DoubleRegs:$dst), (sint_to_fp (i32 IntRegs:$src)))]>,
Requires<[HasV5T]>;
// Convert Floating Point to Integer - default.
def CONVERT_df2uw : ALU64_rr<(outs IntRegs:$dst), (ins DoubleRegs:$src),
"$dst = convert_df2uw($src):chop",
[(set (i32 IntRegs:$dst), (fp_to_uint (f64 DoubleRegs:$src)))]>,
Requires<[HasV5T]>;
def CONVERT_df2w : ALU64_rr<(outs IntRegs:$dst), (ins DoubleRegs:$src),
"$dst = convert_df2w($src):chop",
[(set (i32 IntRegs:$dst), (fp_to_sint (f64 DoubleRegs:$src)))]>,
Requires<[HasV5T]>;
def CONVERT_sf2uw : ALU64_rr<(outs IntRegs:$dst), (ins IntRegs:$src),
"$dst = convert_sf2uw($src):chop",
[(set (i32 IntRegs:$dst), (fp_to_uint (f32 IntRegs:$src)))]>,
Requires<[HasV5T]>;
def CONVERT_sf2w : ALU64_rr<(outs IntRegs:$dst), (ins IntRegs:$src),
"$dst = convert_sf2w($src):chop",
[(set (i32 IntRegs:$dst), (fp_to_sint (f32 IntRegs:$src)))]>,
Requires<[HasV5T]>;
def CONVERT_df2d : ALU64_rr<(outs DoubleRegs:$dst), (ins DoubleRegs:$src),
"$dst = convert_df2d($src):chop",
[(set (i64 DoubleRegs:$dst), (fp_to_sint (f64 DoubleRegs:$src)))]>,
Requires<[HasV5T]>;
def CONVERT_df2ud : ALU64_rr<(outs DoubleRegs:$dst), (ins DoubleRegs:$src),
"$dst = convert_df2ud($src):chop",
[(set (i64 DoubleRegs:$dst), (fp_to_uint (f64 DoubleRegs:$src)))]>,
Requires<[HasV5T]>;
def CONVERT_sf2d : ALU64_rr<(outs DoubleRegs:$dst), (ins IntRegs:$src),
"$dst = convert_sf2d($src):chop",
[(set (i64 DoubleRegs:$dst), (fp_to_sint (f32 IntRegs:$src)))]>,
Requires<[HasV5T]>;
def CONVERT_sf2ud : ALU64_rr<(outs DoubleRegs:$dst), (ins IntRegs:$src),
"$dst = convert_sf2ud($src):chop",
[(set (i64 DoubleRegs:$dst), (fp_to_uint (f32 IntRegs:$src)))]>,
Requires<[HasV5T]>;
// Convert Floating Point to Integer: non-chopped.
let AddedComplexity = 20 in
def CONVERT_df2uw_nchop : ALU64_rr<(outs IntRegs:$dst), (ins DoubleRegs:$src),
"$dst = convert_df2uw($src)",
[(set (i32 IntRegs:$dst), (fp_to_uint (f64 DoubleRegs:$src)))]>,
Requires<[HasV5T, IEEERndNearV5T]>;
let AddedComplexity = 20 in
def CONVERT_df2w_nchop : ALU64_rr<(outs IntRegs:$dst), (ins DoubleRegs:$src),
"$dst = convert_df2w($src)",
[(set (i32 IntRegs:$dst), (fp_to_sint (f64 DoubleRegs:$src)))]>,
Requires<[HasV5T, IEEERndNearV5T]>;
let AddedComplexity = 20 in
def CONVERT_sf2uw_nchop : ALU64_rr<(outs IntRegs:$dst), (ins IntRegs:$src),
"$dst = convert_sf2uw($src)",
[(set (i32 IntRegs:$dst), (fp_to_uint (f32 IntRegs:$src)))]>,
Requires<[HasV5T, IEEERndNearV5T]>;
let AddedComplexity = 20 in
def CONVERT_sf2w_nchop : ALU64_rr<(outs IntRegs:$dst), (ins IntRegs:$src),
"$dst = convert_sf2w($src)",
[(set (i32 IntRegs:$dst), (fp_to_sint (f32 IntRegs:$src)))]>,
Requires<[HasV5T, IEEERndNearV5T]>;
let AddedComplexity = 20 in
def CONVERT_df2d_nchop : ALU64_rr<(outs DoubleRegs:$dst), (ins DoubleRegs:$src),
"$dst = convert_df2d($src)",
[(set (i64 DoubleRegs:$dst), (fp_to_sint (f64 DoubleRegs:$src)))]>,
Requires<[HasV5T, IEEERndNearV5T]>;
let AddedComplexity = 20 in
def CONVERT_df2ud_nchop : ALU64_rr<(outs DoubleRegs:$dst), (ins DoubleRegs:$src),
"$dst = convert_df2ud($src)",
[(set (i64 DoubleRegs:$dst), (fp_to_uint (f64 DoubleRegs:$src)))]>,
Requires<[HasV5T, IEEERndNearV5T]>;
let AddedComplexity = 20 in
def CONVERT_sf2d_nchop : ALU64_rr<(outs DoubleRegs:$dst), (ins IntRegs:$src),
"$dst = convert_sf2d($src)",
[(set (i64 DoubleRegs:$dst), (fp_to_sint (f32 IntRegs:$src)))]>,
Requires<[HasV5T, IEEERndNearV5T]>;
let AddedComplexity = 20 in
def CONVERT_sf2ud_nchop : ALU64_rr<(outs DoubleRegs:$dst), (ins IntRegs:$src),
"$dst = convert_sf2ud($src)",
[(set (i64 DoubleRegs:$dst), (fp_to_uint (f32 IntRegs:$src)))]>,
Requires<[HasV5T, IEEERndNearV5T]>;
// Bitcast is different than [fp|sint|uint]_to_[sint|uint|fp].
def : Pat <(i32 (bitconvert (f32 IntRegs:$src))),
(i32 (A2_tfr IntRegs:$src))>,
Requires<[HasV5T]>;
def : Pat <(f32 (bitconvert (i32 IntRegs:$src))),
(f32 (A2_tfr IntRegs:$src))>,
Requires<[HasV5T]>;
def : Pat <(i64 (bitconvert (f64 DoubleRegs:$src))),
(i64 (A2_tfrp DoubleRegs:$src))>,
Requires<[HasV5T]>;
def : Pat <(f64 (bitconvert (i64 DoubleRegs:$src))),
(f64 (A2_tfrp DoubleRegs:$src))>,
Requires<[HasV5T]>;
def FMADD_sp : ALU64_acc<(outs IntRegs:$dst),
(ins IntRegs:$src1, IntRegs:$src2, IntRegs:$src3),
"$dst += sfmpy($src2, $src3)",
[(set (f32 IntRegs:$dst),
(fma IntRegs:$src2, IntRegs:$src3, IntRegs:$src1))],
"$src1 = $dst">,
Requires<[HasV5T]>;
// Floating point max/min.
let AddedComplexity = 100 in
def FMAX_sp : ALU64_rr<(outs IntRegs:$dst),
(ins IntRegs:$src1, IntRegs:$src2),
"$dst = sfmax($src1, $src2)",
[(set IntRegs:$dst, (f32 (select (i1 (setolt IntRegs:$src2,
IntRegs:$src1)),
IntRegs:$src1,
IntRegs:$src2)))]>,
Requires<[HasV5T]>;
let AddedComplexity = 100 in
def FMIN_sp : ALU64_rr<(outs IntRegs:$dst),
(ins IntRegs:$src1, IntRegs:$src2),
"$dst = sfmin($src1, $src2)",
[(set IntRegs:$dst, (f32 (select (i1 (setogt IntRegs:$src2,
IntRegs:$src1)),
IntRegs:$src1,
IntRegs:$src2)))]>,
Requires<[HasV5T]>;
// Pseudo instruction to encode a set of conditional transfers.
// This instruction is used instead of a mux and trades-off codesize
// for performance. We conduct this transformation optimistically in
// the hope that these instructions get promoted to dot-new transfers.
let AddedComplexity = 100, isPredicated = 1 in
def TFR_condset_rr_f : ALU32_rr<(outs IntRegs:$dst), (ins PredRegs:$src1,
IntRegs:$src2,
IntRegs:$src3),
"Error; should not emit",
[(set IntRegs:$dst, (f32 (select PredRegs:$src1,
IntRegs:$src2,
IntRegs:$src3)))]>,
Requires<[HasV5T]>;
let AddedComplexity = 100, isPredicated = 1 in
def TFR_condset_rr64_f : ALU32_rr<(outs DoubleRegs:$dst), (ins PredRegs:$src1,
DoubleRegs:$src2,
DoubleRegs:$src3),
"Error; should not emit",
[(set DoubleRegs:$dst, (f64 (select PredRegs:$src1,
DoubleRegs:$src2,
DoubleRegs:$src3)))]>,
Requires<[HasV5T]>;
let AddedComplexity = 100, isPredicated = 1 in
def TFR_condset_ri_f : ALU32_rr<(outs IntRegs:$dst),
(ins PredRegs:$src1, IntRegs:$src2, f32imm:$src3),
"Error; should not emit",
[(set IntRegs:$dst,
(f32 (select PredRegs:$src1, IntRegs:$src2, fpimm:$src3)))]>,
Requires<[HasV5T]>;
let AddedComplexity = 100, isPredicated = 1 in
def TFR_condset_ir_f : ALU32_rr<(outs IntRegs:$dst),
(ins PredRegs:$src1, f32imm:$src2, IntRegs:$src3),
"Error; should not emit",
[(set IntRegs:$dst,
(f32 (select PredRegs:$src1, fpimm:$src2, IntRegs:$src3)))]>,
Requires<[HasV5T]>;
let AddedComplexity = 100, isPredicated = 1 in
def TFR_condset_ii_f : ALU32_rr<(outs IntRegs:$dst),
(ins PredRegs:$src1, f32imm:$src2, f32imm:$src3),
"Error; should not emit",
[(set IntRegs:$dst, (f32 (select PredRegs:$src1,
fpimm:$src2,
fpimm:$src3)))]>,
Requires<[HasV5T]>;
def : Pat <(select (i1 (setult (f32 IntRegs:$src1), (f32 IntRegs:$src2))),
(f32 IntRegs:$src3),
(f32 IntRegs:$src4)),
(TFR_condset_rr_f (FCMPUGT32_rr IntRegs:$src2, IntRegs:$src1), IntRegs:$src4,
IntRegs:$src3)>, Requires<[HasV5T]>;
def : Pat <(select (i1 (setult (f64 DoubleRegs:$src1), (f64 DoubleRegs:$src2))),
(f64 DoubleRegs:$src3),
(f64 DoubleRegs:$src4)),
(TFR_condset_rr64_f (FCMPUGT64_rr DoubleRegs:$src2, DoubleRegs:$src1),
DoubleRegs:$src4, DoubleRegs:$src3)>, Requires<[HasV5T]>;
// Map from p0 = pnot(p0); r0 = mux(p0, #i, #j) => r0 = mux(p0, #j, #i).
def : Pat <(select (not PredRegs:$src1), fpimm:$src2, fpimm:$src3),
(TFR_condset_ii_f PredRegs:$src1, fpimm:$src3, fpimm:$src2)>;
// Map from p0 = pnot(p0); r0 = select(p0, #i, r1)
// => r0 = TFR_condset_ri(p0, r1, #i)
def : Pat <(select (not PredRegs:$src1), fpimm:$src2, IntRegs:$src3),
(TFR_condset_ri_f PredRegs:$src1, IntRegs:$src3, fpimm:$src2)>;
// Map from p0 = pnot(p0); r0 = mux(p0, r1, #i)
// => r0 = TFR_condset_ir(p0, #i, r1)
def : Pat <(select (not PredRegs:$src1), IntRegs:$src2, fpimm:$src3),
(TFR_condset_ir_f PredRegs:$src1, fpimm:$src3, IntRegs:$src2)>;
def : Pat <(i32 (fp_to_sint (f64 DoubleRegs:$src1))),
(i32 (EXTRACT_SUBREG (i64 (CONVERT_df2d (f64 DoubleRegs:$src1))), subreg_loreg))>,
Requires<[HasV5T]>;
def : Pat <(fabs (f32 IntRegs:$src1)),
(S2_clrbit_i (f32 IntRegs:$src1), 31)>,
Requires<[HasV5T]>;
def : Pat <(fneg (f32 IntRegs:$src1)),
(S2_togglebit_i (f32 IntRegs:$src1), 31)>,
Requires<[HasV5T]>;
/*
def : Pat <(fabs (f64 DoubleRegs:$src1)),
(S2_clrbit_i (f32 (EXTRACT_SUBREG DoubleRegs:$src1, subreg_hireg)), 31)>,
Requires<[HasV5T]>;
def : Pat <(fabs (f64 DoubleRegs:$src1)),
(S2_clrbit_i (f32 (EXTRACT_SUBREG DoubleRegs:$src1, subreg_hireg)), 31)>,
Requires<[HasV5T]>;
*/
|